A collection of post-mortems
Published on: October 3, 2022
We’ve just completed the 2022 Summer Institutes in Computational Social Science. The purpose of the Summer Institutes is to bring together graduate students, postdoctoral researchers, and beginning faculty interested in computational social science. The Summer Institutes are for both social scientists (broadly conceived) and data scientists (broadly conceived). This summer we had a mixture of in-person institutes and virtual institutes across the world. In addition to SICSS-Duke, which was organized by Chris Bail, there were 31 partner locations run by SICSS alumni.
These post-mortems describe, for each site, a) how the Institute was run, b) what each site think worked well, and c) what each site will do differently next time. We hope that this will be useful to others organizing similar Summer Institutes or future organizers of SICSS sites. If you are interested in hosting a partner location of SICSS 2023 at your university, company, NGO, or governmental organization, please read our information for potential partner locations.
This page includes post-mortem reports from all locations in order to facilitate comparisons, as well as an overview of key themes and takeaways. As you will see, different sites did things differently, and we think that this kind of customization was an important part of how we were successful.
We pull out a few high-level points of commonality and difference between 2022 site-specific post-mortems. This is a selection of highlights, not a detailed document of procedures. The sites referenced in parentheses are only a sample of the sites that may have experienced/done something, as opposed to a full inventory – these parenthetical references to sites are intended to direct towards the site-specific post-mortem documents to read more.
We’ve divided the post-mortem into 5 parts: 1) outreach and application process, 2) pre-arrival and onboarding, 3) the first week, 4) the second week (group projects) and 5) post-departure.
As the first SICSS in Poland, the SICSS-AMU/Law was quite challenging. That is because, we had to create the whole concept of the institute including a schedule, invitations, social media descriptions, graphics, and essential documentation. There were some tasks that we had to do together with double-checking, as the main description, which could be the base for further content. In the other tasks, like creating content on social media we were supported by students. However, even in advertising, there is always the kind of work that you cannot completely commit to students, as you are the most informed person about your project.
Our plan was to gather young scientists from Poland and surrounding countries through a broad advertising campaign and e-mail. Primarily, the traditional way of promotion through the University website, Facebook and some basic mailing didn’t work. That is why we had to move the deadline to at least a month later. The result of brainstorming was that we had to run a paid advertisement and do some massive mailing. Going after the other locations, we tried to keep our Facebook fan page constantly up to date with regular news on speakers, schedules, interesting articles and the team. The information about the event was also disseminated on topical groups of scientific events, conferences and seminars. These traditional methods were supported with paid advertisement, targeted according to the criteria that we declared to consider in the application process. The whole content that we created on Facebook was also used on LinkedIn, however with little success.
As mentioned, we intensified also mailing as it is the only direct channel of communication with potential candidates. For that purpose, we created a database of e-mail addresses from selected countries based on publicly available sites. It turned out to be a huge volume of work, as we decided to copy individual addresses not only these of secretaries, institutes, chairs, and other offices. Probably we couldn’t have done it without great support from students. In the effect, for three weeks we have sent around 15.000 e-mails with invitations to the Institute.
As a result, we received 50 applications ranging from China to the US and Europe. That was enough to run an institute and make some selections. In order to select participants we used the following criteria: 1) research and teaching in the area of computational social science 2) contributions to public goods, such as creating open-source software, curating public datasets, and creating educational opportunities for others 3) likelihood to benefit from participation 4) likelihood to contribute to the educational experience of other participants, and 5) potential to spread computational social science to new intellectual communities and areas of research. Further, when making our evaluations, we accounted for an applicant’s career stage and previous educational opportunities. An extremely important factor for us was the time zone. After some discussion, we decided to give a chance also to participants from Asia. Unfortunately, it turned out to be a mistake as after a few days they had to resign. A factor that was eventually hard to assess is digital skills. We encountered some applications that mentioned explicitly full possession of Windows/Microsoft Office and other software skills. This could have been an indicator that the applicant may acquire proficiency in other software quickly. However, this was not the case. Programming requires slightly different skills than those learned through extensive usage of office software.
What is important is to accept more people than planned. Primarily, we wanted to have 30 participants. Of the 35 accepted, only half confirmed participation. For that reason, we invited the next 10 people. Finally, on the first day of the institute, we had around 20 participants, with 15 that managed to finish. The other problem we had was the first deadline that we had to move at least a month because of an insufficient number of applications. It was, however, before massive mailing and paid Facebook campaign that we regard as the most efficient.
To the selected group that confirmed their participation, various resources were indicated. First of all a book by Matt Salganik “Bit by Bit: Social Research in the Digital Age” was recommended. Second, SICSS regular boot camp (videos) was suggested. Third, we pointed out that special SICSS-AMU/Law Boot Camp would be available a week before the Institute.
Although the information was spread via e-mail and the Institute’s website, it seems that the resources did not enjoy much attention. In particular, the participants did not experience extensively the exclusive boot camp that took three evenings via Zoom. There could be various reasons for that, one of them being limited possession of coding skills, as the first (of three) boot camp workshops was considerably more frequented than the next ones. It is possible, some participants were discouraged with the amount of coding to face.
As the next days showed, the participants could have better familiarized themselves with SICSS videos that covered many of the issues that came when starting to write in R. It seems, that the practical aspect of the school was not obvious for some participants that were (positively) surprised with the number of exercise and programming tasks. This, however, does not mean they were well-prepared to deal with them with ease. Perhaps, a clearer message on that issue could have helped in better understanding of the Institute and skills needed.
We encountered no abundant correspondence regarding organizational matters. As it was pointed out, we received limited response from the candidates accepted to the Institute. Among those who confirmed, there were a few that either did not show up at all or quit after day 0 (virtual get-together) or day 1. We received no contact from them. In such cases it might be good to get in touch and learn about their motives (which can be either personal or due to the Institute’s agenda). We felt, however, that those people took seats from those who wanted to participate but were refused admission due to the maximum number of participants reached.
We began the first week with a virtual get-together on Sunday evening. We abandoned a traditional introductory round, during which every person says a few words about herself/himself. Instead, we launched a Kahoot quiz and asked for their motivation, background etc. In addition, we made two short (ca. 10-minute long) speed dating sessions in breakout rooms (3 persons in each room) with random allocation of participants.
We followed the traditional framework of the two-week SICSS, meaning that the first week was a mix of lectures and group activities. The day started between 12:00-13:00 and lasted until 17:30-18:30. This was supposed to allow the participants to catch up with their professional and personal responsibilities in the morning and/or read the handouts before the group exercises. In addition, this was supposed to mitigate the Zoom fatigue.
Typically, the guest lecture lasted 1,5 hours whilst group exercises took 2,5 hours. In addition, we allocated approximately one hour for brown-bag lunches in the Zoom breakout rooms, where participants were invited to discuss their ideas for the 2nd week’s projects. The latter proved successful as we had two groups established already in the first week (the third group was established in the 2nd week). Brown-bag lunches arguably contributed to the positive group dynamic as most participants stayed online for the whole time and took the opportunity to get to know each other better. At the end of the day, we had a 30-minutes long wrap-up which was sufficient to summarize the exercises and remind about the reading assignments for tomorrow.
The main focus of our Institute was law, however, from the very beginning, we aimed to recruit participants with diverse background. This also meant that the topics, guest speakers, and exercises had to be law-related and at the same time accessible to non-lawyers. Therefore, we covered the following topics: 1) natural language processing, 2) prediction and classification, 3) online survey design, 4) experimental methods in law, and 5) plotting maps. Most participants decided to apply NLP methods in the second week, which suggests that this topic could be assigned at least one additional day. Most of the guest speakers shared their presentations and R code (posted on Slack) which were then extensively used by the participants. One of the lectures was programming-free and focused on the general issues related to computational social sciences (e.g., why a social scientist should learn programming) and turned out to be a great success for at least two reasons. Firstly, it arrived on Wednesday evening when the participants were exhausted with programming (and some of them contesting their ability to learn computational methods). Secondly, the meeting was run in an informal Q&A setting, which created a very comfortable environment.
Group exercises followed a structure where participants were split into smaller groups to work together on activities (which are all posted on our website and Slack). The exercises were programming-oriented and consisted of smaller tasks. Out of five tasks, participants were able to complete between 1 and 3, which induced us to implement the following changes starting from Day 3:
The participants welcomed all these changes.
As most participants struggled with the basics of R programming, we have been thinking about some backup solutions. We decided to introduce two optional workshops on Orange Data Mining software which attracted ca. 70% of the participants. The software is based on Python, however, it provides an intuitive GUI. Most participants utilized Orange in the second week and performed at least an initial exploratory data analysis before moving to R.
During the SICSS-AMU/Law, three groups were formed, each one realized one project. The participants mixed with each other in a very interesting way. The first group was mainly economic in nature and dealt with IMF discourse change and continuity after covid19-pandemic. The second group was of a mixed, legal-economic nature, and carried out the project Transformation of EU’s internal market. The third group was purely legal and dealt with the project Prediction of derogation of legal act basing on duration of vacatio legis.
From the substantive point of view, the topics of the projects were very interesting and well-thought-out. Basically, the research could be continued and probably could end up with successful publications, because the conclusions already presented during the SICSS were really promising.
From the programming point of view, all groups faced similar problems. Both, during the projects and during the workshops (in the first week). Participants were generally very inattentive and made a lot of typos. As a result, correctly written code did not work, which is very frustrating and, unfortunately, can be daunting. Another problem was misunderstanding the data structure and trying to write code without a plan. Unfortunately, this cannot be done. The programmer should have got a clear idea to write a program. Otherwise, the effects will be dissatisfactory. Such problems could be avoided if the participants took part in the general boot camp as well as in our extended one. Unfortunately, everyone openly admitted that they neglected this help. One solution could be to organize a mandatory boot camp for participants. After all, it is difficult to understand why the participants avoided using the materials provided to them (where the answers to most of their problems were found) and the R-documentation.
One of the strongest points of the projects was definitely the use of R at a satisfactory level by some groups. If the group did choose to use R, they did it quite well. The use of Orange can be indicated as a contrast. This is the tool that many participants have used, and this is the wrong way to go.
The groups hesitantly reached for the help of internet materials (Stack Overflow) or the extended boot camp provided by us. Definitely, this is good news, however, it was done rather inaccurately and superficially, which resulted in analyses of low quality. The ambition was also a big problem, the participants wanted more than they were able to do. However, considering that SICSS is an educational event, the confidence and ambition of the participants must be considered a success, even if projects suffered because of this.
It was clear that the closer to the deadline (Friday in the second week), the more intensified the work. Throughout the whole institute, TAs were available on duty hours every day, which did not meet with great interest. The exception was last Thursday, when there was a line of people willing to ask questions and get help. This is an area for improvement, perhaps it would be better to cancel duty hours, especially during the second week, and instead of this, make TAs supervisors of groups. One TA for one group.
The collaboration within the groups was particularly good. Members used two channels of communication. Most of the discussions took place via Slack chat. Sometimes, when needed, the participants organized video meetings at Zoom.
Using Slack as the main communication channel was a very good idea, even the free version was sufficient. The second channel of communication, more direct, was Zoom. To simplify usage and avoid problems, we organized every meeting and duty hour under the same link. Maybe one software for both chat and video meetings is a nice idea for improvement.
Participation in lectures from other locations was minor. The curriculum of SICSS-AMU/Law was very intense, so attending other lectures could be too hard. Nevertheless, such possibility requires more advertising.
The Institute was concluded in a friendly atmosphere with a noticeable vibe of success among those who made the biggest progress coding-wise. We encouraged the participants to continue using Slack as a platform to further develop their ideas, share information about grants or projects and network. We also suggested joining the SICSS world community to reach out to the broader group of scientists who apply computational methods in their work. We provided participants with personal certificates of completion of the Institute.
A good idea to monitor the execution of the Institute was to run short online surveys on Google Forms (anonymously). The participants were eager to answer them which helped make changes in week 1 as well as evaluate our work throughout the whole event.
According to the three surveys executed during the Institute, all presentations given by invited speakers enjoyed high or very high scores from participants. Despite the whole event being online, we noted that students “definitely made some connection with other participants” (70% of answers). A similar percentage declared that the schedule of the classes on a daily basis was well planned (67%) while some argued that they should start either earlier or later during the day. A very promising number of the participants (85%) responded that they loved working on a group project and even more of them (92%) declared that they are definitely going to further explore computational methods in their work. This surprisingly corresponds with answers we received after the first days of the Institute. When asked about the technical challenges of coding (including both very code writing as well as installation of various software and libraries) students’ opinions were divided. Around 36% of participants declared they clearly expected such challenges. Nearly as many said they did not really expect that many technical difficulties, but expected some. There were single voices that having to deal with such computer challenges was a surprise. This all corresponds with an issue clearly brought in the interim survey, being little coding skills and not enough support from the institute regarding coding fundamentals, according to some of the participants. As it was suggested earlier, a clearer approach should have been taken by our organizing team, i.e. whether we take responsibility and teach the absolute basics or we assume that it is up to the participants to bring certain computer skills to the table.
We accepted with gratitude the feedback from participants who kept saying that there were two best things about the Institute: people (open-minded, enthusiastic and cooperative) and the possibility to grow (learning new research methods, and programming). We note that this was possible not only thanks to the support and funding from various institutions and thanks to highly-involved participants. We believe that this was achieved also by a committed group of organizers who looked out for each other and worked as one team, often beyond “working hours”.
To design the program, we relied on the experiences of our two main organizers (Jana Lasser and Ivan Smirnov), who had either already organized a SICSS location (Ivan, HSE, 2021) or participated in one (Jana, Helsinki, 2021). We designed the programme alongside our strengths and research interests as computational social scientists and invited external speakers to bring diversity and fill in gaps. This included lectures offering a critical perspective on the state of the field of CSS, publishing CSS research and common pitfalls in CSS research projects as well as hands-on lectures on how to use APIs to collect data, how to run large-scale surveys on Amazon Mechanical turk and how to leverage large language models for CSS research. These lectures were complemented by two lectures by our senior organizers: an introduction to networks for CSS research from Markus Strohmaier and a lecture on how to combine large-scale surveys with digital trace data by David Garcia. We were also fortunate to welcome Suhem Parack to give us an introduction to using the Twitter API v2 to collect data for CSS research. Since both main organizers mainly work with Python, we decided to commit to teaching our SICSS location completely in Python, aiming to contribute teaching materials in Python to the collection of teaching materials of SICSS.
Our SICSS location was funded by a funding programme of TU Graz, dedicated to fund online teaching activities that are collaboratively hosted by two universities - in our case RWTH Aachen and TU Graz. We, therefore, planned our programme to take place completely online. We designed the outline of our programme similar to other SICSS locations, planning for one week of lectures and hands-on exercises, followed by one week of group project work with talks by guest lecturers in the evenings to have a common event every day.
We specifically advertised our SICSS location within our two host universities TU Graz and RWTH Aachen, since we wanted to attract students from the Computational Social Systems master’s programme at both universities. We, therefore, advertised the summer school in short pitches in relevant lectures during the winter semester and through mailing lists of lecturers and students of the master’s programme. In addition, we advertised our location over social media channels (Twitter and LinkedIn) as well as over the main SICSS website.
We attracted a total of 71 applications. Students had to submit an academic CV and a motivation letter (1 page), describing their motivation to participate in our SICSS location and the potential benefit for their study and research career. We required no prior programming skills and no specific disciplinary background to apply. Students needed to have at least a bachelor’s degree to apply to our location. We also decided not to ask for recommendation letters. We believe that recommendation letters impose a disproportionate burden on students while there is little evidence of their usefulness.
We aimed to admit 40 students, with guaranteed spots for students from TU and KFU Graz and RWTH Aachen in case they meet the minimum requirements and submit a complete application. We received a total of 5 applications from TU Graz and 3 from RWTH Aachen. The rest of the applications came from students enrolled at other universities. We also received a number of applications from PostDocs or people that had left academia. Given the choice of admitting a student or a non-student with similarly strong application materials, we decided to admit the student, since we felt they would benefit more from our summer school. We also paid special attention to applicants from groups that are underrepresented in CSS.
We invited a total of 40 students to participate in our SICSS location. The disciplinary backgrounds of the people admitted to our location were very diverse, including students from political science (7), computer science (6), social science (5), psychology (5), computational social science (4), economics (3), collective intelligence (2) and physics, statistics, linguistics, data science, geography, engineering, journalism, and genders studies (1 student each). Similarly, the country of residence of our students was very diverse, with the rather high prevalence of German (8) and Austrian (5) students reflecting our increased outreach efforts in these countries, followed by the US (5), Morocco (3), UK (2), Russia (2) and Italy, South Africa, China, Portugal, Denmark, Switzerland, Argentina, Turkey, Korea, Belgium, France, Serbia, Greece and Iran (1 student each). Gender identities were almost balanced, with 22 participants identifying as male, 17 as female, and 1 as non-binary. Rejection rates were similarly balanced, with 52% of the rejected participants identifying as male, 45% as female, and 3% as neither.
We did not require any previous (Python) programming knowledge to enroll in our summer school. To get everybody up-to-speed regarding programming in Python, we offered a voluntary two-week Python crash course in the month before our summer school. The course consisted of a curated collection of online learning materials, tailored to fit the needs of participants in our summer school with introductions to general programming in Python (first week) and introductions to the libraries pandas, networkx, statsmodels and matplotlib in the second week. The Python crash course was supervised by two of our teaching assistants, Moritz Erlacher and Florian Hofer, who contributed a great deal of experience in teaching beginner-level programming courses in Python. Since we were aware that our participants had other study or work commitments in the month leading up to the summer school, we designed the Python crash course to be mainly self-paced learning with videos and exercises in Jupyter Notebooks. We offered office hours in the early mornings and late afternoons to help with questions regarding the exercises. The crash course was greatly appreciated by our participants and helped to equilibrate the level of programming skills across our participants at least a bit.
In addition to the Python crash course, we pointed students to the online learning materials provided by SICSS and asked them to apply for academic access to the Twitter API. We provided a guide on how to apply for academic access within the scope of our summer school, which was greatly appreciated by students. A total of 33 students applied for academic access and 6 of these applications were denied for unknown reasons.
During the instruction week, we usually started with one lecture of 60-90 minutes at 9:30, followed by one hands-on exercise or group discussion for 90-120 minutes. We made sure to provide breaks of 10 minutes every 45-60 minutes since student feedback indicated that in the online setting it was often not easy to follow instructions for extended blocks of time. After a lunch break of 60 minutes, we had another lecture of 60-90 minutes, followed by a hands-on exercise of 90 minutes and a group discussion of 15-20 minutes to wrap up the day. We asked students for feedback every day, asking them “what did you like about today’s sessions” and “what could be improved”? We received very helpful comments and ideas on how to improve online collaboration from the students, which we tried to implement on the following day. Students indicated that they greatly appreciated that their feedback was promptly incorporated into our location’s organization. We also received a lot of very positive feedback, which was very uplifting to read when preparing the sessions for the next day.
On Friday afternoon we hosted a brainstorming session of 2x60 minutes in random groups of four students each to generate project ideas. Students had to provide an idea title, short description, and responsible person for follow-up questions. This generated a total of 34 ideas. Students could then indicate their interest in working on one of these projects, with the aim of forming project groups of 3-4 students each for the project work in the following week.
We matched people into groups over the weekend between the instruction and project week and students could start working on their projects on Monday morning. We asked them to self-organize their working schedule during the project week, indicating that we expected approximately 5 hours of project work every day. The matching resulted in a total of 10 project groups. We split the supervision of the project groups between the two main lecturers such that each lecturer was responsible for supervising 5 groups. We asked groups to have daily check-ins of ~20 minutes every day. During these check-ins we met the groups in a video chat and helped them with a clearer definition of their research objectives and methods as well as methodological issues. In addition to the daily supervision check-ins, we provided programming support by teaching assistants, who offered office hours in the mornings and afternoons to help with programming questions. Projects were presented on the last day of the summer school. Every group had 10 minutes to present what they worked on plus five minutes of questions.
Guest speakers We invited a total of 8 guest speakers to host lectures, talks, and tutorials during our summer school.
When inviting our guest speakers, we tried to showcase cool research by young scholars in our field, as well as provide lectures on foundational aspects of CSS research by established researchers that complemented our lectures. We made sure that each guest lecturer had ample time for Q&A with the students and enjoyed lively discussions after the talks.
Next to the lectures, exercises, and project work, our student assistant Alina Kopkow organized voluntary social activities to give students the opportunity to get to know each other and the lectures in a non-work context.
On the Sunday evening before the summer school started we organized an informal meet-and-greet with crowd-sourced questions. This event was appreciated by our students and about half of the registered participants showed up.
On the second day of the summer school we organized an online pub-quiz where students could win a copy of either “Bit by Bit” or “Everybody lies”. We had three student teams with three to four students each competing for the prize. The activity was a lot of fun but we felt that more online activities after a long day of online lectures and exercises are not optimal.
At the end of the first week, we organized a hybrid treasure hunt in Graz and Aachen with “agents” on the ground walking through the two cities to find clues and online teams to solve riddles. Only two participants showed up for our treasure hunt and we, therefore, scrapped the group in Graz and only ran the one in Aachen. While the concept of the treasure hunt worked and is a good way to show online students around the city it was definitely too much after a full week of online lectures.
In addition to these activities, we organized a Twitter list for all participants and lecturers of SICSS Aachen Graz to follow each other and network after the summer school. This was greatly appreciated and almost all of our participants are part of this list.
We received a grant of 15,000€ from TU Graz to organize a “joint online course”. We used this money to hire four student assistants to help with the python programming crash course, python programming support during the summer school, and the organization of social activities. In addition, we paid honoraria of 500€ per lecture to our external speakers from these funds. Since our location was run completely online, we did not need funds for student accommodation or food. Initially, we wanted to compensate our two main organizers for their efforts as well, but this was not possible with these funds.
Our main communication channel during the summer school was the SICSS Slack platform. There, we created one main channel for our location, and two additional channels, one for programming support and one to collect links to useful information. In addition, we sent all important information and reminders for lectures by invited speakers over email.
Our course was organized completely online to accommodate students from different locations. While this was very useful to attract many students from different locations it also resulted in some challenges regarding teaching and social interaction. We noticed early on in the daily student feedback that students needed many breaks to be able to follow lectures in an online setting for an extended period of time. We ended up introducing 10 minutes breaks every 30-40 minutes to help students withstand the “zoom fatigue”.
We tried to offer a number of social activities around the core course activities. We noticed that participation in these activities was very low. This can probably be attributed to the fact that social activities were also purely online. Students were just exhausted after 6+ hours of online instruction every day and spending another 1-2 hours in front of a screen – even for social activities – was not very appealing to them. As for learning, since we think the social and networking aspects of the summer school are very important, we will probably integrate the social activities better into the course itself the next time we run it, at the expense of lectures and exercises.
We were conscious of potential issues with commitments to attend the summer school because we have had experience with low attendance in online teaching settings. As a countermeasure, we asked students to re-confirm their participation in the summer school after the initial round of invitations to attend the school at the end of the application process. We also kept a “waiting list” of students that did initially not make the cut to attend but had submitted high-quality applications. Six out of 40 students did not confirm their participation and we filled the spots with people from the waiting list.
As expected, offering the course as a free elective course in a purely online setting without any attendance fees created a lack of accountability for students to actually attend lectures. As a result, three students did not show up for any of the lectures without offering an excuse. Two students did not attend but excused themselves, citing unexpected child care obligations or sickness as reasons. Another two students did not attend and excused themselves, citing other work getting in the way of attending the summer school. The latter would probably not happen in an offline setting where people actually need to book trips and accommodation to attend the summer school, because this would create at least some level of commitment. In summary, 33 out of 40 students (83%) that signed up for the summer school ended up attending the majority of lectures and completing group projects.
Lastly, hands-on exercises and group projects involved a lot of programming. Providing students with programming help where needed was challenging in the online setting. Sharing screens and typing commands in a chat works to some extent, but it would be much easier to help students if lecturers could interact with them in an offline setting and help them on their laptops rather than through video chat.
Many APIs return JSON as a response format. To be able to easily work with the data, it first needs to be converted to a tabular format. While we taught students about how to handle pandas data frames in the Python crash course, we did not properly teach them how to handle data in JSON format and how to convert JSON to data frames. This created a number of problems and delays when students had to work on their projects and a lot of demand for programming support during the project week. A learning from this is to provide a dedicated tutorial on how to handle data in JSON format in the future, and/or to provide dedicated functions that convert JSON to well-formatted CSV files for a number of APIs.
One surprising observation was the homogeneity of group projects by our participants. All of them used Twitter as the main source of data. All of them applied sentiment analysis as the main or secondary method for analyzing data. Almost all of the projects were related to politics. This was despite our efforts to show during the lecture week how broad the CSS field could be and despite the fact that many other APIs were introduced. However, the Twitter API was in the focus of some of our lectures and exercises and it seems that using it was the easiest option for our participants.
It could be that using the most familiar API was the only way to ensure that participants would be able to get some results given the time constraints of a project week. On the other hand, we feel that we didn’t pay enough attention to teaching participants how to choose a good research question and how to choose data sources/methods that would be the most appropriate to answer this question. Some participants mentioned that coming from a computational background they would like to learn more about social science, to learn how to choose a question, etc. Given the time constraints of a project week, we think the only solution would be to introduce a “proposal track” for projects in addition to a “practical track”. Creating a good research proposal would require work on selecting a good research question, exploring the relevant literature, thinking about the best data source and methods. However, properly done, these activities would easily consume the whole project week and that wouldn’t allow participants to try to get some data and perform preliminary analysis. We are not sure what would be the best compromise between providing a practical experience and devoting more time to general research competencies.
We have received overwhelmingly positive feedback after the school with participants emphasizing not only the quality of our content but also the special atmosphere of the school. We are not sure how to reproduce this because we don’t feel that we have done anything special and we are surprised by the positivity of feedback ourselves. However, we think that the fact that the school was run by two postdocs (not professors) might have contributed to that. This significantly reduced the distance between participants and lecturers and allowed us to openly talk about academic life, share our struggles, and critique the status quo. This openness seems to help to create a “we” feeling at the end of the school — as one participant put it in their feedback.
Principle Organizer: Kat Albrecht
SICSS Atlanta 2022 was held at Georgia State University in downtown Atlanta, specifically in the Andrew Young School of Policy Studies. This was a substantial undertaking since I had only arrived at GSU this academic year and did not have any connections on the campus or logistical knowledge of university policies. These barriers were overcome to have, what I consider, one of the best SICSS I’ve been part of to date (I had previously been organizing SICSS Chicago since 2018). We are also extremely proud to be, to our knowledge, one of the first partner sites to run in the Southern U.S. and were generally thrilled with the community support from our geo-co-located universities. We had participants and faculty speakers from a large number of area institutions and lots of interest in continuing SICSS ATL next year.
I organized the institute with the assistance of Cynthia Searcy, Georgia State’s Associate Dean for Academic Innovation. The Dean of the Andrew Young School of Policy Studies, Sally Wallace, generously essentially excused Cynthia from other duties during SICSS so she could help me. As will become a theme in this report, the support of GSU was absolutely critical to the success of the institute.
The section to follow concerns organizational processes and pre-arrival materials.
SICSS ATL received $6,000 from the main site and over $10,000 from GSU. Money was spent in four key areas: TA stipend, meals, Prolific survey costs, and group project funding.
The budget request from the main site was originally closer to 12k, but I limited it severely at their request, and I do have some personal regrets about doing so. I did not take an honorarium for organizing the institute and did end up spending nearly 2 thousand dollars of my own money on flights/daily transportation to fully engage with the institute. In the past, the organizer honorarium defrayed these expenses. It was particularly difficult for me to do this as a first-year faculty member who was new to 9-month contracts. I did so because I was worried that we wouldn’t get funding for the institute at all but was a little disappointed to hear that honorarium were still given to some sites but not others. Next year, I will request honoraria because 2 of the co-organizers are PhD students, and I would not ask them to give their time without any compensation.
Cynthia and I met weekly from December through the summer institute, making sure all the logistical pieces were in place. Cynthia was enormously helpful during this process, since she had ready access to university listservs, knowledge of the campus and its vendors, and had more connections with area faculty than I did.
We largely followed the procedures I previously used to organize SICSS Chicago for a number of years to good effect. We created a Gmail account for SICSS ATL communications, with the intention of being able to hand it off to any Atlanta-area institution that may wish to rotate in teaching SICSS.
We opted for a high-effort application process to avoid attrition. We asked that participants give a research statement, a CV, a writing sample, and asked graduate students to provide contact information for a reference. We received 40 applications that we considered complete and plausible. We selected around 23 of these applications and with attrition had 20 participants complete the workshop (1 was no longer able to attend due to injury, 1 was not comfortable participating at the institute due to Covid-19, and 1had their partner test positive for Covid-19 during the institute and could not complete it). Participant recruitment focused on Atlanta area schools broadly construed, but participants came from Georgia State University, Georgia Tech, University of Massachusetts Boston, University of North Carolina Charlotte, the Georgia Policy Lab, University of Georgia, Yale University, University of Berkeley California, Mercer University, and Emory University. Participants were a healthy mix of professors, postdocs, and advanced graduate students.
We offered the usual suite of pre-arrival materials for the institute. There was evidence that many of the participants did read Bit-by-Bit as requested, but participants generally did not make use of TA office hours or use the coding tutorials extensively before the institute. Consistent with previous years, 1-3 participants contacted the teaching assistant, Olga Churkina, for help before the institute. However, participants did use the materials provided on the SICSS website during the institute, often mentioning reviewing them the night before a new topic or using them while working on their group projects.
We opened the institute with an outdoor reception at a local Atlanta eatery, where participants could optionally socialize before the first official day of the institute. Most participants attended this event, which seemed to be very positively received.
We held the institute in the Andrew Young School of Policy Studies, in a 12th floor room with a skyline view of Atlanta. The space was very well laid out for both collaboration and catering, with the advantage of making it clear that the institute was valued and prioritized by GSU. Similarly, our final presentations in Week 2 were held at the Curve, a focal feature of the GSU libraries.
We did not use any video content in SICSS Atlanta, instead opting to teach all of our content live. We felt this was a necessity for us in running an in-person institute to differentiate our content from something that could be accomplished virtually. I acted as the principal lecturer for the institute, giving most of the content lectures. While a lot of work, this effort did produce slide decks that can be used by next years’ organizers if they so choose. I focused on nailing the basics that participants would need to understand to fully engage with the research lectures brought by our guest speakers. We taught content across the institute in both R and Python due the stated preferences of our participants.
We had a healthy variety of local speakers with various expertise. Speakers at the institute included Charlotte Alexander (Georgia State University), Omar Isaac Asensio (Georgia Tech), Lauren Klein (Emory University), Jesse Lecy (Urban Institute and Arizona State University), and Dror Walter (Georgia State University). Speakers all gave research talks, and some (Alexander and Asensio) gave additional workshops on natural language processing and web scraping.
We ran the institute from Tuesday-Saturday due the Juneteenth Holiday and covered the SICSS core curriculum. This was a later-stage change, and we would prefer to not run the institute on a Saturday next year, since it was more difficult for local attendees to be available. We did the full iteration of each group exercise in the afternoons, with consistent success across groups. On occasion, some groups did not finish entirely, but we were fine with this, since it was still a strong learning experience. We deviated from the SICSS main curriculum in offering a different activity for digital surveys, just as we did at SICSS Chicago. In our version of the exercise, participants generate research questions, build the surveys, collect pilot data, and analyze that data to produce visualizations. All project groups were successful in completing this ambitious exercise.
We used a modified version of the design sprint activity to facilitate group formation for Week 2. I opted to run the sprint on Easy Retro, a free, no-registration platform where individuals invited to the board can upload cards with their ideas. I had participants form random groups and generate project ideas for 25ish minutes. They added a description of those ideas to the Easy Retro board. I then shuffled the groups and again had them ideate for 25ish minutes. Following this, there was a 10-minute period where any individual could add an idea to the Easy Retro board. Once both sets of group ideas and any additional individual ideas were posted, participants were given 6 votes to vote for projects they were interested in. The top voted projects were then assigned tables in the room where participants could meet and discuss them and develop them further. 5 project groups emerged out of a combination of this exercise and earlier group activities and discussions from the week.
To close the institute, groups presented for 20-25 minutes on their work so far. All groups gave excellent presentations, which were attended by a number of stakeholders (including the Dean and others) in the Andrew Young School of Policy Studies. The presentations are linked on the SICSS ATL schedule, along with other materials from the institute.
Week 2 also featured some optional social events, including a hike on the weekend and a closing happy hour at a nearby restaurant. Participants, especially non-local participants, responded positively to these events.
After the close of the institute, we conducted the SICSS ATL local grant competition. We received 4 applications from 4 groups of SICSS ATL participants. We opted to fund two of those proposals, one studying the rhetorical and social constructions around Juneteenth as a holiday, and the other a survey experiment about school gun violence. Both projects received money to run surveys related to their goals and promised money to defray publication fees. We applied several criteria to getting the money: a timeline in which it must be spent, that the project has IRB approval, that the project be pre-registered, and that the data be made public if it is ethical to do so. Projects that were not awarded received substantial feedback and encouragement to carry on, regardless.
A new feature we were happy to provide to participants this year was a SICSS 2022 LinkedIn badge that is easily updatable in the future, should any other sites/years want to use it.
I am already in progress with plans for SICSS ATL 2023. I am tentatively holding next year’s institute at Georgia Tech, with the help of the TA from SICSS ATL 2022, Olga Churkina, (who is dual enrolled at both GSU and Georgia Tech) and a Georgia Tech PhD student participant, Ryan Ellis, from SICSS ATL 2022. We are currently in the planning phase, building an advisory board of Georgia Tech faculty to assist in our organization efforts. I expect I will be rather involved in the daily runnings of the institute next year as well because it’s still early, but my long-term goal is to have Atlanta institutions trade off holding it with less involvement from me to maximize involvement across schools.
Our current planning priorities also include securing funding for the operations of SICSS Atlanta 2023. We have some positive interest from one Georgia Tech faculty member, Omar Isaac Asensio about sponsoring group projects in particular, which we are hopeful about. We expect that to be the most difficult financial resource to procure, as it is every year, because it is logistically complex.
Participation was restricted to Ph.D. students, postdoctoral researchers, and untenured faculty within 7 years of their Ph.D. Nonetheless, there were no restrictions based on citizenship, country of study, or country of employment. The idea was to invite about twelve participants. Candidates were evaluated along different dimensions: 1) research and teaching in the area of computational social science 2) contributions to public goods, such as creating open source software, curating public datasets, and creating educational opportunities for others 3) likelihood to benefit from participation 4) likelihood to contribute to the educational experience of other participants 5) potential to spread computational social science to new intellectual communities and areas of research. To apply for the summer school, applicants had to submit the following documents: CV (max 2 pages), research statement (max 1 page, single-spaced), project proposal (max 2 pages, single-spaced), a writing sample (no more than 35 pages).
Pre-Arrival and Boarding
On May 10, we sent our decision letter to the admitted applicants. To confirm their participation, we asked them whether they could fully attend our event. A total of 14 participants were selected. Due to the withdrawal of two students, we ended with a total of 12 candidates. Once the admission procedure was over, we asked the participants to provide the organizers with a short biography and a picture to be included in the web site. During the weeks before the SICSS, we sent the participants several e-mails with detailed explanations regarding the schedule and the overall organization of the summer school.
Developing the Program
We planned a 2-weeks program, with a first week fully online, and a second week with activities organized in the classrooms provided by the Department of Political Sciences of the University of Bologna (UniBo). Both weeks have been organised with the help of 3 Teaching Assistants with specific technical skills in Statistical Methodology, Text Analysis and Informatics. For development of the 2-weeks program, we wanted to balance two types of activities: lectures from selected experts and interactive learning through coding. In the first online week we planned to engage guest experts to cover all the core topics in Computational Social Science. Lectures on the first week covered data collection, text analysis, digital surveys, digital experiments, and ethics of research. The second week was focused on guest researchers presenting larger CSS projects of in both private and academic environments. Interactive laboratories have been tailored on the curricula of the admitted students. We took into consideration many possible activities, for example involving Network Analysis and Crowdsourcing (Mechanical Turk, etc.), but we noticed that for many students the primary interest was in Text Analysis. The laboratories have been delivered fully by the 3 TAs. We offered two online parallel tracks. In the Basic track, students have been introduced to coding Text Analysis in R with tidyverse, Twitter API, rvest, and tidytext. In the Advanced track we offered 2-hours classes on GitHub collaborations, advanced API (Wikipedia, Youtube) and inferential design. Most of the technical content provided has been coded in R. In the week in Bologna, interactive laboratories have been planned to prompt discussion of methodology of research projects proposed by students. Teaching Assistants has been planned to stay at disposal of the students to provide first-aid support with coding. For us, it was important that all the students had at least basic experience with R, so we pushed the students to self-train on R Studio with SICCS Online materials. We tested the admitted students on their training, and we found that the overall level of the class was sufficiently good to approach the packages in the Basic track. We also noticed that approximately all the students scored the same mark at the test, and that they failed only the most advanced questions of the test (for example, on R Markdown internal options). Since these advanced questions were not covered at SICCS Online materials, we have the strong feeling that students actually engaged themselves with pre-School training. We set up a Slack channel within the SICCS environment, to pin all the updates and the relevant information for the students. We also included social activities for the second week in Bologna, with one covered dinner and one covered aperitif. The University of Bologna provided spaces of socialization. We gifted one thermal bottle and one SICCS and UniBo t-shirt to each student at the end of the School.
The first week consisted fully of online activities. The week has been scheduled to provide to students the perception of a technical progression. Students were mostly coming from a background in Humanities, so we expected them to be more confident to humanities and data visualisation than statistical methodology. On Monday, at 9 AM the Dean of the Department of Political Science of UniBo welcomed the students, and we run a general round of ice-breakers short presentations among the presents. The first lecture started on 10 AM, on the shift towards new CSS paradigm in the Ethics of Social Research. Laboratories in the afternoon touched two introductory topics. The Basic track covered pipeline programming in Tidyverse and ggplot. In particular, we showed to students functions as facet_wrap() for elegant multiple plots. The Advanced track guided the students to set up their Github account to R Studio and how to manage most of the functions of Git language for repository management. In this day, students spread equally to both Basic and Advanced tracks. From Tuesday to Thursday, we kept a lighter schedule, with Lectures starting at 10 AM until 12 AM, and Laboratories starting at 2 PM. Lectures involved often watching videos together, often with a comment from the guest lecturer. In some cases, lecturers asked to students to participate in a online activity. The response from students to online activities was good. Usually, laboratories lasted from 1 hour to 2 hours. Only the first hour of the laboratory was mandatory for the student, while the second hour was dedicated to specific applications of the covered techniques. In these days, most of the students backed from Advanced track to the Basic. The Advanced track provided different skills for advanced API and concluded the mini course with a discussion on causal inference. The Basic track showed to students how to scrape reviews from Amazon with rvest, and then use the scraped scores to validate lexicons for Sentiment Analysis. As an example of Topic Modeling, Latent Dirichlet Allocation has been performed on R. Students actively participated in the laboratories, for example sharing their code on Zoom. Teaching assistants provided remote technical assistance to students at any hour of the day. On Friday, no laboratory has been scheduled, but guest lecturer Chiara Binelli provided an extensive online activity on online experiments.
The second week of SICSS was held in the traditional format for the pre-pandemic world - in-person. The Department of Social and Political Sciences at the University of Bologna hosted activities during the second week. On June 13, the first day of the offline week in Bologna, we held a welcome dinner.
The line-up of speakers included foreign and Italian practitioners and academics working in the field of computational social sciences. On June 13, Chris Albon, director of machine learning at the Wikimedia Foundation, spoke to the participants of the institute. On June 14, language technology educator Rachael Tatman presented the latest achievements in Natural language processing. The participants were interested in aspects of working outside of the academy with a background in social sciences, cooperation with technical specialists in the implementation of joint projects, incorporation of the theory about human behavior from social sciences in data-driven initiatives, etc.
Italian demographer Nicola Barban presented how digital data is used in contemporary demography. Annalisa Pelizza, in collaboration with Wouter Van Rossem, highlighted the importance of database structures operated by the EU states in profiling and framing migrants contacting with the European jurisdiction. Their lecture was accompanied by a short group exercise in network analysis.
During the second week, participants presented their own projects, after which they had time to work on them, receiving advice from teaching assistants on issues related to data processing, the use of appropriate statistical tools, and further improvement of research design.
We realized that the formation of a schedule of hands-on sessions for two groups - basic and advanced, should not imply simultaneous classes. A number of topics covered in the advanced group were also relevant to the research projects of participants who chose to attend classes in the basic group. In the future, it would be better to draw up a schedule with consecutive classes.
The closing of SICSS Bologna was accompanied by a farewell aperitivo. We also made a gift in the form of a t-shirt and a bottle for all participants, speakers, and organizers.
SICSS-Covenant was hosted by the Centre for Economic Policy and Development Research (CEPDeR) at Covenant University, Nigeria, from 19 to 29 June 2022. The 10-day Summer Institute in Computational Social Science (SICSS) dubbed SICSS-Covenant 2022 was the first ever SICSS to be held in Nigeria and the ECOWAS Region. It was also one of three SICSS sites in Africa in 2022. This post-mortem on SICSS Covenant 2022 is divided into five main sections: 1) outreach and application process; 2) coding bootcamp 3) pre-arrival and onboarding; 4) the summer institute; 5) closing ceremony, departure and post-departure.
We decided that our summer institute would focus primarily on the Sustainable Development Goals (SDGs), and we sought to explore how computational social science methods could contribute to better tracking of the progress towards the Sustainable Development Goals in African countries. We engaged two teaching assistants: one had prior experience with coordinating large academic events at Covenant University, while the other had a computer science background with significant coding experience. In choosing these two, we considered gender, their unique talents, and their potential to enrich the programme and effectively manage the diversities of the selected participants.
With regards to recruitment drive for participants, we solicited applications with several intentions. Chief among this was that we wanted a diverse yet cohesive group of participants in terms of skill sets and interests. We were also conscious that there were only two other sites in Africa and, as a result, were willing to consider and partially fund six applicants from other African countries. We leveraged our networks as alumni of SICSS-Cape Town to reach out to colleagues to spread the word among students and early careers in their network. We also tried to reach potentially interested participants through emails to faculty that we thought might know interested participants. In addition, we targeted other top African universities (the University of Ghana and the University of Witwatersrand) and hosted a post-presentation information session with their students and early career researchers. Also, the CEPDeR fellows’ network across African countries and Covenant University platforms were utilised as the call for applications was showcased on their websites. Lastly, we placed advertisements in the newsletters/forums of the African Population Studies and AuthorAID, and announced the call for applications on our personal websites and other platforms like the R-blogger.
All applications were managed through a customised Google Form that sought information about participants’ evaluation of their own proficiency in R-programming and Python in addition to basic socio-demographic questions. We had a lengthy discussion about whether to accept letters of recommendation, particularly given earlier reports of their limited usefulness at previously held SICSS at other sites. Ultimately, we decided to request for recommendation letters primarily to filter the pool of applicants and focus on those who have carefully read the programme information and are likely to participate fully in the programme. We reviewed our pool of applications frequently and intensified our recruitment efforts targeting specific groups as the need arose.
We believe that the combination of our recruitment methods worked quite well because of the overwhelming number of applications we received despite this being our inaugural SICSS. We received 104 applications, mostly from Nigeria (75%) but also from other African countries, including Ghana (3), South Africa (3), Kenya (2), Zambia (2) and others. Our applicants were also very diverse in terms of research expertise, mostly from Economics (33), Sociology/Demography (32), but also from Computer Science (10), Geography (6) and others. The gender distribution of our applicants shows a higher proportion of male compared to female applicants (Fig. 1).
Fig. 1: Gender Distribution of Applicants by Qualification
Three reviewers comprising one Covenant University faculty, a TA, and an organizer assessed each application independently. During the review process, we considered the diversity of research interests and whether there could be synergy in research collaborations. We also carefully reviewed the applicants’ motivation for participating in the programme and re-evaluated the role of recommendation letters. Although we discovered that the letters were not critical for our evaluation as they did not offer additional information than we obtained in other application materials, we hope to still request for letters of recommendation to limit the number of applications we have to screen in the future.
We then sorted applicants according to the average scores from the three reviewers. During the application review phase, we discovered many participants overestimated their proficiency in R/Python. In contrast, we realised that many of them only had some quantitative and statistical background but with very limited coding experience. We believe this issue cuts across Africa, especially because the computational social science (CSS) community in Africa is not as strong as in many other parts of the world. Based on our own evaluation of the applicants’ coding experience, we decided to follow a two-stage selection process in which applicants were first conditionally accepted and only given an unconditional offer once they had fully completed the coding bootcamp and demonstrated some proficiency in R-programming. This decision significantly reduced our pool of applicants to 29 conditionally accepted applicants who were enrolled in the coding bootcamp and 16 applicants who were offered an unconditional acceptance and ultimately attended the summer institute. Due to funding constraints and high regional flight costs, particularly from Malawi and Mozambique, three participants were invited but could not afford the flight cost to the summer institute even with partial funding.
After conditionally selecting our participants, we decided that our primary programming language would be R, given the relatively few numbers of conditionally accepted applicants who were familiar with or interested in using Python. We did not emphasise one programming language over the other during our outreach activities but think that our team’s collective area of expertise may have influenced the pool of applications we received. In the future, we hope to partner with more colleagues in Computer Science to enhance the representation and mutual benefit of both programming languages, even more so now with the integration of R and Rstudio (now Posit).
Fig. 2: Screenshot of DataCamp Leaderboard with our Participants’ Progress
We received generous support from DataCamp to access over 360 interactive courses through DataCamp for the Classroom. This support lessened the burden of a virtual bootcamp on the organizing team while also providing a valuable opportunity to track our participants’ progress (see Fig. 2). We assigned a total of 19 courses covering a diverse range of topics (data wrangling, data analysis, data visualization and an introduction to programming) in R to the participants to complete over five weeks. Each course requires about four hours to complete, but the completion time could vary depending on the participant’s proficiency in coding. For most, this was manageable (particularly students), but for some, this was quite a lot to juggle amongst other work, family-related demands, and other competing activities at the time. We did not anticipate the burden these courses would have on our participants and the impacts on the attrition rate. In the future, we will allocate more weeks for the participants to complete the courses and conditionally accept more applicants. The teaching assistants supported our conditionally accepted applicants. We also created a dedicated channel #sicss-covenant on Slack and asked the participants to reach us there if they required assistance with any aspect of the coding bootcamp. Participants did not access TA/computational help resources as often as we would have expected. We hypothesize that this may have occurred for a couple of reasons. First, the DataCamp virtual environment made learning fun, and there were hints for each exercise. Second, participants were hesitant about being the first to seek help in an open forum. In the future, we might ask participants to share the most challenging coding exercise that they managed to solve on their own and their winning solution on Slack. We think this may work well, given that participants are more willing to share their successes than failures. Also, knowing where they struggled and how they solved it may be helpful for other participants. We tracked the participants’ progress every week. We sent targeted emails based on whether they were on track (likely to complete the week’s exercise), stalled (yet to start the week’s exercise) or not started (not completed their registration on DataCamp). We also think the DataCamp leader board may have fostered healthy competition among our participants.
About three weeks to the start of the summer institute, we reviewed our applicants’ progress and unconditionally accepted those who have completed a significant number of courses. After the participants were unconditionally accepted, we begun to onboard them into the program. Our onboarding process included several steps, including gathering participant information and populating the people page. We shared a link to a Google form with the participants to fill out the “onboarding survey” to provide bio information for our website. We provided the participants with links to important sections of our website, including the schedule (with reading materials) and the people page (to know who is coming to SICSS-Covenant). We also included a link to the pre-arrival section of our location’s webpage and shared travel guidance to Covenant University and Nigeria (for international participants). We communicated with the participants primarily through email but also via the Slack channel. Due to the popularity of WhatsApp in our setting, our TAs also created and managed a WhatsApp group for emergency/informal updates.
In designing and developing the program, we relied on our collective experience as participants and a teaching assistant at SICSS-Cape Town, as well as our experience organising similar academic events. We set up goals that revolve around three blocks based on our applicants’ motivation: training of the next generation of computational social scientists who will be able to teach others and develop new tools; generate sound scholarly research evidence that leverages computational and digital tools to better understand the society and accelerate progress towards the SDG; build long-lasting academic/research collaborations. To achieve these objectives, we structured our programme to be composed of five components: a research happy hour, invited presentations, tutorials, joint programmes with other sites, and group projects.
Research Happy Hour
Research happy hour emerged as one of the most exciting aspects of our programme as it was intended to prepare our participants for the invited presentations as well as tutorials. All participants were assigned four research articles to read, review and discuss. We asked groups to focus on the rationale of the study, the methodology and potential to replicate the results, limitations, and ethics, as well as how they could tailor the methods to their area of expertise. During the research happy hour, the groups collate their ideas and share their key takeaways with everyone. To avoid having a few participants dominate group discussions, we asked each group to appoint a discussion leader and a rapporteur. No group member can take the same role twice at any time during the summer institute. Participants really enjoyed this aspect of the programme, and several reported that it prepared them for the main sessions of the day. Given the engagement and how the participants found it helpful, we would encourage other sites, particularly those with participants with limited CSS experience, to consider incorporating it into their programme.
During initial planning, we decided it was best to amend the lecture content into a live presentation to encourage participant engagement. We also focused on topics highly relevant to our site theme and collective research interests. We targeted researchers and academics who have an interest in Africa. Unfortunately, our preference for African-focused presentations and speakers limited our pool of potential speakers and made it challenging to find willing speakers. This challenge, however, reinforces the need for increased investment in CSS training and research in African countries. Ultimately, we had nine fantastic speakers who shared their work on different aspects of CSS (including bibliometric analysis, sentiment/text analysis, machine learning, and new methods for survey research) with our participants. We were particularly fascinated by the level of detail in these presentations and their willingness to share their slides and code with our participants despite our inability to offer a reasonable honorarium due to funding constraints. Given the fact that most of our speakers were overseas, we were worried that an all-virtual speaker presentation would limit speaker-participant engagement and an opportunity to have a one-on-one session with the speakers. We allocated ample time (105 minutes) to each presentation so that there was enough time to interact with the speakers. In the future, we hope to attract more local speakers and additional funding to cover travel costs for speakers and offer a more reasonable honorarium in recognition of their efforts and time.
In addition to invited presentations, we incorporated tutorials into our programme, focusing on different aspects of programming and CSS (text analysis, machine learning, spatial analysis, r-package development, reproducible analysis). All the tutorials (except text analysis) were held live and supported by the TAs on-site. We note that the overall rating was very high for all the tutorials especially given that our tutorial leads were highly experienced and adapted their tutorials to suit our context (Nigeria). There were some critical comments as well. Most importantly, our participants affirmed that condensing the programme into ten days rather than a full-two-week summer institute did not allow for a good balance between learning and practice. While virtual presentations also worked very well, the virtual environment for tutorials evidently did not work well for our group. A number of technical and infrastructural challenges (for example, slow PC processing speed and internet connection for installing packages) significantly affected some of our participants’ pace and ability to follow many of the tutorials. In the meantime, we have uploaded some of the tutorials on Youtube and made all teaching materials publicly available while in the future, we will extend the summer institute to two full weeks and focus on a single theme each day. Other programming tutorials (such as spatial analysis, R-package development, and reproducible analysis) will be held in the second week with group projects.
Joint Programme with other Sites
We collaborated with SICSS-Howard, SICSS-JIAS/IPATC, and SICSS-West Central Africa to organise a panel session focused on interrogating data collection strategies in Africa. We had a series of preparatory meetings, and when the event was finalised, we shared the information with our participants. Many of our participants enjoyed the session as it fostered a sense of shared experience. We hope to explore (with colleagues) opportunities for fostering cross-site interactions during this event so that our participants can interact with each other (our third goal - collaboration). Although the mass collaboration/experiments session and the fragile families challenge are indeed exciting and could foster collaboration, we could not incorporate them into our programme because we wanted to tailor the programme specifically for the needs of our participants while also considering their technical skills. Instead, we liaised with an African-led data science company Zindi to provide our participants with a similar experience. Though Zindi was open and willing to sponsor a hackathon and provide our participants with their data science (or CSS) competition platform, we could not see the engagement through because of our extremely late engagement with the Zindi team. In the future, we hope to still liaise with Zindi and partner with other African sites.
A major part of SICSS is group research projects. Our experience from participating in previous SICSS has taught us that leaving the participants to develop CSS-focused research question, particularly among participants with limited CSS experience like ours, often leads to very ambitious projects. While it is important that participants also experience this research development phase, we decided that the most successful approach was to identify important research questions based on the topics covered during the summer institute. In line with this, we identified three important research questions that centre around different dimensions of the sustainable development goals (SDGs): good health and wellbeing (Goal 3), gender equality (Goal 5), and reduced inequalities (Goal 10). In choosing our final projects, we were conscious of financial constraints and only shared projects that did not require funding or fieldwork with the team. Before their assignment to groups, we asked the participants to select two CSS methods they were likely to use in their work. We then assign participants to work on the pre-defined projects based on their interests and technical strength. This way, the participants hit the ground running and were able to complete significant work in the little time available. The participants worked well together, and it was particularly exciting to see how they applied their skills to new problems.
Closing Ceremony and Departure
Our closing ceremony featured presentations of preliminary findings from all the group projects. The presentations and discussions were lively, and it was a great way to wrap up the week. The groups enjoyed sharing their ideas and discussing what worked and what could be done better. We distributed SWAG t-shirts to participants, and all participants were given a certificate of attendance. Ultimately, we were very happy with our cohort. It was particularly heart-warming for us to see the seriousness with which the participants engaged with all aspects of the programme, including the research happy hour, invited presentations, tutorials and group projects. We believe that our rigorous selection process may have contributed to our overall experience and helped us select participants who were likely to participate fully in the programme.
We believe that our work has just begun at SICSS-Covenant. Over the coming months, we hope to support the groups to see their works through to dissemination and publication. We have also encouraged our participants to stay in touch with the SICSS community through our Slack channel, Facebook, WhatsApp group and participation in virtual group meetings.
Centre for Social and Behaviour Change (CSBC), Ashoka University, and CSIR-Central Scientific Instruments Organisation collectively hosted SICSS-Delhi (Summer Institute in Computational Social Science-Delhi) from 13th June 2022 to 17th June 2022. For the first time, an Indian location served as one of the SICSS partner sites. We hosted the summer institute in virtual mode. Even though we were aware that doing SICSS in virtual mode has certain recognized obstacles, similar to other online learning programs, the choice was made with the COVID-19 outbreak’s uncertainties in mind. Below, we have provided a few details and shared our experiences with every process involved in conducting the summer institute.
The first step in this process was to decide on the intake to SICSS-Delhi. Even though we were conducting it online, we preferred not to have too many participants and restricted ourselves to 25 participants. The reason was apparent that a smaller number would help to have one-to-one interaction between the speakers and the participants as well as among the participants. The next step was to set up a Google form so that candidates could fill in their details to apply for the summer institute. Apart from asking for information regarding their educational background, area of interest, familiarity with programming languages, etc., we also asked them to submit their curriculum vitae, a research statement of the ongoing project, and a list of completed coursework that is relevant to the program. The link to the Google form was provided on the SICSS-Delhi website. All the other necessary updates were also made to the website. We also created an official email id for SICSS-Delhi so that we can receive and resolve, if any, queries among the candidates regarding the program.
Apart from the faculties of the summer institute, we invited following guest speakers:
We are thankful to all the guest speakers for accepting our invitation and delivering such insightful talks.
For the success of any learning program, it is essential to have good participants, which requires wide publicity for the program. The activities conducted in this regard can be divided into the following two phases:
Before commencing the advertisement activities, we need to prepare materials that would be used to disseminate information regarding the program. Thus following tasks were done in this phase:
Once all the required things were ready, we kick-started our advertisement on 28th February and continued the process till the last date of receiving applications from candidates. The following tasks were done in this phase:
We received more than 100 applications from participants from different academic fields and research backgrounds. Out of these applications, we were required to select only 25 applications. For this, we employed a screening procedure that involved reviewing the applications and assigning scores to each of them based on several parameters, which are listed below:
We assigned two reviewers for each application and distributed the screening task among all the members of the SICSS-Delhi Team. Every review resulted in some score, along with some comments. Based on the reviews obtained for each application, we took a decision (i.e., select, reject, or wait) through a rigorous discussion among our team for each application. Thus, an email was sent to all 25 selected applicants, and they were asked to fill out a Google form to confirm their participation. We also send emails to the waitlisted as well as rejected applicants. Out of our 25 selected applicants, a couple of them could not accept the offer due to some unavoidable circumstances; hence those two seats were occupied by the waitlisted candidates. Finally, the application statuses of the remaining waitlisted candidates were changed to rejected, and the same was informed to the candidates through an email. If we consider the statistics of our selected 25 participants in terms of educational background, 10 (i.e., 40%) were young industry professionals; six were pursuing Ph.D., four were doing Post-Docs, the remaining 8% (i.e., 2), and 4% (i.e., 1) are post-grads and under-grads students respectively. In the case of area of expertise, seven participants are from computer science, and a similar number of participants are from the Public Health sector. There were five participants from Economics backgrounds also. Besides these, we also had one participant each from Public Policy, Engineering, and statistics areas.
Figure 1. Educational Qualifications
Figure 2. Areas of Expertise
After finalizing the participants for the summer institute, the next step was to initiate the onboarding process. The first task in this phase was to have the participant’s profile on the SICSS people page and the SICSS-Delhi people page. For this, we emailed all the participants the Google form link shared by the SICSS core committee. In this way, we obtained the photo, bio, and profile link of almost all the participants and other details (although after sending several reminders for the same to some participants). Once we had all the required information about the participants, their profiles were added to the website’s people page under the “Participants” category.
The next task was to communicate pre-arrival instructions to all the participants. It involved two main things. First is sharing the pre-reading materials and programming practice links/videos related to Python and R with all the participants. The second thing was to share a link to a slack channel created within the SICSS2022 Workspace. The organizers, as well as TA, were also part of this public channel. The purpose of creating this channel was two-fold. One was to have a standard and official mode of communication between the participants and the team members, where they could make announcements and resolve the participant’s queries. We even announced the TA office hours for the same. The other purpose was to motivate the participants to network with each other through public discussions or private discussions. Following the SICSS traditions, we also conducted an informal Zoom session two weeks before the program and had discussions with the participants (not all joined it as it was an informal discussion to guide the participants regarding the program).
We also shared the program schedule as a part of the onboarding process to give them a complete idea about the contents to be covered and the speakers for the same. The schedule was posted on the website and shared with the participants through the mail and slack channel.
Finally, after all the planning and preparations, our summer institute started on 13th June 2022. The timings were kept from 10:00 am to 5:00 pm throughout the week, except for the first and the last day. On the first day, we organized a keynote on Computational Social Science at 7 pm, and on the last day, the closing ceremony was scheduled at 5 pm. We preferred to use the Zoom platform for the sessions as it is a widely used platform for online meetings. A common link for all the sessions was created beforehand for the entire week and was shared with the participants. The same link was separately sent to all the invited speakers as well. The day-wise activities/sessions of the program are as follows:
13th June: The program began with a Meet and Greet session, during which the participants, TA, and organizers all introduced themselves to one another. The vital goal of this introduction was to get people acquainted with one another and their fields of endeavor, laying the groundwork for networking, which was one of the main goals of the summer institute. The sessions on this day focused on experimental aspects of research. The first talk covered the design constraints involved in experimental surveys, followed by a hands-on tutorial to examine how digital campaigning affects people’s perceptions of COVID safety precautions. We divided the participants into several groups for the tutorial and created Zoom breakout sessions for each group. A discussion on incorporating real-world experiments into the policy was also held. After a break of 2 hours, because of the different time zone of the speaker, we had our keynote session about computational social science in general and how scientific simulations could be used for social and public good.
14th June: On this day, our first session was on highlighting the impact of chemical senses in understanding, diagnosis and treatment of COVID-19. The second session of the day was to understand causality from computer science perspective. The other two session of the day focused on highlighting the importance of ethics in research and the role of AI in public health sector.
15th June: In the market and lemons theory framework, the third day of the summer institute was focused on gaining a deeper understanding of how stock markets operate. Several group activities were also conducted to demonstrate and educate stock market behavior.
16th June: The day started with a discussion on the eco chambers that could be seen on social media and a tutorial to help people comprehend the dynamic nature of opinion that results in such eco chambers. Additionally, we held two sessions with a strong emphasis on agent-based modeling and its practical application. The day’s final session focused on issues when analyzing multilingual texts in languages with limited resources.
17th June: On the final day, we had a hands-on tutorial in continuation of the last session of the previous day. The tutorial focused on different text analysis approaches to dealing with text written in Hindi. After this session, before starting the presentations for group projects, we created breakout sessions. We asked all the participants to prepare a presentation through discussion among their group members. They were expected to present an idea with some literature background and the suggested approach to bring the idea on board. We had some really good presentations from the participants, followed by the closing ceremony.
We had a very positive experience with the summer institute on the whole, despite having conducted it in online mode. We are incredibly thankful to the participants for their active involvement and for making it a success. We were satisfied with our decision to have it for one week as conducting an online program for a longer duration is quite challenging. A of the glimpse of the summer institute, posted on Twitter for promotional purposes, can be seen below:
We have divided this document, per guidelines, into sections relating to: 1) Outreach and application process; 2) Pre-arrival and onboarding; 3) First Week; 4) Second Week; 5) Post-Departure; 6) General Notes.
Outreach: We used multiple avenues to advertise the summer school in Edinburgh. These included Twitter, subject-specific pages (on Teams) for particular postgraduate programmes at the University of Edinburgh, the Scottish Graduate School for Social Sciences (SGSSS) email listserv, and the Social Network Analysis in Scotland mailing list. Ultimately, general advertising through Twitter and the main SICSS website was likely most effective as we ended up with relatively few individuals from Edinburgh. This is something to think about improving in future, particularly out of environmental considerations.
Application process: We set a deadline of early April for applications. As with previous iterations of SICSS that Christopher Barrie has been involved in as organizer, applications tumbled in on the last or second last days before the deadline. We did not require letters of recommendation but we did require letters of support for those applying to the two SICSS-Edinburgh scholarships we were offering to outstanding students who couldn’t otherwise fund their travel and accommodation.
Selection process: We received 46 applications and accepted 31. Ultimately, 25 attended after dropouts. The selection process involved each Organizer (of 5) to rank participants (yes/no/maybe) and the successful participants were those who received most yes votes. For those marginal cases, the Organizers discussed their respective merits in order to reach a final decision. Each Organizer also decided on the merits of those applying for the SICSS-Edinburgh Scholarship. This was a longer discussion and was decided after mutual deliberation. Here, we made judgments based on the same Eligibility criteria displayed on the webpage but with particular on emphasis on likelihood to benefit from participation.
Prior to arrival, we used the Onboarding form materials to make an email list for initial communication. All participant information was then added to the main SICSS webpage. Finally, we created a standalone Slack for SICSS-Edinburgh participants. One mistake here is that we should have added participants to the main SICSS Slack as well. This was only done midway through the first week, meaning there was only limited interaction with the main SICSS Slack (less so than on previous years when Chris Barrie organized SICSS-Oxford, for example but this is perhaps because SICSS-Oxford was an online event).
It would have been worth encouraging the use of this platform before the First Week so that participants could get to know each other a bit better.
Day 1 involved a more general introduction and introduction to ethics. As in previous years, the ethics discussion was a productive way to get individuals to get to know one another. We also asked participants to briefly introduce themselves at the beginning of Day 1, and Chris introduced the other Organizers and TAs.
Each day was devoted to one topic (CSS for Social Scientists—Chris, APIs and Web Scraping—Chris, CSS for Computer Scientists—Björn, Qualitative Digital Research—Karen, Networks—Tod, NLP and Machine Learning—Björn and Walid). These were followed by research talks by three Edinburgh-based researchers across Informatics, Political Science, and Design Informatics (Maria Wolters, Clare Llewellyn, and Benjamin Bach) as well as two invited speakers from the Hertie School and the University of Sheffield (Anita Gohdes and Carol Scarton).
We did not ask participants to fill in feedback surveys after the First Week but Chris did open up the floor to any questions, criticisms or other feedback. The decision was taken not to provide feedback surveys due to low response rates in previous iterations of SICSS and because participants were generally spent by the end of the day!
In retrospect, we could have used more innovative means to get participants to get to know each other in Day 1, such as speed dating or similar used by other sites. We did have success creating a Google Map of things to see in Edinburgh (see here) and organized a treasure hunt across Edinburgh, which one group did complete. The Organizers could also have done more to integrate TA contributions and encourage participants to learn from their considerable expertise alongside asking Organizers for advice.
The major limiting factor was an outbreak of Covid in Week 1, which spread to four of our participants and meant a number of additional participants elected to stay at home in the Second week.
In the Second week, participants divided into groups that were most-similar in terms of interest. They then discussed project ideas for 1.5 hours and then were instructed to select into another group to discuss ideas. We could have done this second step better as participants were generally reluctant to shift group after the first 1.5 hours. In hindsight, we should have also moved them into most-dissimilar groups. This was also made difficult by having to organize hybrid group discussions, which meant some participants felt excluded from the process. This is where more TA oversight should have been encouraged.
We also organized, as we did in the First week, several pub trips for social activity after two of the days in the Second week. In addition, we organized an alternative research talk of Pedro Jacobetty talking about how to create generative music in Python. This was welcome relief to students who were slightly overwrought by this point in the summer school and is certainly something I would repeat. We also had two group trips—one to a poster exhibition of work by Informatics PhD students at the University of Edinburgh and another Mixer event at the INSPACE Design Informatics studios organized by the Centre for Data Culture and Society (thanks to Morgan Currie and Lisa Otty.
Group projects were universally impressive, and spanned a number of topics from: reactions to a change in platform ownership/governance; to gun violence in the US; to attitudes toward sexual behaviour in popular TV drama; to the performance of Transformer models in political science research; to the lexical structure of Chinese epic literature; to reactions to the Depp-Heard trial.
In retrospect, it would have been nice to provide some sort of prize or certificate upon completion of the final day. We did go for a celebratory meal but some participants had to be absent out of health considerations.
The Slack is still open and we have acquired additional research funds that participants can apply for (~£600 on Prolific) to encourage groups to continue pursuing their group work research or other related research.
We plan for a three month check-in event online where groups can present group work research in development and any research that they have gone on to pursue.
It was a challenging thing to organize this. And writing this post-mortem makes its author (Chris Barrie) realize how much work went into its organization. Notwithstanding the Covid outbreak, which was handled as best as it could have been via a DIY hybrid setup, the event went smoothly and very successfully.
In future, the main Organizer Chris Barrie, notes that it’s worthwhile having a main organizing team of 2-3 in order to make delegation of work more simple. Given many are still working remotely, it was sometimes difficult to determine lines of communication required for effective delegation of tasks. Given Chris was the only Organizer who had organized before, he was also the only one aware of the contours of the event and what was required. This meant it was occasionally hard to communicate what needed to be done. This should be easier now that there are multiple organizers who have contributed to the event at Edinburgh.
Drinks at an Edinburgh pub
Informatics poster event with participants
SICSS FGV-DAPP Brazil was held virtually from June 14-26, 2022. We invited 28 participants from a list of around 40 applicants. We ended with 26 participants. Our first week followed the main curriculum from SICSS. The second week was focused almost exclusively on collaborative research projects with our participants. This is a modification from our SICSS 2021, in which we organized several roundtables in the second week. In this edition, we aimed to focus more on research, and we believe the results were very positive. This post-mortem is divided into six sections: 1) outreach and application process; 2) pre-arrival and onboarding; 3) first week; 4) second week; 5) Post-Sicss, and 6) Final Remarks.
Different from FGV DAPP SICSS 2021, we were a bit more limited in advertising our edition. We believed that the network and name reputation built on last year’s edition would be enough for us to achieve a similar number of applicants. However, that was not the case. Although we still received more applications than spots we could offer, we received less than half of the number of applications compared to 2021. Therefore, particularly as it pertains to fields of study, our 2022 edition was less diverse, with a majority of students coming from political science programs that already have a stronger tradition in CSS.
We used three main strategies to advertise our edition. First, we advertised widely our call for applicants using the organizers personal networks, particularly, on social media platforms, like Twitter and Linkedin. Second, we used distinct institutional newsletters from Getulio Vargas Foundation to reach out to an even broader audience than the organizers’ direct network. Third, we sent an email to all the participants from last years’ edition.
Overall, we believe our advertising for SICSS Brazil could be improved. The event received less traction on social media than last year, and we received fewer applications. This might be a constraint related to the supply of Ph.D. students in Brazil who are interested in the field. One idea for the future is to advertise more broadly and also accept more Master’s students in the program.
We selected twenty-eight applicants, and only two declined our invitation to join the SICSS. The majority (about three-fourths) of our participants were Ph.D. students, with some junior faculty members, postdocs scholars, a few master’s students, and one data scientist working in the private sector.
One interesting thing from this year’s experience is that the attendance did not decline at all throughout the weeks. So we believe we are at the right balance for a virtual edition and for the number of students who we invited to join.
After participants were accepted, we sent participants an email asking them to confirm their participation, using the same template provided by the SICSS-Princeton Team. In the same email, we introduced the structure of our SICSS, and asked them to start working on the pre-arrival materials. All but one participant accepted our invitation to join the SICSS FGV DAPP.
A week later, we sent a second email adding all of them to our Slack, and asking for a photo and bio to be included on the website. Both of our emails stressed the importance of reviewing the pre-recorded lectures once they were posted because the institute would be a flipped classroom format.
Overall, we believe most of our participants watched the pre-recorded lectures, and some read a few chapters from Bit by Bit. Students also worked on the codes beforehand. And we believe the majority of students were still actually improving their programming skills as they were preparing for the SICSS. We organized three preparatory meetings before the SICSS, and some of them were more focused on the coding exercises.
We were worried at the beginning that the students would have a hard time following the lectures, since English fluency is not too high among Ph.D. students in Brazil. However, we were positively impressed by the feedback we received from them. The feedback about the lectures was extremely positive, we did not receive any complaint about the topics covered, or any student saying they could not follow the lectures due to any type of language barrier.
For the first three days, we deviated very little from the main curriculum. We started with a smaller group discussion about CSS and expectations for our SICSS edition during the morning, and covered the Pantheon exercises during the afternoon. On the second day, we did the suggested scrapping exercise. And on the third day, we covered text analysis.
There were, however, some changes in the activities on these three days. On the second day, we pushed our participants to work with anything other than the Twitter API. Our feeling was that most of our participants had some experience grabbing data from Twitter, and CSS as a field has an over-representation of Twitter as a data source. The students followed through the recommendation, and worked with different data sources, such as wikipedia edits, YouTube, fact-checking agencies, and official data from the Brazilian government through APIs. We got positive feedback from this activity, although some students complained about being too broad, and spending too much time thinking about what to do, instead of practicing how to collect digital data.
On the third day, we decided to start with more advanced NLP techniques in Python. We covered mostly the use of Transformers for classification purposes. The workshop was a huge success. Although the students were not familiar theoretically with transformers, the presentation and the notebook were easy to follow, and several of the students were planning to use transformers on their final projects.
On day four, we deviated substantially from the main curriculum. Because Mturk is not widely available in Brazil, we decided to focus on the use of Facebook Ads as an alternative to recruit survey respondents. We started the day with a workshop about how to use Facebook Ads, how to connect your surveys with the Ads, and which type of decisions, such as quotas, geolocation, number of ads, Facebook allows one to control. After the workshop, students created their own survey and put them available through a Facebook Ad. Overall, the participants appreciated the learning experience about using the Ads for social science research. However, because the time to complete the activity was too short, most of the groups were not able to collect enough responses, as we would expect from Mturk. We did, however, show how Mturk works with an afternoon presentation in which one of the organizers showed the platform to participants. This is the second year we do this activity, and the results are not what we would like to have. The solution for next years is to have some data collected through Facebook Ads beforehand, so they can practice the Post-stratification methods during the afternoon.
On day 5, we also deviated substantially. We decided to do the research speed dating in the first week. The rationale here was to give the students the whole weekend to think about which projects they wanted to work on the second week. We believe that was a successful choice. Compared to our past edition, engagement and quality of the projects in the second week were far superior.
Throughout the first week, we also requested participants’ feedback via anonymous forms almost daily. In those feedback forms, participants remarked positively about the content and the activities of each day, but showed frustration that the groups and ideas for the daily activities were not continued on the following days. We explained that the rationale for that was to have a diversity of experiences and collaborations in those days, so that they could be inspired to reunite with their preferred groups and elaborate on their favorite ideas for the final SICSS project and even beyond SICSS.
We invited three guest-speakers for the first week:
Unfortunately, Alexandre Siegel had to cancel her talk on the day of the event for personal reasons, but she kindly met with some of our students interested in her work in the following week. We believe participants enjoyed the Guest-Speakers and Workshops. We sensed that the guest speakers were a nice break from the group activities structure, and an opportunity for participants to see some research projects in practice.
Our schedule for week 2 was mostly focused on giving students time to work on their collaborative projects. We only organized two other workshops in week 2, and all the other time was devoted to collaborative projects.
On Monday, we organized a workshop on Network Analysis of Social Media Data in Python. This is the second year we have offered this workshop, and the feedback was overall positive. On Tuesday, we offered a workshop with the invited guest, Fernando Meirelles, on post-stratification of online surveys. The workshop was very instructive and worked well as a supplement for the Facebook Ads activity day.
Unlike last year, our attendance did not decline at all during the second week. In addition, engagement and the quality of the collaborative projects were really high. We think our decision not to pile up too many activities in the second week was very positive. With more time, participants could think carefully and engage in some exciting research projects. This is something we adjusted considering our post-mortem from SICSS FGV DAPP 2021.
SICSS Helsinki was organised in June, 2022 for two weeks. It followed the structure of previous SICSS in Helsinki and globally: first week focused on lecture-based teaching activities while the second week focused on project-based learning activities. SICSS Helsinki 2022 was organised as online-only course.
In 2022 SICSS Helsinki, we continued many practices already developed on SICSS 2021 and previous iterations, including:
In response to last year, we attempted to improve following aspects in our iteration:
Like last year, we conducted targeted advertisements for Finnish and Nordic audiences via list servers. Overall, we received a modest number of applications across disciplines: computer science, physics, communication and media studies and political sciences. We accepted a total of twenty participants, about half from Finland and others from mostly Europe.
As we organised SICSS as online-only activity, some participants also had challenges to participate in all activities of the programme. This is a factor which should also count in to online vs. in-person choices in the future. Committing two weeks of time without a clear in-person element may be difficult for students if they are pulled into various ad-hoc meetings etc. on their home institutions (which is easy, as they are “only” attending an online course).
Action point for 2023: We are receiving fairly low number of applications in SICSS Helsinki. Most likely with a bit more coordinated application procedures throughout all SICSS, the application pressure to some SICSS could be balanced and ensuring that high-quality candidates are considered for maximal inclusion in the programme.
We organised three ice breaking events before SICSS to help group socialisation via Zoom. While participant survey indicated that they did feel welcomed before SICSS Helsinki, I felt that these activities were less fruitful than last year. It is hard to articulate why I felt this way, it might be just too much online activities – but they felt a bit chilly in atmosphere. While the concept is fruitful, it might be that the presence of a teacher – an authority figure of some kind – does break some of the social side of these activities.
In addition to social activities, we asked students to prepare by reading book chapters and doing exercises: for those with less programming experience, doing programming activities and for those with stronger technical background, exercises focused on helping them to think as social scientists. However, the pre-training phase is still an area which requires improvements. The first week programming exercises were still perceived as fairly difficult and students felt not prepared. Comments like:
“I would suggest, for the sake of those unfamiliar with coding, that conducting 10-15 minutes (out of an hour of those pre-institute get-togethers) of familiarisation with the R/Python environments (RStudio/Jupyter) and the basics would go a long way to get people warmed up for the main two weeks. Such sessions are very helpful and easy to conduct even to get people going for doing even the pre-exercises. The starting of programming is so scary and intimidating that a bit of coaching or just a tour around could be of great help. It is also possible for those who have prior knowledge to do the intros in breakout rooms, for instance.”
“Maybe a nice conclusory session in the end of each session, showing how results could look, if Matti did the tasks, would be helpful.”
The exercises were designed so that majority of them could be solved by changing a variable name or reading through a linked documentation. We also noted with the TA that overall, problems with the exercises emerged on topics we felt the pre-assignment was supposed to cover already. We examined the self-reported information on pre-assignment task completion and noted that most students did not seem to do pre-assigned tasks at all.
Action points for 2023:
First week activities followed our standard structure: chapter review, group-work based on chapter learning activities on research design and discussion followed by self-phased programming tasks following our course materials, which had narrated voice-over tutorials this year as well.
Students appeared to enjoy the first two activities quite well and they served in the group formations:
“Discussions were great and gave me a new perspective on many things.”
“People had the chance to get to know each other during the group assignments.”
“Interactive teaching sessions in the theory section.”
The technical programming tasks received more criticism, which partly related to pre-SICSS program and its challenges. However, I do think that the criticism speaks more about the need to ensure sufficient technical background pre-SICSS than radically alter the exercises and their content.
For the second week, we continued the idea of team-based meetings every day but reduced the meeting time for one longer meeting each day. The projects were kicked of previous week by asking groups to formulate their initial ideas for me to review and provide comments.
The instructions for daily meetings included:
“You should have received calendar invitations for daily meetings with me. I will ask the following questions on every day:
On Monday, I will focus on questions 2 and 3 more and read through the few sentence descriptions you have written and provide comments. Just a hint based on personal experience:
This instruction helped students to focus on the group meetings more than last year, at least the student feedback did not have negative sentiments on these. I do feel that this formulation might have made the meeting a bit too teacher-centric reporting events, while the purpose of such meetings is primary to support groups to identify problems and sync regularly (we have adapted these meetings from agile software development). Slight adjustments on the phrasing might help to communicate this aspect more but might also dimmish the learning value.
Student criticism emerged mainly from the group forming phase, where we tried to identify maximum similarity of groups based on the research interests twice and then allowed groups to form after these discussions. Partly this may be me not explaining the process clearly, which lead to some confusion; partly it may be that Zoom did not really afford people to sufficiently wonder around in group forming phase to find interesting topics, partly it might be just a problem that groups need to be formed somehow and ultimately it is hard to make everyone happy.
Action points for 2023: Clarify the group forming process descriptions more and examine how to best people wandering between groups. In 2021 we used Gather Town, which was somewhat easier in this stage, but had other challenges and additional costs related to it.
Overall, I consider that the core elements of the SICSS program are in place and the flipped classroom approach allows focusing more on interactive learning during the SICSS. However, such program puts significant pressure for students’ self-regulation skills, and this is an area for development.
Second area I feel we as SICSS should consider more is expectation management and continued program. The two-weeks serves as an intensive kickoff, but it should not be the last course participants take in computational social science – they journey must continue beyond this. We could help to identify future steps more and try to communicate this idea in the marketing more clearly.
Action point: I will invite SICSS Helsinki participants to my intensive courses in Network Analysis and Data Science taking place in January 2023. Sadly, we cannot offer these courses free of charge, but they are highly compensated in costs (75 euros each).
The second Summer Institute in Computational Social Science at Hong Kong was held from June 20 to June 30, 2022. It was (virtually) hosted and sponsored by the Computational Social Science Laboratory (CSSL) and the Centre for Population Research of the Department of Sociology at the Chinese University of Hong Kong.
From 177 applications, we admitted 29 participants. There were 18 PhD students, 6 MA/MPhil students, 5 Postdoc/professors, coming from disciplines as diverse as Communications, Political Science, Economics, Sociology, Psychology, Geography, History, Law, Hotel management, etc. Origins include Hong Kong, mainland China, UK, Singapore, Germany, Iran, among others.
This post-mortem is divided into six sections: 1) Outreach and Application Process; 2) Pre-arrival and Onboarding; 3) First week; 4) Second week; 5) Guest Speakers and Recording; and 6) Final Remarks.
Learning from our experience from the first SICSS-Hong Kong that we started late (but successfully organized), we secured enough time to ask for funding from both the SICSS headquarters and Jaemin (lead co-organizer this time)’s university. We wanted to give sufficient time for applications, considering that the pandemic situation got worsened early 2022. Our deadline of application was April 15, 2022.
During planning, our main target audience are research students and faculty in Hong Kong and the Greater China Region. The rationale is that there has been a growing body of faculty in Hong Kong doing computational social sciences, and a clear growing interest among students. However, there lacks intercollegiate conversations among local universities and researchers on computational social science. Therefore, fostering a working group on computational social sciences is a goal.
We advertised our event to all universities in Hong Kong, as well as major research universities in mainland China. This time we got help from CUHK’s admin staff – they have a large pool of emailing lists that they often use when they organize monthly CSS seminar. The advertisement was mainly spread through emails to university departments.
Last year’s action point was “We could have people read lecture materials and watch videos further prior to the event (say, 4 weeks before)” and we did exactly that. About one month prior to the SICSS, we emailed our participants to encourage them to watch videos of the lectures on the SICSS bootcamp website and train themselves about programming. We hired a teaching assistant, Yujie LI, and he offered virtual office hours twice a week. A few of the participants told us that they actually got help from the virtual office hours from our TA.
However, some feedback survey responses indicate that the coding skills that they learned from the boot camp were not quite linked up to the contents of our workshop. Indeed, we only spent the first four days on the standard SICSS modules like ethics (Day 1), collecting digital trace data (Day 2), automated text analysis (Day 3), and digital field experiment (Day 4). For the rest of the days, we covered special topics such as image and audio analysis, mainly learning from guest speakers’ talk.
Action point is that we could have asked guest speakers to share their pedagogical contents or useful readings, which can be combined with the standard bootcamp materials.
|1||June 20||Introduction and Ethics||Ethics discussion|
|2||June 21||Collecting Digital Trace Data||Idea development|
|3||June 22||Automated Text Analysis||Coding exercise|
|4||June 23||Digital Field Experiment||Mini conference|
|5||June 24||Image Analysis/Computational Experiment||Making project teams|
We began Day 1 with “meet and greet,” providing the overview of the workshop program. The Dean of the Faculty of Social Sciences at CUHK offered a great opening remark.
We followed the standard SICSS modules and group exercise materials for the first four days. In “mini-conference” on Day 4, we allowed participants to introduce and present their work in 20 minutes. We had two presenters. Day 5 we encouraged and advised grouping for group projects, which we encouraged to present on the last day in Week 2. Organizers and TAs occasionally jumped into their breakout rooms to check and help the progress.
Each day we had a speaker’s talk (guest speakers or co-organizers).
An important organizing task in the first week was to facilitate networking in a virtual environment through group exercises. Last year, we only used random grouping in each exercise. We had received mixed responses because, on one hand, it gave more chances to know each other with many others but, on the other hand, it prevents them from fostering a stronger tie and makes them often encounter someone whose interest is often too different.
To address this feedback, this year we prepared “buddy groups” – a pre-filled list of groups composed of 4-5 people matched upon their declared interests, discipline, and career stages – what we named “Applied Sciences”, “Chinese Politics”, “Econ/Development”, “PhDs”, “SNS/Misinformation”, and “Sociology/Population”. Then, we used both random group assignments and buddy groups.
Reflection: The former was effective on Day 1 and the latter was effective when developing a collaborative idea/project. We believe that most buddy groups naturally evolved to the development of group projects. They were able to present their research project in the end, which was our goal.
Along with the buddy groups, we adopted a technology called Gather for “Research Speed Dating.” Gather is a fluid video chat system where participants could bump into each other while walking in and out of conversation. After creating our own town in Gather, we conducted speed dating on Day 2 and Day 3.
Reflection: With a large number of participants, we weren’t able to give all participants an opportunity for a “flash talk.” Gather was an alternative. Research Speed Dating with Gather seemed more effective at least for casual conversation. One drawback, however, was that participants’ interactions were not under our control. We noted some participants just wandered around the side of the town and were not really in communication with their peers.
For those staying in HK and available to socialize in person, we treated them with a luncheon on 1 pm, Sunday, June 26 at Chung Chi College Staff Club, CUHK. We were joined by five alumni of the 2021 SICSS-HK.
|6||June 27||Social Networks/Survey||EGroup project|
|7||June 28||Audio Analysis||Group project|
|8||June 29||Image Analysis||Group project|
|9||June 30||Presentation/wrapup||Group project|
The second week we only had morning sessions, featuring guest speakers specialized in the respective area and released participants to work on their own group project. On Day 6, we had another session for “Mini-conference” for two participants to present their work.
On the last day, we had a total of five groups presenting.
|Baiqi, Haixin, Jing||online civility in the context of Hong Kong Anti-ELAB movement|
|Seong Hah Cho, Sean Guo, Zhengyi Liang, Angel Chang||wiki data & transformation of political ideology/identity|
|Han Li, Wenyu Li, Hakimeh Nasiri, Huanyu Bao, Jessica Li||mental health and social resilience during COVID-19|
|Xiaoqian Yue, Qing Lan, Shengbin Wei||Personnel flows in officialdom in China, 1880-1905|
|Tianwen Du, Guo Cheng, Xiaotong Li||Knowledge sharing and online consultations: Evidence from online health communities|
The quality of presentation was, in fact, very good. All groups have come up with well-documented, well-thought, and executable plans for their projects.
Reflection: Some presentations seemed to rely on a member’s own, ongoing project (rather than a product of collaboration). It would still be great if they found a collaborator on the existing project, because they expect to develop it further after the SICSS efforts. An action plan is that we preemptively and openly discuss such an option and help clarify/figure out what they can do on the project during the days of working on group projects.
We had a total of 8 guest speakers – for international speakers we offered an honorarium. Each of the co-organizers also presented a lecture. We received very positive ratings for what participants learned from those presentations.
For 5 international speakers (teams) we invited, our sponsor, CUHK FSS, wanted to live-stream their talks on YouTube. With the generous consent of four speaker teams for live streaming and recording, their talks have been uploaded to the CUHK Social Science Soundbox YouTube Channel (playlist). The links can also be found in our schedule page: https://sicss.io/2022/hong-kong/schedule
Reflection points: Some participants expressed interest in applying the learned techniques right away, but substantive applications may be a long way because the talks focused on the introduction and overview of methodologies. We can consider sharing tutorial materials in advance, if possible. But at the same time, understanding “what it is” might also serve an important goal of the workshop.
Great memories. We believe that the 2022 SICSS-HK was successful particularly in terms of:
We’ve divided the post-mortem into five main sections: 1) Outreach and application process, 2) Pre-arrival and onboarding, 3) First week, 4) Second week, 5) Wrap-up.
The Center for Computational Social Science at IIIT-Hyderabad, India, hosted its first edition of the Summer Institute in Computational Social Science on 11 July-25 July 2022. We aimed to invite participants across diverse disciplines. To start with the promotions, we designed a flier with the necessary information. Our major outreach effort began on February 15, 2022, on our social media handles. We did rigorous rounds of advertising on Twitter, Facebook and Linkedin. In addition, we also tried spreading the word via sending email lists to the faculty and institutions that we thought might know the interested participants.
For the application process, the flier constitutes a QR code that directs to a Google form. To ease the application review process, we asked participants to submit their CV and SOP describing how they are the best fit for the school. We have received 200+ applications across 8 - 10 disciplines. The applicants ranged from bachelor’s students to PhD Students. During the review process, we account for the diversity of research interests and whether they can create a synergy in collaborations.
Finally, we accepted 32+ applications that comprise 41% females and 59% males, spread across CS, Computational Biology, Journalism, Economics and Social Sciences backgrounds.
On June 20, 2022, we sent out a decision letter to the admitted applicants. Our summer school was an offline event. Hence we rolled out the email with the following details:
The first week of the summer school was a mix of guest lectures, hands-on tutorial sessions and group activities. The week kick started with a welcome note by the organizing team head. The welcome message was meant to familiarize participants with the following:
Coming to the guest lectures, we had distinguished speakers like Prof. Niloy Ganguly talking about tackling one of the most critical social issues of the time, Misinformation and that too during the Pandemic time. Then we had Prof. Alison Noble talking about the emergence of Sonography via Data Science. Dr Rajiv Ratn Shah from IIIT Delhi demonstrated various applications that leverage multimodal signals in solving the tasks. We also had Suhem Parack from Twitter introducing participants to the Twitter Academic API and its attributes. Another prime aspect of learning the CSS or any other field involves evaluating your solutions. To fill the gap, Aditya Medury talked about correlation, association and causation, and we had Fabricio Benevenuto illustrating the deployment of a system in combating misinformation.
We also conducted discussions and tutorials to encompass both theoretical and hands-on knowledge. For each discussion session happening the next day, participants were reminded to view the recorded lecture. The prime agenda of the discussion was to have an in-house discussion about the course where the assigned mentor prepared the extended content taking cues from the recorded video. Not only was each concept highlighted in the recorded version elaborated, but the session was meant to solve doubts and ask-answering questions. Next, each discussion session was followed by a hands-on tutorial session comprising rigorous training on the topics learnt. The first week had sessions related to ethics, collecting data from the web, understanding the pandas library via the lens of the Python language and learning tools to interpret or infer the survey results.
Besides bombarding participants with the information, we also had dedicated hours in the schedule where they could put their thinking caps and come up with innovative ideas for the projects. The plan for the projects in the first week was solely to focus on formalizing the ideas and coming up with the todos to proceed with the same. Keeping the two weeks timeline in mind, we had an introductory session where we informed participants about the expectations and outcome of the projects. Since it was a group activity, we advised participants to utilize each other’s expertise in forming the projects. Each project was handled by a TA that helped participants not only by brainstorming the ideas but also by directing them to think in the right direction. We kept two slots in the schedule for validation and feedback sessions with the panelists to measure progress. Not only this, but being an offline event, we encouraged participants to utilize the lunch and networking hours for one-on-one interactions with the speakers.
Last but not least, we also had paper reading slots as part of the group activity designed with a plan of imparting knowledge but enjoying the company of each other too. The first paper reading session happened on the second day of the first week. Taking inspiration from Alec Jacobson and Colin Raffel, we designed paper reading as role-playing. We made it mandatory for every participant to read the paper beforehand. Being a group activity, a team of TAs demonstrated the paper reading session and the different role-playing in the first few sessions. The remaining sessions were led by each team.
In a nutshell, the jam-packed week introduced participants to the following:
A significant part of SICSS is participant-led group research projects during the second week. We kept numerous feedback/project work hours where students could spend time on their proposed ideas. Mentors/TAs were available at the venue to solve any form of queries and roadblocks. We also scheduled a mid-week project progress evaluation slot with the panelist.
Likewise, the previous week, we continued on the distinguished lecture series and discussion sessions. The week had Micheal Bailey and Harshil Sahal talking about the social networks and spatial mobility-evidences from Facebook in India, Jisun An from SMU speaking about tackling the social challenges with data science and AI. We also had Munmun De Choudhary enlightening the audience about how to combat mental health via social media. Rikin Gandhi introduced the niche area of digital green to the participants. We also had Vasundhra Kaul talking about understanding data privacy through socio-legal lenses. The discussions and hands-on sessions taught various concepts like mass collaboration, conducting experiments and introduction to the causal inference libraries.
The D-day of the summer school was scheduled for July 25, 2022. The day began with the final project presentations by each group. The panelist not only evaluated the work but also provided pointers on improving it further. Next, we conducted a 30-min anonymous feedback session to jot down the feedback from the participants. A google form was shared with them that demanded feedback on the lectures, discussions, hands-on sessions, group activities and logistics. Following were some sample feedbacks: (i) The guest speaker talks were the best part of the curriculum for me as I got exposed to various research questions people have been tackling in the area and just exposed me to how the research pipeline works in real projects. (ii) I liked all of them [guest speakers]. I learnt that what can be the research perspective behind each and every thing, even behind very simple things. (iii) The feedback from my TA really helped us to modify the problems statement in a much and define the question in a proper way. (iv) This [reading papers with personas/roles] I felt was a fairly interesting way (multi-faceted) to read and and analyze papers.
Being the first edition of the SICSS, we have the vision to make it better for the coming year and feedback from the participants will aid in it. Last but not least, we conducted the certificate distribution ceremony, where we acknowledged and appreciated the dedication and sincerity participants showed during the timeline. We gave them participation certificates along with hoodies as a token from the SICSS community. Finally, we took loads of photos and said the final goodbye.
From 20 June - 8 July 2022, The Johannesburg Institute for Advanced Study (JIAS) and The Institute for Pan-African Thought and Conversation (IPATC), both based at the University of Johannesburg, sponsored the Summer Institute in Computational Social Science (SICSS), held in a hybrid format. SICSS-JIAS/IPATC was organised by malebo sephodi, with Zimasa Ndamase as sessional facilitator, Xolani Makhubele and Sven Lampe as support. The first two weeks took place online. The third week combined remote and in-person participation. SICSS-JIAS/IPATC was intentional in centering African narratives in its curriculum. The first week focused on African context research in the digital age. The second week followed the SICSS main curriculum closely with updated group exercises, enrichment lectures and workshops. The third week was in person and included lectures and group work. We used UBUNTU as a foundational framework in how we approached the planning and implementation of the programme. We applied radical Black Love in our approach to the site.
We’ve divided the post-mortem into six main sections: 1) outreach and application process; 2) pre-arrival and onboarding; 3) first week; 4) second week; 5) third week (group projects); 6) post-departure.
Our application process took place between January and April 2022. We advertised our institute to a large, diverse group. We were intentional about prioritising applications from previously disadvantaged universities in South Africa (HBUs) and the disciplines of the social sciences. This meant deliberately having meetings with faculties in the specific institutions to share the information with their postgraduate networks. We also reached out to a varied network of scholars and universities and academic websites that serve the African continent. Websites such as Opportunity Desk and Academic Hive advertised on their platforms. Most of the applicants heard about our programme from their supervisors.
We regularly met with site organisers from SICSS-Covenant in Nigeria, SICSS-Central/West Africa in Cameroon and SICSS-Howard Mathematica in the United States (US), through weekly support meetings. This was to build a strong network of Africans and African Descendants in Computational Social Science. This translated in the sites planning a joint event that would address data collection methods of Africans and African Descendants. Each site provided a panelist and we agreed on discussion points and how the event would run. We would pre-record the panelist opening statements and then host a live Q&A session at an agreed time. We all agreed to host it on 27 June 2022, this was the final week for most of the collaboration sites and the second for ours. All the sites were first-time organisers except for Naniette Coleman from SICSS-Howard/Mathematica who provided great tips and insights from a structure and organisational point of view. Having this support group proved fruitful as it made the organising journey less lonely/daunting. Although each site had its own unique structure, we had a common thread and that was to amplify the voices of Africans and African Descendants in Computational Social Science.
Our applicants were expected to provide a detailed CV, Personal and Research statement of no more than 2000 words. Our application stream was slow and low and this could be due to the fact that in 2022 there were 32 sites who most had a virtual model as that allows for applicants to apply to a varied group of sites. Originally our application deadline was set for 31 March 2022 but was extended by a weekend to 04 April 2022. The extension helped the many applicants who missed the first deadline. Many sites such as ours considered in-person sessions, dependent on covid updates. We specified on the website that should restrictions adjust, we would be adopting a 100% virtual model.
We invited 17 applicants out of 52 applications and ended up with 14 participants. Most of the applicants were physically located in South Africa, although from various African countries. We accepted three international participants from Uganda, Cameroon and the United States. We looked at accepting mostly participants who had an interest in data science and computational social science methods regardless of background. Although fairly new in Africa, we were not particular about programming skills as we would be offering a coding Bootcamp.
We used the onboarding form provided by SICSS to collect participant info such as their brief biographies and photos and filtered the information to our website. In compliance with the South African Popi Act ( Protection of Personal Information Act) personal information was not shared without the applicants’ consent.
We made use of Slack and a Shared Document to communicate with participants. Although Slack communication was slow in the beginning, it picked up as the days went by. We arranged a meet and greet on 4 May 2022 and invited Emmanuel Olamijuwon, SICSS-Covenant founder and organiser, for a Q&A session. This session was 100% attended by the participants. The reason for the meet and greet was that we anticipated the difficulty and overwhelming nature of the coding Bootcamp and to remove the anxiety of dropping out.
Because many of our participants did not have any coding background, we created a coding Bootcamp schedule that would assist them in catching up with the basic skills needed to complete week two and three of the SICSS programme. The Bootcamp was meant to be a self-learning self-paced programme but we created a schedule as a guide running from 13 May to 13 June 2022. We also had on-hand programming support and participants could schedule virtual one-on-one consultations with our support. Those who did not need coding support were required to submit a book report.
Participants were required to send a report each week detailing their Bootcamp progress. This assisted in giving us insights on how to structure the SICSS-JIAS/IPATC three-week curriculum.
We used MSTEAMS as our meeting platform. It is a widely used platform in South African universities. We also offered daily updates on a shared document where we posted resources from the theme of the day and the following day’s programme. This was a great addition to Slack, as the uptake of Slack proved to be slow.
The focus of this week was to deal with question of what Computational Social Science is and what it looks like in Africa since it is still a new field. We incorporated the lesson of research/experiments in the digital age for African researchers. The sessions dealt with the digital landscape of Africa, data collection methods, ICT4D, Decolonising Artificial Intelligence and the ethics and responsibility of African researchers. Sessions were also geared to speak to each of the participants’ research areas. We exposed the participants to varied scholars with different presentation and ideological approaches. Meetings were held once to twice a day to minimise digital fatigue and to allow participants to focus on their coding exercises. Participants were strongly encouraged to continue seeking one on one support.
The participants were quite dynamic and vibrant and engaged with great zeal. The impact of the sessions proved to be beneficial as there was a golden thread among each speaker. This created a great foundation for the forthcoming main curriculum and group work weeks. For example, we hosted Sabelo Mhlambi who builds AI and works with NLP but has written extensively on decolonising AI, using UBUNTU theory and building ethical AI. This was a great precursor to a later session delivered by Prof Vukosi Marivate who dealt with the technical side of NLP and African languages.
As anticipated, this was a challenging week for participants, but we kept on encouraging and giving them tips on how to tackle many of the challenges they faced. We created an open environment where participants could freely share their learning journeys, this assisted in avoiding dropouts. We also prioritised the well-being of the participants, in that we used their feedback to adjust the schedule where necessary.
To minimise digital fatigue, we expected the participants to watch the main curriculum videos before coming to the sessions rather than watching the videos together, but this proved challenging for the first few days as many participants were falling behind with the video material. Participants struggled with coding prompting us to organise an emergency two-hour R workshop which proved helpful. We found that hosting live workshops compliments the Bootcamp. Participants reported that things made better sense when they went back to the Bootcamp and curriculum material after the live workshop. Participants who made use of one-on-one support reported the same.
Our online sessions were not without their challenges. Our institute coincided with a difficult time in South Africa as we experienced harsh bouts of load shedding. This meant power cuts up to three times a day, two hours at a time. In most areas, power cuts came with network issues. Since participants had to shift to their own networks after power cuts they experienced challenges in connecting to the sessions. The disruption was not so evident in the first week as we had a shorter programme, but we felt it in the second week. For example, our sessional facilitator would need to leave the sessions earlier due to the cuts in their area. Most areas had load shedding schedules and participants communicated their schedules ahead of time. This was frustrating particularly for this important week as it meant participants were falling behind. Mobile data is expensive in South Africa and some participants could not connect once they experienced these cuts. Many of the participants tried to manage load shedding as much as they could by moving venues during expected cuts, but it still proved to be a huge challenge.
Due to this, we had to adjust the programme, for example, we repeated day two’s lessons on day three. We allowed participants to rewatch the videos and reattempt the group exercises. This was a great decision as we saw a major shift in the ways participants gained confidence in engaging with the material. We were also joined by Suhem Parack from Twitter who was generous and accepted our invitation at the last minute.
We updated the group exercises for context purposes and used a more African approach to the case studies. Instead of Ethical articles that centered on the West, we focused on Africa and instead of Donald Trump’s tweets, we used Prof Mamokgethi Phakeng’s tweets. After extensive discussion, we agreed that we would collaborate with Howard/Mathematica on their mass collaboration exercise. We were then joined by Naniette Coleman, the organiser, who explained the background of the exercise and answered any questions that arose. Participants were keen to do the exercise but needed time to reflect on the ethics and the American context of the exercise and asked for an extension to look at it the following week.
Our local partners supported our site with local travel grants and the Social Science Research Council (SSRC) assisted our international participants with flights to South Africa. The participants from Uganda and the US arrived on Sunday 3 July 2022 and were checked in at Pathways Guest House. Unfortunately, our Cameroon participant’s flight was overbooked and this caused a delay in his arrival. All participants arrived early morning, Monday 4 July 2022, from different cities across South Africa. After participants settled in at Pathways Guesthouse, the UJ shuttle picked us up to have lunch on the JIAS campus. This was followed by a welcome from the JIAS director, a JIAS campus tour and a logistics meeting with the participants on their group work and the rest of the week.
On the same evening, IPATC hosted a welcome dinner where the participants had an opportunity to meet with the director. This is where it was communicated that participants would be appointed as junior fellows at IPATC and writing incentives were put into place for the institute’s group research output.
Each day would begin with group work, followed by an in-person lecture then the group work would continue. Participants could use both the guest house conference facilities and the JIAS campus to work from. Most participants preferred alternating between the venues.
On Tuesday 5 July 2022, Prof Vukosi Marivate spent the day with the participants. He delivered a two-hour lecture on NLP and then facilitated a fireside chat where each group pitched their research projects; he provided extensive feedback with assistance from sephodi. Most group projects focused on using Twitter API v2 and we encouraged participants to continue exploring other platforms for data collection.
Other in-person sessions included topics such as feminist theory, African Data collection methods, digital politics and wellbeing. Our online sessions included Olamijuwon who gave a lecture on a project using Facebook data and Chris Bail who engaged the participants mainly on professionalisation and a general discussion on SICSS.
Participants also received a surprise visit from the University’s VC (Principal) who engaged them in a discussion about the importance of interdisciplinary research and giving back to communities. About nine participants got to participate in a podcast recording with IPATC sharing their SICSS experiences.
After many sleepless nights, the Cameroon participant joined us on Thursday, 7 July through the generous assistance of Chris Bail, Meghann Norden-Bright and the SSRC. Although the participant faced numerous challenges while stuck at the airport in Cameroon, they made use of the virtual platform to join lectures and meet with their group.
On the final day, groups delivered their project presentations. We agreed not to record the presentations. We were joined in by Sarah Mulaji from UCT who provided extensive feedback to the groups. The presentations were impressive and we were bowled over by the kind of progress all the participants made. Each presentation was followed by a certificate handout ceremony as requested by the participants. Certificates were provided by UJ.
The participants then sat through a final presentation by Dr Sibongiseni Thotsejane who took them through the different ways of engaging communities through their research. She also addressed issues of self-care, well-being and reflexivity that researchers should be conscious of.
Then it was time to celebrate and relax. JIAS organised a Braai and some wine at the JIAS campus overlooking a spectacular view of Johannesburg. On Saturday, 9 July 2022, the participants departed except for the Cameroon participant, who departed on 12 July 2022.
The participants are still using slack to communicate. Although convenient and cost-effective, we can’t stress enough the importance of having face to face sessions. An online-only summer school is not the best option, as digital fatigue quickly sets in and there is a risk of participants dropping out. A hybrid approach not only helps reduce costs but it also allows for a more flexible approach to summer school design. The three-week period may have been long and challenging overall, but for this beginner group, it proved fruitful. It may be possible to change the approach to one week online and two weeks face-to-face to avoid the difficulties of week two, even if it increases the cost.
SICSS-Jogja 2022, hosted by Universitas Islam Indonesia, is Indonesia’s first SICSS partner site. The main event took place between July 18 and July 29, 2022. As it was also the first time, we (Firman and Ahmad) had run a partner site, we learned from other partner sites by reading post-mortems and reflecting on a personal experience as a participant in another partner site, particularly for Firman, who attended WICSS-Tucson in 2020. We adopted some approaches, incorporated some takeaways, and made some local adjustments in running SICSS-Jogja 2022.
In the following, we will describe some highlights in SICSS-Jogja, what went well, what could be improved, and our reflection that might help future SICSS events.
Following Salganik’s lead, we believe that great participants will result in a fantastic SICSS experience. As a result, we used a variety of tools to publicize SICSS-Jogja and recruit as many potential participants as possible. With the assistance of the TAs, we created a poster containing important information about SICSS, such as the dates, speakers, and website. Then, we distributed this poster and additional information about SICSS, such as what computational social science is and what it can do, via email, Twitter, Instagram, Facebook, WhatsApp, and Telegram, among other channels. Our efforts were noticed by Indonesian podcasters, who invited us to be guest speakers and promoted the event on Apple Podcasts, Spotify, and other social media channels.
Since the number of PhD holders and PhD students in Indonesia is relatively low in comparison to other countries who also run SICSS sites, we modified the eligibility criteria slightly. We invited master’s students, researchers working for research institutions or government agencies, and lecturers with a PhD less than three years old or no PhD to apply, regardless of tenure status. Participants were required to submit their applications through Qualtrics, which we accessed via our university’s subscription. Qualtrics, unlike Google Forms, does not require applicants to own or create an account in order to attach their materials. In addition to CV, personal statement, and writing sample, we also asked some demographic questions to participants. We used these questions as a means to evaluate our outreach efforts and application process.
Our application form was filled out by a total of 185 applicants, but many did not complete all the questions, nor did they submit the requested materials. As a result, we were able to identify 85 unique applications from four countries (Indonesia, the Philippines, Malaysia, and Hungary). Then we discovered that not all applicants met the eligibility requirements. For example, we received applications from participants who work in industry or have a PhD that is more than three years old. We also received a significant number of serious inquiries from participants who are still undergraduate students. After eliminating all the ineligible applicants, we were left with a shortlist of 43 candidates to consider in the next round of selection.
Given our limited resources, we hoped to accommodate 20 to 30 participants at SICSS-Jogja. As a result, we classified these shortlisted applicants as either unconditionally accepted (18), conditionally accepted (9), or rejected (16). We saw potential in those who were conditionally accepted, but there were some aspects of their applications that needed to be improved, so we asked them to revise their applications before allowing them to participate in SICSS-Jogja. Unfortunately, 3 out of 9 conditionally accepted applicants failed to revise their application, so we ended up with 24 invited participants in SICSS-Jogja 2022. The following table summarizes the demographic information of all applicants and invited participants in SICSS-Jogja 2022.
|Gender||Female: 39 (46%)
Male: 44 (52%)
N/A: 2 (2%)
|Female: 16 (67%)
Male: 8 (33%)
|Disciplines||Computer Science, Informatics, or Information Technology: 42 (49%)
Math, Statistics, or Engineering: 14 (16%)
Communication Science: 6 (7%)
Economics: 4 (5%)
Education: 1 (1%)
International Relations: 1 (1%)
Linguistics: 2 (2%)
Political Science: 3 (4%)
Psychology: 5 (6%)
Sociology: 2 (2%)
Other: 5 (6%)
|Computer Science, Informatics, or Information Technology: 10 (42%)
Math, Statistics, or Engineering: 2 (8%)
Communication Science: 3 (13%)
International Relations: 1 (4%)
Linguistics: 1 (4%)
Political Science: 2 (8%)
Psychology: 2 (8%)
Other: 3 (13%)
|Status||Master’s student: 7 (8%)
Doctoral student: 18 (21%)
Untenured faculty: 2 (2%)
Junior lecturer: 33 (39%)
Postdoctoral researcher: 2 (2%)
Researcher: 14 (16%)
Other: 9 (11%)
|Master’s student: 3 (13%)
Doctoral student: 3 (13%)
Junior lecturer: 13 (54%)
Postdoctoral researcher: 2 (8%)
Researcher: 3 (13%)
Given the high level of interest among undergraduate students and those working in industry, we recommend that future organizers consider some changes for future SICSS events, particularly in Indonesia. Undergraduate students, for example, can apply and participate with a reference letter from their professors. These letters serve as proof that the students have potential and will most likely be able to follow the materials. Those working in industry can also apply and participate if their companies agree to sponsor or donate to SICSS. Donations can be in the form of money, which is literally required to run the site, or in the form of research collaboration, which can contribute to a greater good for the public.
About a month before the main event, we began the onboarding process. We held a Zoom meeting to discuss this. Using a proportional randomizer code written in R by Firman, we divided the participants into six groups. This ensures that there are some people with computer science or engineering backgrounds and some people with social science backgrounds in each group. We believe that peer learning is one of the most effective methods of learning. As a result, we encouraged participants to first ask their peers in their group if they had any questions about social science or coding before approaching the TAs or organizers.
In addition, we created two WhatsApp Groups. One for organizers and TAs only, where we communicated, coordinated, and discussed almost everything about SICSS-Jogja. Another for all SICSS-Jogja participants, as well as the organizer and TAs. We used it as a communication channel, as well as to share some opportunities. We believe that the participants have their own WhatsApp groups through which they can coordinate. We did not use Slack because people in Indonesia are more familiar with and prefer to work and communicate with colleagues through WhatsApp Groups. We acknowledge that using this channel will prevent participants from getting exposure from other SICSS participants in other partners’ sites. Thus, we still invited participants to the Slack group even if we did it after the event. Additionally, we also asked participants to join the SICSS Facebook group.
In general, we divided the talks into three categories: main talks by either Firman or Ahmad, guest lectures/workshops, and lightning talks also by guest speakers. The first two ran in Week 1, while the third ran in Week 2.
Week 1 was mostly made up of lectures and workshops. Day 1 began with an introduction to computational social science and its ethics, delivered by Firman. In the afternoon session, Ahmad presented Surveys in the Digital Age, combining the third chapter of Salganik’s Bit By Bit book with his personal experience in doing survey and big data research.
On Day 2, Jason J. Jones gave lectures on how to design and run digital experiments and where to publish them. He also gave a hands-on experience of participating in an experiment study run on Qualtrics. Jason, together with the participants, also analyzed the data. In the afternoon session, Arie Wahyu Wijayanto and his student, Salwa Rizqina Putri, taught participants how to collect data from various sources on the internet using web scraping techniques.
Day 3 was an open webinar, which we will go over in more detail later.
On Day 4, Firman introduced text analysis to the participants focusing on both theoretical and practical aspects using RStudio in the morning session. Following that, Dhomas Hatta Fudholi demonstrated machine learning and deep learning to participants using popular datasets such as the IMDB dataset and the Fashion-MNIST dataset in the afternoon session.
Finally, on Day 5, which is the final day of Week 1, Firman introduced participants to agent-based modeling and how to develop one using R in the morning session. Following that, in the afternoon session, Ahmad discussed the basics of causal inference before asking participants to start thinking about their past or current research, or ideas for future research, which they will present the following week as part of the participant-led sharing sessions. He also requested that participants begin working on their group projects, which would be presented on the final day of Week 2.
On Monday, July 18, 2022, we had a welcome dinner after finishing the first day. Ahmad was joined by fourteen other people, including participants, guest speakers, and TAs, for a lively evening in a nice cottage-style restaurant serving a mix of Western and Indonesian dishes.
On Day 3 of SICSS-Jogja, we made the event available to the public via webinar, with the theme of “Making Sense of Societal Issues Through Data and Computational Science”. Two of our guest speakers (Steven S. Skiena and Ismail Fahmi) joined Firman and Ahmad as keynote speakers in this webinar. We organized this event to introduce computational social science to a larger audience and to familiarize them with computational techniques that can be used to understand and reveal social phenomena.
We co-hosted this event with Universitas Islam Indonesia (UII) to commemorate its 79th anniversary. For the publication of this event, we created another poster and distributed it via UII social media channels (Instagram, Facebook, and Twitter), as well as broadcasting webinar information via mailing list and WhatsApp groups. We are overjoyed to announce that we received 411 registrations from people in 10 different countries expressing their interests in attending this event.
The number of questions that sparked discussions between participants and speakers during the webinar demonstrated that participants were very enthusiastic about the webinar. The webinar concluded with a quiz with prizes, which the participants eagerly participated in. This webinar’s recording is available on YouTube at the following link.
We spent more time in week 2 allowing participants to discuss, work on, and present their group projects. Each day, we had one lightning talk from our guest speaker in between.
The main goal of lightning talks is to expose participants to various ideas of research in computational social science in various settings. As a result, we invited those from research institutions and industry.
The first lightning talk was given by Nurvirta Monarizqa, a data scientist at Microsoft, Seattle. She shared her past research on spatial analysis and urban data science. The second lightning talk was given by Claudia Flores-Saviaga, a research fellow at Carnegie Endowment for International Peace. She shared her ongoing research on the collective actions systems to mitigate disinformation. The third lightning talk was delivered by Alëna Aksënova and Sandy Ritchie, linguists at Google New York and London, respectively. They shared how to build better speech recognition for everyone. All lightning talks took place in the morning, with the exception of the final one, which took place in the evening Indonesia time due to the speakers, organizers, and participants being from at least three different time zones (US, UK, and Indonesia).
In general, all lightning talks went well. Both participants and the speakers were happy. Participants get to know a wider application of computational methods. Speakers were happy seeing the interest of participants who asked lots and thoughtful questions.
We should have had another speaker who, unfortunately, had to cancel his talk for personal reasons. We tried to find a replacement at the last minute but couldn’t find anyone who could fill his spot. We then decided to spend the time doing group activities.
We joined our sister site, SICSS-Taiwan, who organized the SICSS Asia Pacific International Panel, in the afternoon session of Day 7 on Tuesday (July 26, 2022). In this international panel, Ahmad represented SICSS-Jogja alongside four other speakers, all of whom are SICSS 2022 organizers: Olga Boichak (SICSS-Sydney), Yu-Hui Chang (SICSS-Taiwan), Lanu Kim (SICSS-Korea), and Jack Qiu (SICSS-Singapore). Here, they shared their experience in organizing SICSS at their sites, their research, and interdisciplinary collaboration situations in their country/region. They also discussed how the SICSS Asia-Pacific region can perform differently or make a unique contribution when compared to other SICSS regions.
We have always emphasized that at SICSS-Jogja, there are many different ways to learn. In this regard, we, too, hoped to learn from the participants, as well as for the participants to learn from one another. In order to facilitate this, we asked participants to share their research in Week 2. This could refer to their previous, ongoing, or future studies. Each participant received constructive feedback from their peers, organizers, and TAs, which can help them improve their current and future work.
Group projects focus on putting the knowledge gained throughout SICSS to use. Participants preferred to remain in their existing onboarding groups rather than form new ones. This activity piqued the interest of the participants. Some even reported working extra hours to complete the activities. While we did not recommend it, given that we had already provided ample time during SICSS-Jogja, it could be an indication of how enthusiastic the participants were.
We were pleased that participants could produce results in such a short period of time. We also encouraged participants to continue working on this project after SICSS. We’ll talk about it after we get back.
As we have stated since the inception of SICSS-Jogja, this event serves as a springboard for future collaboration. We still communicate, primarily through a WhatsApp group, which is, once again, the most popular group communication channel in this region.
We also intend to hold a follow-up meeting to discuss the projects. Some participants stated that they intend to continue working on their group projects, either individually or collectively.
It is also worth noting that many, if not all, participants require a certificate proving their participation in SICSS-Jogja.
Overall, all sessions went well, though one important point to note is that the time difference made it difficult for some speakers and participants. Participants in Indonesia (UTC+7) had to begin at 8:00 a.m. In the United States (UTC-4), speakers were required to begin at 9:00 p.m. Meanwhile, participants in other time zones, such as East Indonesia (UTC+9) and the Philippines (UTC+8), had to begin earlier. And a participant living in Hungary (UTC+2) had to begin at midnight.
Initially, we intended to invite all participants to Yogyakarta and provide accommodation for them. This would have simplified the time difference issue, leaving only the difference between Indonesia and the United States, where Firman and some guest speakers reside. Our resources, however, were limited, so we were unable to provide participants with lodging. As a result, we gave participants the option of participating in Yogyakarta with self-accommodation or joining online. The majority of participants (54%) chose online. Some chose hybrids, in which they participated in person on several days and online on other days. Only six participants joined the event completely in-person.
Unfortunately, there were two participants who could not join SICSS-Jogja almost entirely. During the main event, one participant was confronted with a natural disaster, a flood. Another participant was having technological difficulties because he lives in a remote area of eastern Indonesia. This is out of our hands. Were we able to provide them with accommodations, it might have been easier for them to participate fully. Regardless, we sent them links to the recorded lecture videos and copies of the materials.
We cannot express how grateful we are to SICSS and SSRC, especially Chris Bail and Meghann Norden-Bright, and our local partners for making SICSS-Jogja a reality. Apart from UII, we also secured a sponsorship by Qiscus, a leading Omnichannel Customer Engagement Platform in Indonesia which operates as a B2B Software as a Service enterprise. While we were extremely grateful to all our local partners, the additional funds raised were insufficient to help provide accommodation support for some of our participants, as we had hoped. Given that this was the country’s first event, it is possible that we didn’t have enough exposure and appeal yet to attract more sponsors to help us run the event. With the success of SICSS-Jogja 2022, we hope that if we run similar events in the future, we will receive more support from industry or other organizations, allowing us to raise enough funds not only to provide accommodation for some participants who really need it, but also to provide research grant opportunities, as in other SICSS sites.
Finally, apart from venue, infrastructure, and human resources for technical support during the main event and the open webinar session, we were also able to get mass media coverage through UII’s network. SICSS-Jogja 2022 were reported by several online media such as Beritabernas and Jogpaper, as well as two major newspapers in the region, Republika and Kedaulatan Rakyat.
The first Korean partner site (SICSS-Korea) was co-organized by Lanu Kim, Wonjae Lee, and Jae Yeon Kim (SICSS-Princeton 2019 participant and BAY-SICSS 2020 co-organizer). The institute was held at the Korea Advanced Institute of Science and Technology (KAIST) from June 19 to June 29, 2022, and was co-hosted by the KAIST (School of Humanities and Social Sciences, Center for Digital Humanities and Computational Social Sciences, and Graduate School of Culture Technology) and the KDI School of Public Policy and Management (Applied Data Science Lab). Our goals for our site were to build a community of computational social scientists in and outside Korea and introduce participants to a wide range of computational social science work in academia, the industry, and the government.
In this post-mortem, we describe: 1) advertisement, outreach, and logistical planning for the institute, 2) programming for the first part of SICSS-Korea, which focused on group exercises and skill-building, 3) group projects pursued in the second part of the program, and 4) the reflections from the participants.
Given our goal of establishing an enduring network of computational social scientists in and outside Korea across academic disciplines and sectors (e.g., academia, industry, and government), we first sought to engage with local academic institutions. KAIST is a premier STEM school in South Korea, and KDIS is a premier policy school in the country. We secured funding from both institutions to symbolize broad support for the very first Summer Institute in Computational Social Science in South Korea. It took a while to settle the total amount of funding. From both institutions, we confirmed the amount of funding and the possible usage of it around April, which delayed the specific planning of the summer institute. Despite the pandemic, we decided to host an in-person version of the institute at KAIST because the Korean government tracked and controlled the COVID-19 situation with relative success.
Our process of program development took place in three rounds (speakers, TAs, and participants). First and earliest, we reached out to potential guest speakers before their summer schedules were fully booked. 7 out of 11 contacted people accepted our invitations (64% yield rate). Four are domestic speakers (Korea University, Yonsei University, KAIST, and Naver), and three are international speakers (Northwestern, Code for America, and Twitter). As designed, these speakers are from various disciplines (political science, communication, business, network science, industry data science, civic tech) and institutions in and outside Korea and across multiple sectors (academia, the industry, and the government). After confirming the guest speakers, we recruited a total of three TAs (all graduate students) from Seoul National University (Korean male, sociology), KAIST (Korean male, culture technology), and KDI School (Ukranian male, finance). As for the KDI school TA, Jae Yeon Kim interviewed five applicants and examined their technical knowledge and fit with the program. We recruited TAs to have different programming skills; two TAs used R and one TA used Python as a primary programming language. Finally, to inform potential applicants regarding the application process, Lanu Kim and Jae Yeon Kim held an online information session for prospective participants on March 4, 2022. 61 people from 11 countries RSVPed to this information session (Figure 1 shows these people’s country and institutional backgrounds). We asked these people to share what they would like to know about the application process and the program and explained them the during the webinar held via zoom.
As a result, over some months, we received 56 highly qualified applications from South Korea, the United States, the United Kingdom, and many other places. Lanu Kim and Jae Yeon Kim read these applications, created a long list of 22 candidates, and sent out 18 offers with the expectation of having 15 participants. However, all admitted applicants accepted the offers (100% yield rate). Our criteria for selection were as follows:
Our participants were highly diverse in their gender, country, institution, and disciplinary backgrounds. 11 out of 18 participants were female (61%), and these participants came from South Korea, Japan, China, and Turkey. Even among the Koreans, some of them were currently affiliated with Korean institutions (KAIST, KDI School, Seoul National University, Korea University, Pohang University of Science and Technology, Chungnam National University), and others with the US (Harvard, Columbia, Michigan State University, UC Berkeley, UC Davis, Washington University in St. Louis) and UK institutions (Cambridge). The participants research a wide range of subjects (e.g., political science, sociology, public policy, international studies, industrial and management engineering, business, economics, psychiatry, city and regional planning, and cultural technology) and are at various stages in their careers.
We adapted the group activities from SICSS-Duke with a focus on active learning; The schedule for our program is found here. We applied a flipped classroom method. We strongly encouraged the participants to watch the video lectures from the SICSS website at least one day before the relevant group activities. Each session was divided into a lecture and a group activity session. In the lecture, an institution briefly walked through the main concepts related to a particular theme (e.g., data ethics). The lecture was designed to be interactive, so the participants could think critically about the main issues. During the group activity session, the participants were divided into small groups and collectively solved part of a research design puzzle (e.g., data collection). We emphasized that each group activity is a building block of computational social science research. We acknowledged that the participants came from diverse backgrounds. For instance, some students have high conceptual knowledge but lack technical training. Other students are in a reversed situation. To provide enough training for every student, we asked a member with the least experience in programming should be in charge of typing (typing, not programming) code for a group project so that even the least skilled member could make a contribution and learn from the experience. From a pedagogical perspective, a sense of belonging is as important as skill acquisition as it relates to participants’ self-efficacy. Finally, we believe that every curriculum design is imperfect and needs to adapt to participants’ various and changing needs. So we used a simple form of survey called KSS (KEEP, Start, and Stop). At the end of each day’s activities, we asked participants to fill out a survey that asked what they would like us to keep, start, and stop. The next day, we debriefed their responses and made necessary changes. This format increased the institute’s responsiveness.
In addition to seven guest lecturers, we planned one domestic and one international panel and a lightning talk session. One guest lecture was canceled due to the speaker’s schedule conflict. Each guest lecture lasted one and a half hours, followed by a break and a one-on-one meeting session. Three guest speakers came from academia (Korea University, Yonsei University, and Northwestern), two from the tech industry (Naver, the Google of South Korea, and Twitter), and one from civic tech (Code for America). We asked the speakers to focus more on the “process” of their research rather than the “output” so the participants could learn from the complex decision-making process involved in highly impactful research. During the one-on-one meetings, participants could network with these guest speakers, ask about the speakers’ research, and receive feedback on their own research and advice on their career paths. The domestic panel included Inbok Rhee (KDI School, moderator), Jaesung Choi (Sungkyunkwan University, panelist), and Jae Yeon Kim (KDI School, panelist) and discussed the strengths and weaknesses of the Korean data infrastructure for empirical social science research. The international panel included Lanu Kim (KAIST, moderator), Cecilia Liu (SICSS-Taiwan, NYC-based tech lawyer, panelist), Tiago Ventura (incoming Georgetown AP, NYU postdoc, and previously Twitter researcher, panelist), and Jae Yeon Kim (KDI School, panelist) and discussed academic and non-academic career paths for computational social science PhD students. We also held a lightning talk session in which a presenter needed to talk about their research projects in 5 minutes. 12 participants and all of the co-organizers presented at this event.
Our site took a systemic approach to participant projects by structuring the project development process. We provided ample opportunities for participants to find best teammates aligned with their research interests with complementary skill sets and develop their group project based on the following step-by-step guideline:
Day 2: Group project overview + Research speed dating We introduced the concept of a group project by using a startup incubation analogy. We also highlighted the benefits of initiating a project (e.g., starting a project that requires diverse perspectives and complementary skillets). We initially matched participants to projects based on the list of the projects that some participants proposed and all participants’ ranked preferences and a simple clustering algorithm. Participants confirmed their participation in a particular project through deliberation. At the end of the process, five projects were created.
Day 3: Scoping the project We encouraged participants to scope the project so that they could deliver a minimal viable product (e.g, one or two main figures) by the end of the institute.
Day 4: Data collection We recommended participants to check data availability and other feasibility issues by Day 4.
Day 5-6: Preliminary data analysis During these two days, participants fully focused on developing preliminary data analysis.
Day 7: Data communication One day prior to the presentation day (Day 8), participants focused on sharpening and polishing data communication strategies.
Day 8: Demo day (feedback) On the presentation day, participants received feedback on their research from the organizers and other participants.
We provided moderate financial assistance to three selected teams based on the quality of their research proposal and preliminary results through the KDI School. Each of these three teams received KRW 1,000,000.
Based on the exit survey, we knew that the SICSS-Korea was quite successful in multiple quantitative and qualitative measures. 12 out of 18 participants filled out the exit survey. Regarding the question on the program’s quality, 75% of the participants said they are “VERY POSITIVE” in recommending the SICSS-Korea to other colleagues and friends. One participant commented that they learned from the SICSS-Korea “Many things. Each lecture and guest talk was eye-opening, and group practices and group projects taught me the practical difficulties and concerns that are difficult to learn from textbooks. I really learned a lot from informal conversations with the organizers and participants as well!” The same participant also commented that “ (1) I want to write a paper using one of the methods I learned through SICSS by learning more about it. (2) In the near future, I hope to learn more to be confident and knowledgeable enough to offer an introductory course on computational social science at my home institute.” Another participant expressed a similar enthusiasm: “I would pursue to grow to be a part of css fields and follow up-to-date fast-changing trends of computational methods.” All in all, the SICSS-Korea inspired a new generation of computational social scientists in and outside Korea and a strong network among them that could last even after the SICSS-Korea is over. We believe that our participants and their experience with the SICSS-Korea are the most valuable impact produced by the program.
Organized by Qiwei Han and Filipa Reis
The SICSS-Lisbon program is a partner location of SICSS 2022 in Lisbon, Portugal as the second edition. It is organized by Qiwei Han and Filipa Reis and co-hosted by two leading Portuguese business schools Nova School of Business and Economics (NOVA SBE) and Catolica-Lisbon School of Business and Economics (CLSBE). As the Covid risks eased, the program is held in person from June 19th to June 29th at Nova SBE campus for the first time. Continuing with the program initiated since 2021, the focus of the Lisbon program is to nurture a community of computational social science (CSS) researchers in Iberia and neighboring areas, equipping them with essential skills for CSS research.
In this post-mortem, we share our hosting experiences and the whole process in the following sections: 1) advertising and application process; 2) logistics issues and onboarding process; 3) main programs; 4) group projects; 5) extra activities and post-departure.
As this is the first-ever SICSS program held in Iberia (Portugal and Spain), promoting the program in the area is the top priority to attract the great participants that are interested in CSS research. At the beginning of the year, we have received support from the host schools to help advertise the program to the local Portuguese schools. Meanwhile, we also reach out to other schools through organizers’ professional networks in Europe. Although later many participants reflected they applied to the program because “SICSS is quite well-known”, we realized it is not the case in Iberia, that SICSS is much less-known compared in other regions. Thus, we decided to extend the application deadline to April 7th, and at the same time to circulate the message through social media, email lists.
All applications are managed through a customized Google Form and are reviewed by both of the organizers. The applications are reviewed by both organizers and decisions were sent on April 21st. In total, we have received 25 applications and admitted 17 of them. Those who are not admitted mainly have submitted insufficient application material or do not meet the requirement (we are looking for participants who are currently enrolled in the Ph.D. program, post-doc, or pre-tenured faculty). However, 3 of them decided to withdraw due to personal reasons, as this is quite common this year when the restrictions like travel restrictions limit people’s schedules. As a result, we formally welcomed 14 participants with very diverse backgrounds. They come from 12 nationalities from U.S., Europe, Middle East, China, and Africa, 14 universities, and 5 major disciplines, including Communications, Sociology, Business, Economics, Political Science. However, this year we only have 4 female participants and we will strive to make a more balanced group in the coming years.
All program days are scheduled to start at 9 am Western Standard Summer Time (WEST), which potentially posed challenges for several participants from the U.S to participate, although a majority of them are in various European countries. However, our participants managed to participate in the full program to overcome the time zone barrier.
We have recruited two TAs to help with logistic issues and the onboarding process. Both TAs have prior teaching experience in the data science courses and thus provide good help throughout the program. All participants are notified of the pre-arrival preparations right after they are accepted. First, they are required to submit the bio and profile photos that are added to the SICSS website. Second, they are provided with instructions such as how to join the Slack channel #SICSS-Lisbon and to follow the instructional materials from the SICSS website. According to the application material, all participants have some programming experience, while a few have quite good R knowledge. For those who need to practice R programming, our TAs have set office hours to work with participants for individual concerns. In particular, participants reflected that they find the instructional material provided by SICSS are very helpful and allow them to jumpstart the program later.
We also hosted the welcome session the day before the program at the restaurant on the Carcavelos beach next to the campus. All attending participants have the chance to get to know each other and break the ice. Meanwhile, we brief the logistic details of the program to the participants during the welcome session.
For the participants attending remotely, we followed the experience from the SICSS-Duke 2020 as one organizer was the participant there. We created the Zoom link for each day of the event separately, including both morning and afternoon sessions, guest talks as well as virtual social activities. The feedback forms and daily notes are created in the form of Google doc and have been put together in a shared Google excel that all participants have access to.
Thanks to the support from the school, we are able to provide free lunch and two coffee breaks during the whole program. Meanwhile, we managed to invite a machine learning professor to give guest talk and lecture through the organizer’s professional network. Prof. Jinjun Xiong Empire Innovation Professor with the Department of Computer Science & Engineering, University at Buffalo to talk about “The Flashy and Earthly Sides of AI”.
As the SICSS program has universally adopted the “flipped classroom” model since 2020, we have followed the main program at Princeton with the same teaching schedule. Participants are required to watch program videos before each day (if available) and work together on the group exercises. Organizers and TAs then help them with the challenges on site.
Furthermore, we have invited guests that are faculty members or former SICSS participants to attend afternoon social hour to talk about a CSS-related topic, in order to make the coffee break more productive. The schedule is as follows:
|2||Joao Baptista, Professor, Chair in Information Systems at Lancaster University Management School||Qualitative Research|
|3||Daniela Schmitt, Assistant Professor of Quantitative Marketing at Nova SBE||Academic job market tips|
|4||Dan Tran, Invited Assistant Professor, Católica-Lisbon School of Business, SICSS-Lisbon 2021 alumnus||Post-SICSS experience|
|5||Jinjun Xiong, Professor, University at Buffalo||How to turn to AI research|
|6||Rodrigo Belo, Professor of Information Systems at Nova SBE/Rotterdam School of Management||Causal inference|
|7||Maximilian Kaiser, PhD student at Frankfurt School of Finance and Management||Large-scale data analysis|
|8||Bernardo Costa, PhD student at Nova SBE, SICSS-Lisbon 2021 alumnus||Post-SICSS experience|
Given the well-established program schedule of the Princeton site, generally, we have detailed instructions on each day’s learning materials as well as exercises. In the first two days, participants joined the group exercises through random assignment that takes into account of online presence, i.e., one team has one online participant and rotates every session, for the purpose of getting to know others in both online and offline mode. Following the last year’s experience, we asked participants’ confidence in the subject in order to perform skill matching program but it turns out that participants are well-equipped with necessary skills and they are flexible with working with each other.
Both organizers and TAs joined all sessions with the participants, and we can tell that the in-person collaboration is much more effective than the last year’s program with all virtual participation. The small talk during the social hour also help enrich the discussion between sessions. It would be great if the in-person attendance can be enforced in the future, if COVID-19 situation completely is resolved.
Similar to last year, we also encouraged all participants to share their interests in special topics that are discussed during the lunch break. Essentially, we ended up with just one lunch topic per day. In fact, this practice becomes less effective than we did last year, mostly because participants tend to exchange ideas during the lunchtime. Most of the days we ran the daily activities independently. On the second day of the first week, we worked on the survey design and used the Amazon Mturks to collect the data. However, the Mturks made some changes, which makes it harder to work with and pay the workers. It would be necessary if SICSS could update the instructions according to the updates of Mturks or consider using some other crowdsourcing tools, such as Prolific, which is more friendly to European sites.
On Friday, as the Fragile Families challenge is no longer available, our invited speaker Prof. Jinjun Xiong gave the lecture on the introductory machine learning and designed the challenge of recognition of sign language. Overall, the participants find the lectures and challenges very useful for practicing their programming for the related CSS topics.
Following the last year, the participants mainly work on the group research project during the second week. We have created a spreadsheet for participants to express their interests, and then participants further discussed in the most similar and most dissimilar groups. In total, there are 3 topics that are initiated by the participants, and 2 of them end up with a research presentation with preliminary findings.
The group projects are presented and discussed on the second week’s Wednesday afternoon. One topic analyzed the public debate script in European parliament and participants are able to identify the groups of parliament members that agree or disagree. They show the parliament members have formed several clusters consisting Southern European countries and Northern European countries. The second topic studied the determining factors of success in crowdfunding projects. They also retrieved the social media data from Twitter API to show that the engagement between project owners and patrons is positively associated with the project success. Both have enjoyed the presentation as well as the discussions afterward, attended by two former SICSS participants. Last, all participants have attended the farewell dinner together.
Meanwhile, all participants are encouraged to attend the tutorials offered at IC2S2. One team of four participants and TAs also participated in the Datathon to work on the COVID forecasting problem.
We have prepared site swag for the participants have received their swag but it does not arrive on time. We instead decided to mail the swag to them. Meanwhile, we have prepared for the participation certificate signed by both organizers. In Europe, many universities may need the proof to cover the expenses. As this is the first year of on-site program, we find that having such a program at a nice location would be more useful than virtual format, because the face-to-face communication would be a key component of the program as well. In the next year, we hope we could keep promoting the SICSS in Iberia to have more participants and a growing CSS community.
We’ve divided the post-mortem into 5 main sections: 1) outreach and application process; 2) pre-arrival and onboarding; 3) first week; 4) second week; 5) post-departure.
Three of the five organisers from last year’s SICSS-London (Joshua Becker, Nicola Perra & Mike Yeomans) were enthusiastic about running it again this year - we were also the most active last year so this was encouraging as a start. Mike took the lead this year, having arranged support for a location and funding from his host institution (Imperial College), though Joshua also found some extra funding from his school as well.
As before, we advertised our event to a large, diverse group. Our major outreach effort began in February. Our site was scheduled a bit earlier than other sites, since London takes a different tenor in July, with most residents clearing out and tourists pouring in. We also had an application deadline that was a bit later than many sites (end of March), to make sure we could pull from the largest pool. We emailed former participants and former speakers. We also advertised through professional societies and other colleagues across London. Finally, we tried to reach potentially interested participants through social media, email lists, and emails to faculty that we thought might know interested participants.
This year while reviewing the applications, we did not ask for letters of recommendation, to lessen the burden on our community. We received many high quality applications - almost 100! We heavily prioritised students from schools in or near London, with a few people with strong CVs from out of town who were passing through in June. We had excellent diversity for gender, ethnicity, home institution, topic of interest, and expertise. Our students were a bit younger (mainly early/mid-PhD) than last year, though this may have been because we took the best senior students last year. The choices were not hard - each co-organiser did an initial triage through 2/3 of the list ( so everyone would be seen twice) and we ended up selecting a few more than planned - 24 total, though two dropped out due to other conflicts. We deliberated on the full list, and had almost no conflict on who we chose. We did not need to make extra efforts to select diverse applicants because the initial pool was so varied. The students were fabulous.
After participants are accepted we begin to onboard them into the program and provide them with pre-arrival materials. We did all the standard onboarding stuff - the main SICSS survey, the website profile uploads, slack, a site-specific survey on topics of interest, etc. The one tricky bit was that we had people from out of town who had some work to do arranging accommodations. We did not provide this, though if we had more funding in future this might make it easier to recruit more broadly. That being said, because 3/4 of people already worked and lived in London, that was not such a big deal. People from out of town were encouraged to work together to share ideas on where to stay.
Because Mike is at Imperial, he took the lead on arranging most of the logistics. This posed a lot of (solvable) challenges that didn’t come up during the virtual event last year. Lots of conversations with admins on campus about rooms, funds, meals, etc. We had some trouble securing some of our first-choice keynotes but eventually got a really strong mix of people, including one non-academic (Alex Sutherland from BIT) and some other local profs from UCL and ICL (Neave O’Clery, Yves-Alexendre de Montjoye, and Stephen Hansen). The topics were broad (behavioural science & policy; geospatial analysis; data privacy; and text-as-data) and students had great questions for all of them when the time came. In retrospect, we should have been more proactive about deadlines for keynote commitments, and could have started this process earlier (Joshua provided helpful nudges and suggestions throughout). Offering small honoraria to all of the keynotes was helpful though, as their time is quite valuable, and as a token of appreciation it was well-received.
We offered office hours via our TA (Burint Bevis, one of Mike’s PhD students), though in practice people did not use this before the event (he was very helpful during the event though). We shared lots of tutorial info on getting R and Python ready to go. We taught in both, as we had different expertise in our organising team (Joshua & Mike were R people, Nicola and Burnit were Python), and managing different languages ends up being a key part of many CSS workflows! Students really appreciated this diverse expertise, and many projects drew on both languages in the end.
The first week followed the typical format. We arranged five teaching days based on our relative expertise and availability - Monday ethics (Mike); Tuesday NLP (Mike); Wednesday APIs & Scraping (Joshua & Nicola); Thursdays online experiments (Joshua) and Friday networks (Nicola). The division of labour was very smooth - Joshua mostly worked from the SICSS base materials but Mike & Nicola used teaching tools from their own classes. Everyone had code examples and datasets prepped for each day, and students built on these resources throughout the whole camp. This was all a success, I think! It also let us be more flexible about time commitments, so we could duck out for meetings, other responsibilities, taking a day off here or there… Mike & Burint were there every day, and Joshua most days (but was always available via slack/whatsapp as needed). Nicola had to be in Italy for some of the session, but was essential when he was there as he had expertise we couldn’t replicate.
We prioritised group work and mini-presentations that first week. The hackathons and breakout groups were the most invigorating experience for students, as they got to mix and match with many other students over the week. This also allowed students with expertise in the domain of the day to teach the others - arguably they learned more from one another than from us instructors! We really tried to reduce the amount of lecturing and instructor-led discussion compared to last year. This was arguably much easier in-person than on zoom. It also felt more collaborative among the instructors, as we each had focused lead time on different days, rather than the rest of us slotting around a single leader. The first day had a lot of get-to-know-you exercises in breakouts, since ethics isn’t really a technical topic, and that was a great way to start the discussions and get people learning about one another’s interests and abilities.
Imperial College sponsored meals and that was great for getting students to spend more time on campus and with one another. Rather than balancing everyone’s dietary requirements and pre-ordering food, we bought stacks of vouchers that were redeemable at locations across campus - this ended up working very well. It also gave more time for people to walk around outside of the classroom and get to know one another more informally. We kept each day’s schedule to 10-5, to allow for travel time, socialising, etc. And mostly kept punctual (with a bit of mid-day flexibility).
In the first day of week 2 we did the speed dating group matching process. We used a google spreadsheet for people to add research interests and then clustered them into maximally similar groups and maximally dissimilar groups. We did the same with their self-reported skills expertise from the onboarding survey. Many of these conversations turned out to be generative of project ideas.
We were more open-minded about attendance - we had a tube strike one day, and other folks had a class here and there, or other disruptions and we had one or two students virtual each day. This worked out great in the MBA classrooms, as we could zoom them in easily. During the flex hours, the instructors who were there that day mainly cycled between the different groups to provide advice, coding tips, etc. Students were quite busy each day working in their pods.
Our keynote speakers were mostly in the second week and this was great to break up the long flex hours devoted to group projects. We solicited ideas for breakout tutorials and got four - three from students with special expertise not covered in our first week of teaching (geospatial, large language models, and pre-registration) and Mike did a session Thursday on data visualisation to prep for the group presentations the next day.
A big addition to the week was to bring in four alumni from last year’s SICSS-London for blitz talks. This was a great way of building more community! Each student had a half-hour, and two of them presented projects related to their work from last year’s group projects. It was very positive.
All five groups presented on Friday afternoon. The talks were excellent, and everyone had a lot of questions for one another. Some of the alumni visitors stuck around for this year’s group project presentations which was nice. The work done that week was really impressive. We expect at least a few of these projects will end up as publications and/or long-term collaborations.
After SICSS was over, we followed up with the students who submitted grant applications. We funded most of the projects - basically anyone who proposed something that came out of the camp itself, rather than asking for funding for the normal work they were doing before the camp. The students set up a WhatsApp group to stay in touch (tbh, the slack channel was not as active as we would have liked during the sessions, and basically went dormant after our last day together). We were swag-less, sadly. But I think the connections will outlast any mug. We will keep checking in periodically as we plan for next year (including looking for new organisers) and hopefully encourage more alumni connections if SICSS-London 2023 actually happens.
SICSS-NDSU 2022 took place from June 19 – June 30, 2022. Dr. Shuning Lu, Assistant Professor of Communication led the workshop in collaboration with Dr. Zoltan Majdik, Associate Professor of Communication at NDSU. This was NDSU’s first Summer Institute partnership as well as the first one held by an upper-midwestern university in the United States. Our workshop received additional funding from NDSU’s Department of Communication. This post-mortem will give a detailed account of our application process, onboarding, first week, second week, post-workshop activity, and challenges.
We began our outreach in March 2022. We aimed to recruit a wide range of participants from different disciplines, cultural backgrounds, institutions, and career stages. For recruitment, we sent out a call for applications through NDSU’s Office of Research and Creative Activity (RCA) and Graduate School. We also reached out to administrative staffs of several departments and colleges at NDSU to announce the workshop. To reach out potential non-NDSU participants, the lead organizer further sent the announcements to various program directors and academic friends from a few schools, including University of Minnesota, University of North Dakota, Minnesota State University at Moorhead, University of Texas at Austin, and University of Utah. The announcement was also posted on social media such as Twitter, WeChat, and Facebook.
For the application process, we used the Qualtrics platform. The application asked for name, demographics, email address, institution, career stage, discipline, self-reported frequencies in using text analysis and statistics for research, self-reported skills in R, and propensity to attend the workshop in person. To smooth the application process, we limited the application materials to two documents: (i) a curriculum vitae, (ii) a statement of interest, describing current research using computational social science and/or interest in computational social science for future research (500 words maximum). We believe that dropping the writing samples and recommendation letters made both the application and review processes more efficient, and a CV combined with a statement of purpose proved to be sufficient for us to evaluate and select participants from the application pool.
We received 24 applications from 10 institutions across the United States, Denmark, and the United Kingdom. Among the applicants, 12 were male, 11 were female, and 1 was non-binary. For ethnicities, a majority were Asian (n=15), followed by White (n=7), Black/African American (n=1) and Hispanic (n=1). They came from a variety of disciplines, including communication, sociology, anthropology, economics, education, social work, and public policy. Most of the applicants were graduate students (n=17), followed by early-career faculty (n=6), and postdoctoral scholar (n=1).
The team of two organizers evaluated applications based on applicants’ research and interest in computational social science, the likelihood the applicant would benefit from the experience and contribute to the learning experience of others, as well as the potential for the applicant to teach computational social science to new intellectual communities. A total of 21 participants were selected and 20 confirmed their participation. All non-NDSU graduate student participants received financial support from NDSU’s Department of Communication to partially offset their travel or lodging costs.
We sent our decision letter to the admitted applicants on April 19, 2022. We asked them to confirm in-person attendance by April 22, 2022. In that email, we also provided a sketch of the format and content of the workshop, and the amount of travel grants non-NDSU graduate student participants would receive. Out of 21 admitted participants, 20 accepted our admissions.
In the meantime, we built a slack channel, which was used for quick messages and informal interaction. Participants who committed to joining the workshop were invited to the channel. The channel was not very active during the pre-arrival stage. It warmed up when participants started to come to Fargo, ND and served as a platform for previewing the next day’s schedule, code-sharing, troubleshooting, groupwork, and announcements of informal gatherings during weekends.
On April 27, 2022, we sent out a formal onboarding email to all confirmed participants. In the email, we asked participants to provide their brief bios, keywords about their research interests, Twitter handle/website, and a profile picture. We also included the suggested schedule for going through the learning materials during the pre-arrival stage.
On May 3, 2022, we updated information on on-campus housing for non-NDSU participants. Thanks to NDSU Residence Life, non-local participants could live in studios with affordable prices in a location that was just a 10-minute walk to the SICSS-NDSU classroom.
To help our participants easily adapt to life in Fargo, we curated a list of grocery shops, weekend/evening events, restaurants, and places of interests for them to check out during the days off. We pinned the information in our main Slack channel.
Finally, we launched our social media channel on Twitter to live tweet each day’s events. Please check that out: https://twitter.com/sicss_ndsu
Our first week was composed of five main components – lecture/tutorial, research speed dating, small group activity, participant research presentation, and guest lecture. The first week’s schedule generally ran from 9:30 a.m. to 5:00 p.m., with a lunch break and several small breaks throughout the day. We also fielded daily surveys to collect feedback from the participants and adjusted our module in the following days accordingly.
On Monday, Dr. Stephenson Beck, the chair of NDSU’s Department of Communication, delivered welcome remarks, followed by organizers’ introduction of the workshop and participants’ self-introductions. After the welcome session, we overviewed the field of computational social science and discussed key issues of research ethics in computational social science research. We had our first research speed dating among participants before lunch break. Our afternoon was reserved for small group exercise on ethics developed by SICSS-Princeton. Participants discussed their thoughts on the case study and research ethics in different disciplines more broadly. Before the end of Day 1, organizers worked with participants on setting up Twitter Standard API, which would be used for Day 2’s module.
Day 2 focused on collecting digital data. We began the day with a quick overview and then started live tutorials on web scraping and Twitter API. Our afternoon was devoted to a small group activity that asked participants to develop a research question, collect digital trace data relevant to that question, and describe the data they collected. Groups presented their findings on topics such as trending discourse among U.S. political representatives, virtual linguistic landscape on Twitter, debates of gun control and gun right issues, Twitter discussion on Imposter Syndrome, abortion issues, and immigration policies. Day 2 was concluded with participant research presentation, in which our participants showcased their own recent research projects and stimulated cross-disciplinary conversations.
Day 3 centered on natural language processing. We began our live tutorials on NLP, followed by a guest talk on video as data from Dr. Kaiping Chen during lunchtime. We provided a few datasets for the group to work with for small group activity in the afternoon. Groups presented their word frequency and sentiment analyses on Black Lives Matters, Harry Potter novels, Anti-Asian tweets during COVID-19, and COVID-19 vaccine discussion on Twitter. In the end of the day, several participants presented their own research in the workshop.
Day 4 focused on network analysis. Like previous days, the morning started with lecture and live tutorials on social network analysis in R. During lunchtime, two speakers from NDSU, Dr. Dan Pemstein and Dr. Dane Mataic, visited SICSS-NDSU and shared their research and datasets. In the afternoon, participants collected and analyzed network data from Twitter in the group activity. They presented their findings on BTS fandom discussion, animal rescue, capital riot hearing, football conferences, and gas tax holiday, among others. The day concluded with participant research presentations.
Day 5 was lighter as compared to the previous days. The morning focused on live tutorials on data visualization. During lunchtime, we followed Chris Bail’s approach and conducted cluster analysis to form groups for final projects. In the afternoon, participants joined a campus tour at NDSU. Then, they were divided into the final project groups to discuss potential ideas. Finally, we had our last round of participant research presentations.
Day 6 & Day 7 were weekend days off.
The second week of the workshop was primarily devoted to final group projects. A typical day was composed of lecture or live tutorial in the morning, guest lecture at noon, and group work time in the afternoon. It ran from 9:30 a.m. to 4:00 p.m.
Day 8 covered basics of machine learning, followed by guest speaker Dr. Tian Yang’s talk on network analysis of news consumption behavior. Day 9 centered on topic modeling, with a noon guest lecture by Dr. Scott Graham from UT-Austin. Day 10 discussed deep learning, followed by Dr. Saif Shahin’s talk on combining qualitative methods and computational social science. In the afternoon, participants utilized the small rooms in the building to spend focused hours on their final projects. The organizers provided guidance and troubleshooting to each of the groups. Slack was a helpful tool for participants to call out the organizers for immediate help. Participants were then asked to come back to the main classroom and share their progress and challenges by the end of the day.
Day 11, the last day of the institute, was devoted to group presentations. Each of the five groups gave a 15-min presentation, in which each member had a role. Other participants listened and gave feedback to the presenters. Participants accomplished incredible work during such a short time period, encompassing topics such as COVID-19 vaccine discussion, coffee drinking and stratification, education policy analysis, public perception of AI tech, and Yellow Stone flooding, see here. At the closing of the workshop, we encouraged participants to stay in touch by continuing their collaborative work and conversations. We also encouraged them to spread the word of SICSS-NDSU and hold SICSS workshops at their home institutions.
After SICSS-NDSU was over, we secured additional funding from NDSU’s Department of Communication. We encouraged participants to apply for the post-SICSS-NDSU grants to continue their work during the workshop or embark on new projects by August 15, 2022.
Here are some challenges we encountered from organizing SICSS-NDSU:
“You’ll see, it is as draining as it is rewarding” -(Matt Salganik, about organizing a SICSS).
We organized a Summer Institute in Computational Social Sciences in Paris in June 2022. It took place at ENSAE (Institut Polytechnique de Paris) from June 20th to July 1st. Below, we provide a brief summary of the different stages of the institute. But before this, let’s say that Matt was definitely right: it was both a draining and a great experience.
The outreach had several phases and channels. We advertised applications through several French-based social science listservs. We also posted on computational social science and digital humanities listservs. We used Twitter extensively, on different occasions, as well as LinkedIn. That said, we believe the main draw came from persons looking at the SICSS webpage.
In 2020, we had planned to organize an event in French to reach out to the emerging French-speaking CSS community. We had to cancel it due to the pandemic, and we organized this 2022 event in English, but we still had a massive outreach toward the francophone sphere. In particular, in addition to Belgium and Switzerland, we directed some efforts towards Africa and the Middle-East, via some contacts we had in universities and research centers.
Another flexible rule we decided on early was to select people who could travel to Paris by train/bus. Given that SICSS happens on many continents, we felt that there were enough satellites on various continents (except for Africa, for which we made an exception) for us to apply this rule. We publicized it early on in our application guidelines.
We accepted applications via a google form. We decided to keep this very simple, to just ask participants a dozen questions and a short paragraph about their research. We specifically did not ask for letters of recommendation, which are a burden for the recommender as well as for the selection committee, and are often not so relevant. In addition to some questions about career, we asked about the CSS projects the applicants may have, and their motivation to attend. We also had questions about the knowledge of code, although we did not select on this (we wanted to adjust the teaching to the crowd).
We decided to publicize SICSS-Paris widely, so we placed few requirements on the application process. In terms of discipline, we received mostly applications from the human and social sciences, but also a few from computer science/maths. In terms of career stage, we focused on Ph.D. students, postdoctoral researchers, and untenured faculty, but we subsequently decided to extend at both ends (motivated Masters students, and interested faculties). In the end, we received close to 100 applications, and we selected 22 persons - 5 extra were placed on a temporary waiting list. Find the participants here.
After participants were selected, we sent participants an email asking them to confirm their attendance. To our surprise, absolutely all participants accepted our invitation, which made the waiting list redundant. In addition to these two dozen persons, a handful of graduate students and faculties from one of the schools on campus came to attend some sessions, as we had broadly publicized the summer school. They participated mostly in the evening talks, but some attended other sessions.
One noteworthy issue: two scholars came from countries where a visa was required to enter Europe. Both had serious issues with this. One had her application declined at first, and we had to appeal to the French Consulate. A second one did not manage to have an appointment in due time and had to cancel his trip (which led to unnecessary costs on both sides).
Our onboarding process took place between April and June. After applicants confirmed their participation, we reached out via email to ask them to: (1) provide a short bio and photo for our website, (2) provide information for various logistics (e.g., dietary restrictions, attendance to social events that required booking), and (3) prepare for the summer institute by completing either some social science readings or coding training. We also offered TA help to our participants—through our own TA and the TAs at Duke. And we made sure to send out a reminder email to participants about a week before the summer institute.
4 interns, first-year students from ENSAE, were hired for 4 weeks. They helped us with the logistics, the preparation of course material, and also fully participated in the summer school. They were extremely helpful to alleviate the everyday logistics, which are menial but time-consuming. When they double-checked the code for the training sessions with their 4 sets of fresh eyes and were able to spot a few critical inconsistencies in the script that we had missed before. During the summer school, they sat in class, went to social events, and even did their own project. They also got, early on in their training, an opportunity to get familiar with CSS. They seem to have enjoyed their presence, and so did we.
Thanks to generous grants from various sponsors in France (CREST, Hi Paris), we managed to subsidize some aspects. In particular, we offered to pay for the accommodation, which was located in a new and large student housing facility, with numerous common spaces for cooking or hanging out/working. 16 out of the 22 participants stayed there, the others either living in Paris or staying with friends in the area. We also paid for 2 dinners (welcome, goodbye), and a series of goodies (tote bags, reusable cups and water bottles to limit waste, pens, and notepads). In addition to this, we offered grants to cover travel and daily expenses to students who specifically applied for them. 4 of them benefitted from this. Overall, we are delighted we were able to do this, as it seems that it enabled some to come, and it alleviated some financial stress from many others.
Before arrival, the figures for the new covid wave were on the rise in France. We emailed the participants to be safe, and we suggested that they wear masks and use hand sanitizing lotions whenever possible. We made them available every day.
This did not prevent Covid to make a small appearance in our institute. Two persons were sick, one had to skip the entire institute after two days, and the other one was able to return after a few days of isolation.
Once we were notified that we had cases, we took some extra measures. We moved from a classroom to a large amphitheater; we urged participants to wear masks; we provided antigen-tests; we canceled some conviviality moments and suggested that lunch be taken outside. We also broadcasted every lecture on Zoom so that students could participate remotely. To our surprise, the epidemics did not spread any further, so we were able to continue with the summer school and we even maintained some social activities in the second week.
Throughout the 2 weeks, we structured the institute in 3 moments: morning lecture, practice in the afternoon, and guest lecture in the evening. The full program is available here, and below is a summary:
Morning: Everything scraping & data cleaning (E. Ollion, 4 days); Text analysis (G. Gauthier, 2 days). In the second week, we had invited lectures: Text analysis (L. Girard, 1 day); Agent-Based Models (G. Manzo, 1 day); Experiments in the Social sciences (M. Keuschnig).
Afternoon: 1st week: labs about the morning session (F. Lennert). All the materials (R notebooks) are available here.
Evening: 1h invited talk
The evening talks were followed (except during the time of covid crisis) by drinks served on the terrace of the building, which allowed the participants to further socialize and chat with the guest speaker.
Overall, we believe that the program worked well. As could be expected, not everyone was interested in all topics, some knew about some aspects, but the evaluation form we circulated shows that everyone got a lot out of the institute.
Ratings from participants to SICSS (yes, there were other options, including negative ones)
Still, adjustments could be made. Some have to do with teaching, topics, and packages and don’t need to be detailed here. The main change we will consider has to do with the schedule, which was maybe packed. The days were long, and if this has its plus sides (we cover more, more in-depth), it is also a bit intense. One suggestion we received was to remove some lectures from the second week, or maybe some evening talks (esp. since the latter tend to cut into the afternoon’s group work).
With a bit of nudging, groups were formed quite easily. They covered a gamut of topics that ran from sexism in rap songs, partisan bias in news consumption, coverage of climate change in newspapers, success of scientific proposals in being funded, effects of #MeToo implicated actors on the outcome of movies, ….
On the last day, the groups presented their ongoing work, and we gave them some comments. At the beginning of the summer school, we were approached by the editor of a European social sciences journal who said they would be interested in publishing a special issue on CSS, potentially with the papers produced by the different groups of the institute. We announced this early on, which had a positive effect on the overall motivation. We are now in the process of collecting abstracts to send a proposal to the editorial board.
SICSS-Roma Tor Vergata was organized by Laura Brandimarte (participant at SICSS-Lisbon 2021), Simone Borra and Gustavo Piga at the University of Rome Tor Vergata in Italy from July 18th to the 29th. Additional support and experience were provided by Alessio Huma in the early stages, and later on by Niranjan Nair (Pachu) and Vlad Simon as TAs during the sessions.
This post-mortem is broken down into the following sections: 1) advertisement, application, and acceptance; 2) pre-arrival and onboarding; 3) the first week; 4) the second week; and finally, 5) concluding remarks.
Applications for SICSS-Roma Tor Vergata opened on February 4 with advertisements on Twitter and university mailing lists – specifically, to undergraduate students of the Global Governance program within the Department of Economics and Finance. We believe we are the first location to experiment with applications open to undergraduate students. We hoped and expected that this would lead to as much diversity as possible in applications across Italy, but we were positively surprised by the many applications from all over the world.
We received a constant stream of applications until the deadline (April 7) and evaluated a total of 39. From these applications, we accepted 24 participants, coming from very different backgrounds (economics, sociology, psychology, development studies, political science, international relations, urban studies, tech policy, data science, medicine) and locations (Italy, Switzerland, Germany, UK, Denmark, Turkey, US, Brazil, Australia). Two applicants were not able to participate due to visa issues, so the final number of participants was 22.
Although we did not originally expect applications from so many countries requiring an entry visa to Italy, we anticipated this may be an issue for a few applicants. However, given the mostly educational nature of our Institute, targeted to participants who had little to no experience with R, we decided to hold our sessions exclusively in person, since learning a programming language from scratch usually causes all sorts of problems (some hardware-related, some related to hardware-software compatibility issues) that would have been very difficult to solve in a distance-learning format. We also figured that participants who were comfortable with learning the basics of programming online could take advantage of the excellent material available on the SICSS website (and other sources suggested therein), and the value added of our Institute would reside in in-person interactions.
Most participants were able to come to campus for the full two weeks, but in certain circumstances we accommodated requests of remote connections. Indeed, given the fast spread of COVID in Italy during the summer of 2022, we encouraged participants to stay at home if they experienced any flu-like symptoms or fatigue caused by the sever heatwave that lasted for the entire Institute. Wearing a mask in class was mandatory for everyone unless they had a medical condition that did not allow them to do so. We were truly appreciative of the understanding by all participants that health and safety were our first priority, and of the compliance of everyone to our rules.
The guiding principle we applied in accepting participants was based around the content of the virtual event. We looked to accept participants where their background and CV suggested they would benefit from learning computational science skills. To this end, we did not accept applicants who had career experience and advanced skills in data science, finance, or physics, and whose CV suggested significant prior experience in R, or computational social science broadly interpreted. We also tried to give students who were still in their undergraduate studies more opportunity than those who had completed their studies or were close to doing so. This meant that we had to reject some strong applications that were from senior academics who felt that, as a growing field, computational science was something they should learn more about. While they were not a suitable fit for SICSS- Roma Tor Vergata, we engaged with them further to expand the field of computational social science in Europe.
Given our target population of mostly R-newbies, we limited our requirements for onboarding to the collection of information required by SICSS and ourselves, as well as the installing of R and RStudio. We made use of our own platform to collect participant information to ensure that we adhered with privacy regulations in Europe and shared this information with the core SICSS team as was necessary after having received consent from participants. As standard procedure in the SICSS community, we required participants to submit biographical information and photographs as part of their acceptance to the program and posted all information to our page on the SICSS website.
Where requested, we also coordinated room reservations on campus, or provided suggestions for alternative accommodations to participants who preferred to stay downtown (as campus is not centrally located).
We met every day, Monday through Friday, from 10am to 1pm, take a 1-hour break for lunch, and resume 2-5pm.
By design, we expected that the vast majority of applicants to SICSS-Roma Tor Vergata were unlikely to have had much prior experience in R and that most will have had little formal experience with programming. While it has previously been a pre-arrival task for participants to learn R themselves, we felt that participants would benefit more from instructor led training in R than they would from self-study.
We therefore created slides based on the content of the R for Data Science book and spent the first week teaching the material via live lectures and giving participants many opportunities to pair up and work on exercises together. While we advertised this first week as an R bootcamp for those who had no experience with R, some more experienced participants joined us nonetheless from day 1. From the 2 surveys we ran in week 1, we could see that managing this diversity was quite challenging: the pace was good for R newbies, but more experienced participants would have preferred more advanced exercises and a faster pace. For the future, we plan on adding to each section a set of more challenging exercises, so even participants who are more familiar (but not experts) with R can learn more.
We kicked off our program giving everyone an opportunity to introduce themselves and share learning goals and research interests. We then moved on to the teaching part, using slack to post code, errors, successes, and so that everyone could respond to each others’ questions. The week was intense, especially for people who had never used R before, so to unwind we spent Friday morning downtown and visited a museum where we could watch the history of Rome in 3D. We wrapped up week 1 with a final session on Friday afternoon with a case analysis for participants to apply everything they had learned so far.
Feedback received from most participants was highly positive about the content of the first week, with many stating that this content is where they will benefit most in future. Most participants also indicated that this week’s content was necessary to prepare them for much of the content in week two. This feedback should be viewed as unsurprising, given that we knowingly accepted participants who we felt would benefit most from this content. As mentioned earlier, we also had some participants suggest more advanced exercises.
In the second week, we focused on the core SICSS curriculum and presented the content that was pre-recorded by Matt and Chris. We largely followed the suggested program on Monday through Thursday, and invited Prof. Casey Fiesler as a remote speaker for day 1. For the future, we may include some content suggested by our participants: a brief primer to programming (beyond R) at the beginning of the week, as well as network analysis at the end of the week, which some people were very interested in.
It was particularly heartening for us to see the seriousness with which the participants engaged with the content on ethics (Laura’s field of expertise), sparking significant discussion throughout the day and with our speaker at the end. It was also exciting to see how the participants started applying the R skills and the trace data and text analysis content to their own fields and research.
We dedicated the final Friday to group projects and concluding lightning talks. Feedback from participants indicated that we could have started group projects a bit earlier, since they enjoyed the interaction very much. The difficulty, again, laid in the presence of both undergrad and grad or post-graduate students. The former appreciated the learning experience, while the latter enjoyed research interactions more. Perhaps this was especially thanks to the in-person format, which we were truly grateful we could put in place for most participants.
Overall, we are proud to conclude that the first SICSS-Roma Tor Vergata was a success. It was a fun experience for the instructor, Laura, who is incredibly grateful to our excellent TAs for their huge help all throughout the Institute, and to her co-organizers, Simone and Gustavo, and the university for the support. Participants also indicated that they had a positive experience and that they would recommend our Institute to their friends.
From June 20 - July 1, 2022 we held the Summer Institute in Computational Social Science at Rutgers University in New Brunswick, NJ. The organizers included Michael Kenwick, Katie McCabe, and Andrey Tomashevskiy. Teaching Assistants included Jeffrey Coltman-Cormier and MJ Strawbridge. We hosted the site in-person and shifted to hybrid online/in-person format. This post-mortem will discuss our application process and onboarding, first week, and second week.
We recruited participants by reaching out to past SICSS-Rutgers participants, the political methodology listserv, and colleagues in other Rutgers departments to expand the disciplines represented among applicants. Our application deadline was April 8, and we informed participants we would get back to them by April 15. The application asked for name, email address, institution, career stage, discipline, self-reported comfort with R and statistical analyses, as well as a CV, research statement, and writing sample. We had optional fields for gender and race/ethnicity. We did not require letters of recommendation. We used a google form to gather the information, which worked well. To facilitate communication with applicants, we created a free gmail account for the institute.
We received about 24 applications for the institute. While the applicants were highly qualified, this was a smaller number of applications than previous years and may have reflected the need for participants to be physically in New Brunswick for the institute relative to the past two years where the institute was virtual. We would encourage future in-person institutes to increase recruiting efforts. The team of three organizers evaluated applications based on the applicant’s research in computational social science, contributions to public goods and creating educational opportunities for others, the likelihood the applicant would benefit from the experience and contribute to the educational experience of other participants, and the potential for the applicant to spread computational social science to new intellectual communities and areas of research.
Due to space constraints, we limited the institute size to be within 20 participants and ultimately had 18 participants accept and attend our two-week event. Approximately two-thirds of participants were women. The plurality of participants were from political science, with a broad range of other disciplines represented among participants, including communication, psychology, social work, urban planning, demography, public administration, and ecology. We also had multiple career stages represented: While the majority of participants were doctoral students, participants ranged from incoming first-year doctoral students and included multiple faculty participants. About half of participants were from the Rutgers University system, and half came from other institutions throughout the Northeast and Mid-Atlantic.
After notifying participants of their acceptance, we began onboarding participants to the SICSS Slack platform and GitHub website. We used google forms to collect additional information about participants’ backgrounds and logistical information about their need for parking, travel information, and dietary restrictions/other accommodations during the event.
In addition, we provided them with links to a suggested pre-arrival preparation schedule, drawing on the SICSS bootcamp and main curriculum materials to gain familiarity with R and the computational social science concepts discussed during the first week of the institute (schedule here). During May and June leading up the institutes, our TAs provided a small number of virtual office hours to support the participants in their preparation.
The first week of the institute focused on building greater understanding of several computational social science topics, extending upon the SICSS main curriculum with small group exercises and supplemental live tutorials. The schedule for the first week is here with links to the materials.
On Day 1, during the morning, we welcomed participants to the institute, had opportunities for participants to introduce themselves and begin to find common ground with their co-participants, and brought two SICSS-Rutgers alumni (Mike FitzGerald and Katie Krumbholz) to the institute to present on research they developed during SICSS, which has now been published. After breaking for lunch, the afternoon focused on ethics, and we followed the case study small group exercise provided by SICSS.
Day 2 focused on the collection of Digital Trace Data. We had live tutorials from the organizers and teaching assistants on web scraping and using the Twitter API. Participants then had extended small group time to develop a research question and collect digital trace data on their own. We concluded the second day of the institute with a community building activity at a happy hour outdoors at a restaurant in New Brunswick.
The third and fourth days of the institute proceeded similarly to Day 2 in format. We had live lectures on using text as data on the third day of instruction and live tutorials on supervised machine learning tools on the fourth day. On each day, participants broke into small groups to practice these skills and then concluded with a short presentation on their group’s work. We were also fortunate to have a guest speaker, Sarah Shugars, who gave a talk on the value of machine learning for computational social science.
The final day of instruction was focused on surveys and experiments. After a short overview of the materials, participants gained practice designing survey questions and programming and setting up a Qualtrics survey to be launched on Prolific. While the surveys were being deployed on Prolific, we had a second small group session where participants practiced developing an experimental design based on a research question and causal claim of their choosing.
The second week was primarily devoted to the launching of group projects. Similar to the previous two years of SICSS-Rutgers, prior to the start of the second week, we asked participants to fill out a spreadsheet indicating their research interests. Participants completed two rounds of small-group sessions where they were tasked with coming up with research project ideas. For the first of these sessions, small groups were formed using a clustering algorithm on the research interests mentioned in the spreadsheet. The second session included a more random formation of groups. After the two sessions, we then gave participants approximately 20 minutes to choose the project idea they were most excited about joining, with the rules being that each group should have approximately between 2-5 participants. We ended up with 4 small group projects. We told participants that their group projects could include the launching of a traditional academic research project, the development of public goods (e.g., a database or Shiny application), or in-depth small group study of a topic for which they want to gain additional experience.
From Tuesday-Thursday, participants spent the bulk of their time within their small groups working on their projects. The organizers and TAs monitored group progress by requesting short written project updates, by floating into different groups throughout the day, and encouraging groups to use Slack to provide updates and ask troubleshooting questions. In addition to small group work time, the second week of the institute included a variety of guest speakers: P.M. Aronow, Jessica Hamilton and Melissa Dreier, Robert Kubinec, and Tamar Mitts, as well as a supplemental geospatial tutorial provided by our teaching assistant, Jeffrey Coltman-Cormier.
The last day of the institute was devoted to group presentations and closing remarks. Each of the groups gave a 30-minute presentation in which each member had a role. Other participants attended and gave feedback via Zoom chat and Slack. Each project demonstrated a variety of tools learned throughout the institute, including the use of text analysis skills to detect media framing in coverage of abortion, sentiment analysis of political tweets from different media outlets, the development of an online repository of Canadian electoral officials on Twitter as a public good, and the integration of administrative data and social media data to study government responsiveness through multiple perspectives.
We concluded the institute by encouraging participants to “pay it forward” by continuing to share their knowledge and skills with others at their own institutions or at future SICSS events.
The 2022 event was the first time that SICSS-Rutgers was held in person on the New Brunswick campus. With this in mind, below we offer some special considerations for future in-person events:
SICSS-Singapore 2022 was the first SICSS partner location to be held not just in Singapore, but in the South East Asian region at large. Hosted at the National University of Singapore, and funded by the Centre of Trusted Internet and Community (CTIC) and the Faculty of Arts and Social Sciences (FASS), the core-organizing committee comprised three standing faculty members at the Department of Communications and New Media at NUS: Professor Jack Qiu, Assistant Professor Renwen Zhang, and Assistant Professor Subhayan Mukerjee.
Once Singapore was confirmed as a partner location in late 2021, we set up the website in December, and began outreach and publicity in January 2022 (https://twitter.com/wrahool/status/1478556498158514177).
Applicants could be master’s (by research) students or higher (including PhD students, postdocs, and early career faculty members within seven years of getting their PhDs). They were required to send their CV, a 500-word statement, one writing sample of not more than 35 pages, and (for students and postdocs) the name of one faculty advisor who is familiar with their work.
Since we were organizing SICSS for the first time, we set a modest target of receiving around 40 applications, with an aim to accept around 20. Moreover, while anyone around the world could apply, we made a conscious effort to prioritize local candidates (not just from Singapore, but the larger ASEAN region). The application deadline was set for February 2021 with an expectation to have SICSS-Singapore as a hybrid event. We later extended the deadline to March 1 to hit the required number of applicants. In the end, we received 41 applications from 8 countries. Publicity was done via the social media handles of the department, funders, and co-organizers. For future iterations, we plan to increase our target pool of applicants in order to make SICSS-Singapore more selective. One oversight on our part (that we plan to correct in the future) had been to ignore Chinese social media platforms, as those are widely used by many students, their friends and relatives in Singapore.
In parallel, we also worked on inviting guest speakers, with a special focus on showcasing practitioners of computational social science from in and around Singapore. We finalized six speakers (3 local, 3 international; 3 male, 3 female; 4 from academia, 2 from industry).
Since COVID-19 rules had been relaxed in Singapore, we pivoted to a fully in-person event (barring the international speakers who would be delivering their lectures via Zoom). We shortlisted and invited 22 of the 41 applicants - who we confirmed would be able to be physically present in Singapore.
One oversight on our part was to not consider the visa requirements of international applicants: since all the participants had said that they would be in Singapore, we had assumed they would be able to enter Singapore without any difficulty. This was largely true, except for one participant (from Malaysia) who held a Yemeni passport, who emailed us the day before SICSS began that he needed a visa to enter Singapore. Even though we sent him a visa letter the next day, he was unable to make it to the event in person.
We opted to not onboard the participants to Slack as we got to know from other organizers (through Slack) that Slack onboarding hasn’t always been effective. Instead, we decided to create a Google Group for all discussions.
We had arranged for catering for every day of the event (snacks in the morning, lunch, and snacks in the evening). We surveyed the participants regarding their dietary requirements before placing the catering order.
We also hired three NUS undergraduate students to assist us in our day-to-day logistics for the event.
We had one conference room booked for all days for the guest lecture and video sessions, while there was ample space outside the conference room for both catering and group activities on all days. The conference room had large TV screens, and we procured the necessary AV equipment for the in-person guest speakers.
Monday began with an opening (virtual) guest lecture by Prof. Sandra Gonzalez-Bailon from the University of Pennsylvania. While the session itself went smoothly, the timing could have been better. Owing to the time difference between Philadelphia and Singapore, the talk was scheduled for 8:30 AM, which meant the participants had to arrive by 8 AM for logistical debriefing. For future iterations, we will try to have a local speaker deliver the first day’s lecture - preferably later in the day.
We had four other speakers over the following four days:
We tried to reach a balance between using the main SICSS curriculum (because we were organizing SICSS for the first time, and didn’t want to experiment too much) while adding some local, Asian/South East Asian flavor to the event. Our speaker line-up for week 1 reflected this balance. Thus, while most of our speakers have global training in CSS, some of them apply CSS to local/regional problems. We plan to increase the number of locally relevant speakers in future iterations of SICSS-Singapore.
We largely stuck to the main curriculum except on Friday, when we deviated by incorporating network analysis (to bring it more in line with the speaker’s topic and the expertise of the organizing committee. For future editions, we will play fewer of the pre-recorded videos and have the organizing team deliver similar content as “live” lectures to make the sessions more interactive. During the afternoon activities, the participants worked on the material provided by the main curriculum while the organizing team helped debug coding issues that the participants faced. Now that we have experience organizing a partner location, we plan to have locally-relevant problems for the students to work on in future editions as well and thereby, reduce dependency on the main curriculum.
One recurring problem that we encountered was the microphones running out of batteries during the in-person guest lectures. While they were all resolved eventually, this is something we should plan for in the future.
The second week was devoted to group projects, with the first day starting with a short and fun quiz (Kahoot!) assessing participants’ understanding of the things they learned in the first week. We included two questions for each week’s topic covered in the first week, totaling 10 questions. The participants generally did well, with a few students getting full marks. Thus, we believed that the lectures and exercises in the first week had provided a good basis for the group projects. After the quiz, we explained the rubric and requirements of the group projects and then asked participants to work on their own group projects.
We also arranged a special session about open science on Wednesday, as open science is an important area that CSS focuses upon and embraces. The panelists included our organizers, Subhayan and Renwen and a participant (Aimee Pink from A*Star) - all of whom had substantial prior experience in practising and advocating for open science. They discussed some of the best practices of open science and how they can help make CSS research more transparent, replicable, and meaningful. Participants actively participated in these discussions.
Given that we gave participants the flexibility of working on their group projects outside of the event venue, the attendance was generally lower than the first week, with fewer than 10 participants showing up at any time. This reduced the opportunities for participants to chat and learn from each other. Moving forward, we would probably distribute guest speakers over Week 2 and arrange some hand-on sessions to encourage participants to come to the event venue.
The highlight of the last day was the concluding talk by Dr. Kokil Jadka, followed by the group presentations from all participants. We had 5 groups in total and they covered a wide range of topics, including public opinion on food security in Singapore, identity politics, suicide prevention, network effects in quitting academia, and online customer engagement with AI chat bots. All these group projects were well-suited to the CSS methods they had learned over the past week, and the quality of presentations and study designs went far beyond our expectations. We asked them to submit formal project proposals for potential funding a week later. Four of the groups submitted well-thought-out, insightful proposals. Two of the groups were awarded a seed fund of $2,000 to execute their projects.
We ended SICSS with a social event, where we had beers and dinner and chatted with each about our research, life, hobbies, and potential collaborations.
Results from an anonymous post-event survey that was administered to the participants revealed very positive feedback. Most of them complimented the arrangement and structure of SICSS and said how it helped them get a solid foundation in CSS and put this knowledge into practice.
The summary of the survey results is as follows:
Crucially, there were two respondents who gave a 1/5 rating to their overall level of satisfaction. To the question “What are some ways you think in which SICSS-Singapore can improve?” one of them said “it was hard to have to watch multiple youtube videos every day even though we were meeting in person.” The other revealed extreme difficulties in working on the group project that they did not appreciate.
This allows us to segue into the difficulties we faced while coordinating the group projects. One group ended up not submitting a proposal due to disagreement among group members, and another group had some conflicts due to some members being absent from most discussions. Reflecting on this, we think these issues mainly stemmed from participants not being able to work together in person because some of the members had other obligations and prior commitments. As a result, they were not as engaged as we would have liked. In future iterations, we will arrange guest lectures and in-person hands-on sessions in the second week as well, to encourage participants to come to the venue and work together on their group projects. To that end, we will clarify in-person participation expectations during the application process and not accept participants who cannot commit to being physically present at the venue for the full two-week period.
We’ve divided the post-mortem into 4 main sections: 1) outreach and application process; 2) pre-arrival and onboarding; 3) first week; 4) second week (group projects).
Information Sessions and Email List
For our 2022 Program, we wanted to recruit a group of participants who are a mix of gender and geographic diversity. To achieve that goal, we hosted 4 different information sessions from mid-January through mid-April in 2022 to directly explain to potential applicants our mission, our application process and our programs. Each information session usually lasts around 30 minutes.
In our experience, it’s an effective recruiting method because we were able to interact with our potential applicants throughout the application process, and eventually maintain an email list (around 300 people) from which the recipients can receive information about our applications. According to our data, 55% of our participants attended at least one info session.
Single Rounds of Applications
As a result of our recruiting effort, we have received over 70 applications this year. We experimented with two rounds of application deadlines, hoping to better estimate the applications we would receive this year and evaluate if we need to double down our recruiting effort if the first round of applicants didn’t attract the expected numbers. Upon reflection, given that we already had a 300-people email list around the first-round deadline, we probably could have had just one round of deadline as the evaluation process took significantly longer than we anticipated.
25% Attrition Rate
We originally admitted 29 applicants and 5 wailisters. Two admitted applicants withdrew due to concerns about covid surge (as we had planned in-person sessions this year), and we were able to fill the spots with our top two applicants. Five admitted applicants eventually didn’t complete the program (one didn’t show up in the program at all, while the other three withdrew in the middle of the program because they felt they were falling behind the program). We recommend that the future organizers account for 25% of the attrition rate and bake such attrition rate into the final number of admitted participants.
After the admitted participants have confirmed our offers, we begin to onboard them into the program and provide them with pre-arrival materials. We implemented a four-week bootcamp this year. The goal is to provide necessary foundations so that the participants can develop fundamental skills in data science and develop common language in computational social science.
We formalized the bootcamp learning experience by providing four weeks of suggested learning content prior to the two week synchronous event. We budget each week’s learning time to be between 5 and 10 hours. Video content accounts for 3 to 3.5 hours per week.
We also asked participants to complete two pre-arrival exercises. The completion rate was 85% for assignment 1, 75% for week assignment 2. Completion of the bootcamp assignment turned out to be a key predictor for participants to stay through the two weeks and participate in research projects. The detailed curriculum can be found here.
We used a mix of email and Discord server for onboarding communication.
We used the standardized onboarding form for participants to populate their name and profiles onto SICSS directory. However, one participant had trouble with her photo due to HEIC format. In the future year, it will be important to remind participants to use jpeg when uploading the picture.
We added people to the workspace and told them they should have received an email inviting them to join. In case they couldn’t find that email, we also included a link via which they could join the workspace. However, that was not the primary ecosystem we use.
Aside from the SICSS-2022, we set up a Discord server for participants. participants used the Discord server to ask questions regarding the bootcamp exercise. During the synchronous, virtual component of our program, Discord was also a useful bulletin board for us to share information regarding the learning content.
Another component of the pre-arrival support is office hours run by our TAs. We provided 4 weeks of office hours from 3 TAs. Around 25% of the participants attended the sessions. For the ones who attended, TAs helped with troubleshooting technical issues or addressing any other questions they had.
1) We find the formalized bootcamp to be a successful component. We recommend using the similar workload, 5-10 hours of learning content per week including video, reading and assignment. 2) We recommend future organizers to implement a virtual meet-and-greet prior to the bootcamp to make participants more familiar with one another, get comfortable with asking TAs questions, and build up relationships organically, which would help team research projects later on.
We implemented a hybrid format to encourage participants from different geographical locations in Taiwan to apply for this workshop and allocated enough budget to subsidize housing during our three-day in person sessions . Week 1 of the SICSS-Taiwan was fully virtual. Monday, Tuesday and Wednesday of Week 2 were also virtual. Thursday, Friday and Saturday of Week 2 were held in-person.
We continued to use the “flipped classroom” model according to which we asked participants to watch a subset of the pre-recorded videos during our four-week bootcamp. During week 1, we had scheduled activities from 9AM to 4PM during most days.
Similar to our experience in 2021, collaborative coding over Zoom was challenging. There were three sessions that involved Text as Data, Data Scraping with API, and Chinese Text Mining. We split the groups in different ways. One of the methods we used is to group participants of similar level together, which in the other courses the participants were grouped randomly. Three TAs and a few organizers acted as workshop facilitators to rotate through the virtual rooms to help folks move along the discussions. Another challenge to shared participation was the choice of programming language. Most participants preferred R and some Python. In one small group session, we had a group where the preferred language was Python. In other sessions, we did not explicitly split by preferred language.
To make the curriculum more relevant to research related to Taiwan, we added one afternoon session on Chinese text mining and removed the session related to sample weighting.
We attempted to expand beyond text as data. Another new addition was on data visualization. Besides the workshop facilitated by our organizer, we had two guest lecturers who are data journalists to give lectures on data charting principles and cartography.
This year, we took a very active role in planning the lecturers / talks with our guest lecturers. On average, wel had two meetings with each lecturer before the program started. In the first meeting, we explained to the lecturers our participants’ composition and brainstormed what directions of the talk would be relevant to and can inspire our participants. In the second meeting, we went through an outline of their talk and gave our feedback on how to structure and present their talks. The outcome of this lengthy, curated process was positive. Our participants were impressed with the quality of our guest lectures. Our lecturers also felt this process helped them connect and understand their audience better.
There were 5 participants who decided not to fully participate in the workshop during week 1. A main common feedback was the time commitment. They did not expect the workshop to be as intensive.
1) Our week 1 schedule was quite intensive and participants expressed Zoom fatigue. We recommend more breaks in next year’s program if we continue to have a hybrid format. 2) We received feedback on moving beyond text data in the future. We may explore adding in network analysis or analysis of video or sound data.
A major part of SICSS is participant-led group research projects during the second week. In 2021, due to the pandemic, the SICSS-Taiwan was a fully-online event. Presenting a group project was an optional activity. This year, the group projects were mandatory for every participant, and we arranged a 3-day in-person event to facilitate more constructive group projects. Group members formed by the beginning of Friday.
We have three approaches for matching participants for the research groups. First, we encouraged group registration during the application stage. We received a number of group applications and accepted two groups of two applicants to participate in our program. While we were initially worried that pre-formed groups might exclude others from collaborating, these two groups ended up incorporating additional members. Second, on day 2 of the first week of the program, though in a virtual format, we used a google spreadsheet for people to add research interests and opened virtual chat rooms to let participants know the interests of the others and assist them to find suitable partners for group projects. Third, in our bootcamp assignment, we asked participants to prepare a 7 minute presentation on a topic related to computational social science not addressed in our curriculum. During the synchronous event, we asked 5 participants to give their short talks. This method was helpful for generating conversations since we didn’t have opportunities to have a side conversation in the virtual format.
Benefiting from the last three days of the in-person sessions, we found participants could easily adjust their thoughts as well as their demonstrations without the barrier of screen during a virtual meeting. 75% of the participants ended up working on a different research topic from their proposals.
One improvement over last year is that we had a session where organizing committee members shared their experience in organizing SICSS-Taiwan 2022. The goal was to make the process more transparent and encourage the 2022 participants to become co-organizers for the 2023 program.
On Friday there were five groups presenting projects, and these groups involved 21 participants. The presentations and discussions were lively, and we think this is a great way to wrap-up the week.
List of the group projects is as follow: Ivermectin Discourse in Social Media, Analysis of Web Forum on College Entrance Exams, Panel Data on Chinese Foreign Policy and Chinese Diplomats’ Tweets, Analysis on Graduate Student Theses and Dissertations in Taiwan, Analyzing the Press Release of Social Movement in Taiwan.
SICSS-Tokyo was a one-week event from August 1-5, 2022. The organizers were Hirokazu Shirado and Makiko Nakamuro. The teaching assistant was Atsushi Ueshima. We hosted the site virtually via Zoom. This was the second time for SICSS-Tokyo. The post-mortem discusses our application process, onboarding, event week, and self-review.
We prepared the SICSS-Tokyo website in December 2022. In addition to the English version, we included the Japanese version on the website because we were concerned that some Japanese candidates might hesitate to apply due to the language barrier.
We began our outreach in February 2022. To recruit a wide range of participants, we sent the advertisement to several professional email lists as well as posted it on social media via organizers’ networks. We also gave a remote talk to introduce SICSS-Tokyo at Japan’s annual computational social science symposium.
Our application deadline was March 11, 2022, and we informed participants we would get back to them by early April. The application asked for name, email address, institution, career stage, discipline, preferred language (English or Japanese), a CV, research statement, and writing sample. We had optional fields for gender. We did not require letters of recommendation. We created a free Gmail account for the institution and used a google form to gather the information.
We did not recruit enough participants from humanity-major Ph.D. students in the first recruitment cycle. We thought that the application due might be too early because we would start the event in August. Also, some potential applicants might hesitate to apply for the program because of the language barrier. Thus, we deiced to have the second recruitment in June with some modifications to our application site.
In total, we received 14 applications and accepted 11. In keeping with the last year, about 25% of applications came outside of Japan, such as the USA and Turkey, even with the time difference. We completed the selection process in June.
We let the applicants know the result on March 22 and June 24, respectively, and confirm the attendance of accepted applicants via another google form. All 11 applicants accepted our offer. We created a google group for SICSS-Tokyo and used the mailing list and Slack to inform participants of the pre-requirement and updated schedule. We also requested participants to provide a picture and a short bio for the SICSS website using the mailing list.
The main organizer (Hirokazu Shirado) and the TA (Atsushi Ueshima) had weekly one-hour meetings from June 24 through August 10. The meetings helped us prepare the event contents, check the progress on time, and share improvements.
SICSS-Tokyo was a one-week event in 2022, from August 1 to 5. We decided on this period considering the typical academic schedule in Japan, where a semester ends in late July. However, the schedule might not affect the participants’ decisions. Therefore, we will reconsider the event timing in the future.
We set up three goals for participants for SICSS-Tokyo:
We provided them with three components: live lectures, research-idea discussions, and participants’ talks. We decided not to run analytical and programming exercises. Instead, we laid weight on developing research concepts in SCISS-Tokyo. We also decided to give live lectures rather than recorded ones.
Our schedule was as follows:
The name in brackets presents who led the component. When we completed the SICSS-Tokyo, we asked them to give us the post-survey.
Overall, participants actively engaged in all the lectures and discussions. They gave many questions and comments on each content. More importantly, these questions and comments allowed the participants to go forward with their own work or/and find a new research direction. Thus, we think that, like the last year, the event structure worked well for our goals of SICSS-Tokyo. In the future, we might want to add a couple of guest lectures.
We think this year’s group size worked very well. As all participants understood research interests with each other, we discussed the lecture contents in the context of their research. We added the contents about the word embedding NLP technique in social-science contexts and the discussion about the IRB processes of participants’ institutions. Many participants showed their interest in these topics. We also tried to use our Slack channel more intentionally than the last year. It also yielded results so that participants communicated and exchanged research information even after the event. We also increased the presentation time for participants in the schedule.
We found several details that we might want to change:
SICSS-Rochester took place from May 9-20, 2022, as a two-week event. The organizing committee members were Cantay Caliskan (SICSS 2019), Ezgi Siir Kibris (SICSS 2020) and Bahar Zafer. We hosted the site in person at the University of Rochester campus. This was the first ever SICSS in Rochester. The post-mortem discusses our application process, onboarding, event week, and self-review.
We started the application process on January 15, 2022. To recruit a wide range of participants, we sent an advertisement to several professional email lists as well as posted the event on social media via organizers’ networks. We widely advertised the event at the University of Rochester. We deliberately chose the event to start on UofR campus immediately after the finals week hoping that this choice would be the most suitable one for local applicants who will eventually leave Rochester for traveling home or starting for an internship in the summer.
Our application deadline was March 18, 2022, and we informed participants that we would get back to them by early April. The application asked for name, email address, institution, career stage, discipline, as well as a resume, statement of purpose, and a writing sample. We also asked for the names and email addresses of two referees.
We received a range of applications coming from several different countries including the USA, Turkey, Brazil, China and Nigeria. We had applications coming from undergraduate, graduate, post-doctoral students, as well as one assistant professor. We completed the selection process by April 10, 2022.
We let the applicants know the result of their applications on April 10, 2022, and confirmed the attendance of accepted applicants via another Google form. 20 applicants (all people who were admitted) accepted our offer. One of our participants had problems with flying to the USA because of covid-19 related restrictions and was not able to attend the event. Another applicant eventually withdrew because of financial concerns. As a result, SICSS-Rochester had 18 participants (5 advanced undergraduates, 5 Master’s students, 6 PhD students, 1 post-doctoral researcher, and 1 professor). We used the mailing list as well as Slack to inform participants of the pre-requirements and the updated schedule. We also requested participants to provide a picture and a short bio for the SICSS website using the mailing list.
We divided the event schedule into three components: research talks, lectures and tutorials, and project-related presentations. Our goal was to introduce our participants to the field of computational social science, as well as helping them to get introduced to cutting-edge research in the field. Having these objectives in mind, we aimed to recruit researchers working at the cross-section of data science and political science, such as the use of noise to analyze corporate data, the use of audio and image data for analyzing political events, the detection of bots, map-making, and the use of webscraping for collecting original social science datasets among others. In terms of lectures and tutorials, we covered NLP, machine learning, data visualization, network analysis, and various API and webscraping methods.
The first week of the event was more strongly dedicated to introducing participants to the field, whereas the second week more heavily focused on completing the projects. For a more detailed overview of the program, please check the links below:
Our detailed schedule and the names of our speakers can be found here. The abstracts of the talks (there were nine of them) presented can be found here.
Overall, most of our participants actively engaged in all the lectures and discussions. There were a few research talks in which the audience posed many detailed questions and comments. We also felt that these questions and comments allowed the participants to go forward with their own work or/and find a new research direction. An important characteristic of the event was that the educational backgrounds of the participants were more diverse compared to other previous SICSS events in various locations; initially, we expected that this would be a challenge, but, in the end, participants were happy about getting to know and working together with people from different backgrounds.
Participants gave flash talks on the first three days of the school, and we asked them to form research groups by Wednesday of the first week. The quick and dense schedule at the very beginning gave participants more time to work on their projects and released some part of the uncertainty since they got information about other participants from the very beginning of the summer institute.
We had the best project award (chosen by the organizers [Cantay, Ezgi, Bahar]) and the popular award (for which the participants voted). Some competition created more incentive to make the projects better. Thanks to the funding we received from SSRN and the Goergen Institute for Data Science at the University of Rochester, we were able to motivate students with high quality best paper awards. On a related note, abundance of food and snacks was also well-received by the participants and motivated them to be engaged. In the end, all the projects were great and our participants fulfilled all the expectations presented (submission of progress presentation, final presentation, code, and final paper). All groups produced actual first paper drafts; and, some of the projects offer promise for publication in the future.
Lectures and Tutorials
We felt that the lectures and tutorials were well-balanced, and there were at least three topics that every participant was interested in learning more about.
Food and Room
We had an abundance of snack food, and our lunch meals were quite satisfactory. We also had a quite large and modern classroom with powerful AC in which our participants had a chance to interact with each other comfortably.
Possibly, in our (organizers’) opinion, the comparative advantage of our program was the high quality and research excellence of our speakers. Especially, the keynote speech delivered by Prof. Gary King of Harvard University received great praise from participants; and, throughout the event the quality of the talks motivated our participants to come up with thought-provoking questions.
We sent out a survey to our participants two days after the end of the program. Our response rate was 66%. The survey will be helpful in organizing the event again in the future.
Announcement of the program
As previously mentioned, we aimed for a higher number of participants coming from UofR because of the timing of the event. We were happy with the pool of our applicants; but, we believe that we could have received a greater number of applications if we had electronically announced through various channels and contacts in different continents. The number of applications was lower and less diverse than expected. We think that one possible (big) cause for this could have been Covid-19 and related concerns.
Speakers from Industry
One area that we should have done better was inviting one or more speakers from the industry. The people we contacted from various big tech companies were not available for talks; so, we had to create a purely academic event at the end.
Delivery of Food and Coffee Machine
We had 11 lunch meals (in total) throughout the event. On three of those days, lunch arrived considerably late due to some confusion of the delivery people, almost disrupting our schedule. We learned that for the next event we need to order the meals much earlier than needed (for instance one hour earlier than the actual lunch time) to streamline the process. On a related note, we had a few problems with our coffee machine on the first three days - but the coffee machine fixed itself (to our surprise and satisfaction) without any further need for repair.
If we have the chance to organize SICSS again, we are planning to cover image processing and deep learning, and possibly the use of APIs for new social media outlets such as TikTok. We did our best to cover some cutting-edge methods and tools, but we also felt that there are also other interesting topics that we did not have a chance to discuss.
We could have helped our participants to get to know each other better by organizing a ‘bonding experience’, such as a daily trip to a museum, a national park, or similar. We couldn’t do this because of our compact schedule and the summer plans of our participants.
Scholarship for Accommodation
We tried to use our budget for the program as efficiently as possible; nevertheless, we could have increased the number of participants by up to 15% if we had a chance to cover expenses for accommodation for participants coming from other cities and countries. We will do our best to find a solution to this issue in the future.