A collection of post-mortems
Table of Contents
We’ve just completed the 2026 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 various partner locations run by SICSS alumni, we continue to grow our network of computational social science researchers.
These post-mortems describe, for each site, a) how the Institute was run, b) what each site thinks 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 2027 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.
Skip to:
Adam Mickiewicz University, Poznan — September 7-18, 2026
Universidad del Rosario, Bogota, Colombia — June 4-12, 2026
FGV Communication, Brazil — July 27 - August 6, 2026
National University of San Martin, Buenos Aires — August 10-21, 2026
Chicago State University — July 13-23, 2026
Boca Raton, FL — June 1-5, 2026
Habib University, Karachi, Pakistan — June 8-19, 2026
SICSS-Habib 2026 was the first edition of the Summer Institute in Computational Social Science held in Pakistan. It ran at Habib University in Karachi from June 8 to June 19, 2026, with 17 participants.
The program centered on the societal impacts of technology in the Global South, and that focus ran through the guest talks and every group project. This sat within a broader ambition for the institute spanning causal inference, digital polarization, and civic discourse. For a first edition in a country with no existing Computational Social Science community, we judged breadth more useful than a single tightly defined theme.
Funding
SICSS-Habib 2026 was funded by Tech for Open Minds (TOM), the SICSS initiative supporting research and training on how digital technologies shape open-mindedness, humility, and polarization. The grant covered the institute in full, and no participant paid a fee.
On the participant side, it paid for travel for the seven funded participants, all meals and refreshments for the two-week period, and some merchandise for the institute consisting of t-shirts, notebooks, and pens. On the program side, it covered travel and accommodation for invited speakers, speaker honoraria, and wages for the student teaching assistants who supported the two weeks.
Organizing team
SICSS-Habib 2026 was organized by Dr. Muhammad Qasim Pasta, Associate Professor of Computer Science at Habib University, and Asad Tariq, a Dean’s Fellow in the same department. Asad handled correspondence with guest lecturers and ran day-to-day operations across the two weeks.
Two student teaching assistants, funded through the grant, supported the program. Their role was primarily logistical, including keeping sessions running and absorbing the practical problems that surface hourly across a two-week institute.
However, the support participants needed the most was technical, and for that, in future editions, we wish to employ a TA dedicated solely to the programming side: working alongside participants during the workshops, debugging code, and resolving the technical problems.
1. Outreach and application process
We advertised through LinkedIn, which produced 59 applications from across Pakistan against roughly 25 seats. Dr. Qasim Pasta also circulated the call directly through his own network, which included professors at other universities, and contacts across departments within Habib. For the next edition we would add a targeted mailing list to reach an even larger number of people.
Of the total 25 seats, half of the seats were reserved for women, a deliberate response to regional gender disparities in computational research that we built into the allocation from the start.
Applications comprised of a CV, a statement of interest, a writing sample, and questions about prior exposure to computational social science. We evaluated each on:
Both organizers reviewed every application jointly. Beyond the formal criteria, what we weighed most heavily was ambition and motivation. We looked for applicants who wanted to apply computational methods to socially relevant problems, and who showed an interest in building a computational social science community in Pakistan. For a first edition in a country with no existing SICSS presence, we cared more about finding people likely to carry the field outward afterwards.
We extended 24 offers and 17 attended, an attrition of roughly 29 percent. Some of the seven stopped responding altogether, and others replied but did not appear. We had no means of recovering those places. For the next edition we will maintain a ranked waitlist so that late declines can be backfilled and also follow up actively with anyone who goes quiet after accepting.
Seven of the 17 received travel funding to reach Karachi. Participants came from Karachi, Hyderabad, Sukkur, Islamabad, Multan, Lahore, Balakot, and Gilgit — a reach from Sindh through Punjab and the federal capital up to Khyber Pakhtunkhwa and Gilgit-Baltistan.
Who attended
The cohort was evenly divided by gender: 9 women and 8 men among the 17, bringing the balance to 9 women and 8 men so the reserved-seat commitment held through to the room itself, not merely to the offers we made.
Institutional spread was wide, and the 16 participants came from 14 distinct institutions. Alongside Habib, participants came from IBA, UIT University, and Bahria University in Karachi; Forman Christian College in Lahore; The Women University Multan; Muhammad Ali Jinnah University; National Defence University and the International Islamic University in Islamabad; Quaid-e-Awam University of Engineering, Science and Technology in Nawabshah; and Government Degree College in Balakot. Two were Pakistani nationals who had studied abroad, at the University of Manchester and at Universitas Diponegoro in Indonesia. One came from outside the university sector altogether, from The Citizens Foundation, an education non-profit. By province of origin, we had 10 participants from Sindh, 2 from Punjab, 2 from Gilgit-Baltistan, 1 from Khyber Pakhtunkhwa, and 1 from Islamabad.
The disciplinary backgrounds of these participants were mixed and frequently interdisciplinary, as depicted by the fields named by the participants in the application form. Data science was the most common (5), followed by computer science, linguistics, sociology, and economics (3 each), communication (2), and single participants from computer engineering. Counted by mentions, social science and humanities fields outnumbered computing fields roughly two to one.
The career stages of the participants ran from recent graduates through faculty, with faculty being the largest single group with 6 of the 16 participants indicating that they held a teaching position, in some cases alongside ongoing graduate study. Two were doctoral candidates or recent PhDs, three were master’s students, and the remainder were recent graduates. None of the 16 had attended a SICSS before.
Prior programming experience
Participants who applied rated their own Python and R ability on a 0–10 scale. The distribution explains a good deal of what followed. The mean self-rating was 4.1 for both languages, but the spread was extreme: 8 of 16 rated their Python at 3 or below, 7 of 16 rated R at 3 or below, and 6 rated themselves at 3 or below in both. At the other end, 3 participants rated themselves 7 or above in both Python and R. Several reported an outright zero as well.
We therefore ran a two-week program of hands-on computational workshops for a cohort in which more than a third were close to beginners in both languages those workshops relied on.
2. Pre-arrival and onboarding
We published the schedule and a pre-arrival page ahead of the institute. The materials asked participants to read Matt Salganik’s Bit by Bit: Social Research in the Digital Age, to work through the online SICSS Boot Camp if they needed beginner-level coding skills, and to install R and RStudio and create a GitHub account.
We ran two communication channels: the official SICSS Discord as the formal one, and an informal WhatsApp group for faster logistical announcements. Participants used WhatsApp considerably more, and future editions could plan on that rather than treat it as an overflow channel.
The gap we would close next time is expectation-setting about the sessions themselves. Our materials prepared participants for R, but the program also required NetLogo and Google Colab, neither of which appeared in the preparation. In the future iterations of the institute, we will circulate fuller guidance on what sessions to expect and which tools each will use.
Accommodation was the clearest gap, and by a wide margin the most common subject of pre-arrival questions. Habib University has no on-campus housing, and participants traveling from outside Karachi had to arrange their own lodging. Several said it was the most stressful part of attending. We intend to pursue this issue on two tracks: requesting an allocation for participant accommodation in our next funding application, and, regardless of whether that succeeds, treating accommodation as ours to coordinate, by circulating a vetted list of nearby guesthouses and hostels with indicative prices and travel times to campus, negotiating a block rate with one or two properties, and connecting out-of-town participants with each other early enough that those who want to share a booking can arrange it.
3. The first week
Each day of the first week followed the same shape, including a hands-on workshop, a guest talk, and an afternoon group exercise, built around a daily theme: introduction and research ethics, text as data, agent-based modeling, social network analysis and digital trace data, and AI and large language models.
Workshops were led by Dr. Qasim Pasta (Habib University) on research ethics and, separately, on network analysis in R; Dr. Muhammad Yaseen Khan (FAST National University of Computer and Emerging Sciences, Karachi) on quantitative text analysis; Zain Ahmed Usmani (Habib University) on agent-based modeling using Python’s Mesa library; Asad Tariq (Habib University) on social simulations using NetLogo; and Dr. Tafseer Ahmed Khan (Mohammad Ali Jinnah University, Karachi) on using large language models in Google Colab. Guest talks came from Dr. Umberto Mignozzetti (University of California, San Diego), whose opening keynote made the case for the field; Dr. Muhammad Bilal (FAST NUCES, Islamabad) on social media data and Society 5.0; Dr. Shah Jamal Alam (Habib University) on simulating social complexity; Dr. Ihsan Ayyub Qazi (Lahore University of Management Sciences) on doctor-AI collaboration in diagnostic reasoning; and Dr. Abdul Samad (Habib University) on AI and LLMs for social science.
The week closed with a Friday afternoon “Bouncing Ideas” session, where most of the research speed-dating took place; it carried over into a dedicated group formation session on the Monday.
The skill spread described above made itself across the hands-on sessions. In practice, the less experienced leaned on the more experienced, asking for help whenever they got stuck, and the organizers were on hand to guide anyone who needed it. This worked, though informally rather than by design, and it depended on the more confident programmers being willing to guide others.
Participants told us that the first week was too dense. Each day had roughly six and a half hours of structured content, with little room to absorb one session before the next. Each day started and ended with a social gathering over tea and refreshments, allowing participants to interact and build networks.
4. The weekend
We held a communal breakfast on the Saturday between the two weeks. It was a small thing, but it gave participants unhurried time with each other and with us outside a classroom setting, and the tone of the second week was noticeably easier for it. We would add more of this next time including a city tour, and further social gatherings including dinners.
5. The second week: group projects
The second week shifted from structured instruction to independent group work.
Dr. Rabeea Jaffari (Mehran University of Engineering and Technology) opened the week with a talk, “To Be Or Not To Be Human,” on the dilemmas raised by anthropomorphic large language models.
Participants grouped together into teams of two to four for their projects that they worked on for the rest of the second week. We deliberately asked them to generate their own research questions rather than offering pre-packaged project ideas or datasets, and every group ended up pursuing a question and a data collection strategy of its own. This took longer to get moving than a pre-supplied brief would have, but the resulting engagement was noticeably higher.
Five teams presented on the final day:
Who Speaks About Education? Mapping the Voices: A computational content analysis of four years of education coverage in Dawn, The Express Tribune, and The News, coding every quoted individual by institutional role. State actors dominated, international donors were quoted as often as teachers, parents were absent entirely, and grassroot voices appeared only during strikes.
Beyond Policy on Paper: Using Large Language Models to Detect Anti-Christian Discrimination in Pakistani Social Media Discourse. LLM-based sentiment and discourse analysis of Facebook content about Pakistan’s Christian minority, mapping discriminatory discourse across employment, housing, blasphemy, and social exclusion. The team set this against the PICA Act and minority quota regulations to locate the gap between protection on paper and practice.
Politics, News, and Generative AI: Analyzing how LLMs contain Systematic Bias in Political Content. A comparison of how DeepSeek, ChatGPT, and Claude handle contested issues, including that of Palestine, Iran, the climate, US tariffs, development, using prompts written to mirror journalists across the left–right spectrum in Pakistan, China, and the United States.
The Digital Visibility Index: A Digital Transparency Audit of Sindh’s Provincial Departments. Every Sindh departmental website scored 0–100 on whether it functions, carries accessible contact information, has been updated within twelve months, publishes RTI-mandated disclosures, and offers any interactive channel. The team produced a ranked index, a dashboard, and a policy brief in four days, using a method scalable to other provinces.
Brokering in the Headlines: How different newspapers like Dawn and the Guardian construct Pakistan’s role as mediator in the recent US–Iran conflict, built by turning thousands of articles into a network of who deals with whom. The measure was whose coverage places Pakistan at the center of the story.
The institute closed with a group photograph of participants and organizers.
6. Post-departure
Every participant received a certificate of completion along with school merchandise. Participants remain on the official SICSS Discord, and several groups have stayed in touch with the organizers to continue their projects further, and the cohort remains in occasional contact with one another.
7. Feedback and evaluation
We gathered feedback verbally from all participants on the final day. Participants suggested that a written form would let them respond at greater length, and we will add one for the next edition.
Responses about organization and logistics were positive overall, except for accommodation for out-of-town participants, discussed above. Participants valued the mix of lectures, hands-on tutorials, and project work, and the range of disciplinary perspectives among the speakers.
Two criticisms came through consistently:
Too little social science. Participants across the cohort wanted more dedicated sessions on social science research methods and theory. The program was not devoid of this as the opening day included a two-and-a-half-hour ethics seminar with a group exercise, and several of the guest talks were framed around social questions, but the workshops, which formed the backbone of each day, were weighted toward computational tooling. Given that social science and humanities fields outnumbered computing ones roughly two to one among the cohort, and that the institute exists precisely at the intersection of the two, this is the most substantive piece of feedback we received.
Group projects started too late. Participants wanted group formation and ideation to begin earlier. For this edition, the process began with Friday afternoon’s “Bouncing Ideas” session on day five and concluded with the dedicated speed-dating session on Monday of week two, day six of ten. Teams therefore had four working days between settling on a question and presenting results which they felt was too little time.
8. Plans for the next edition
The organizing team reviewed the feedback in full. Alongside the changes described in place above including rebalancing the two weeks, adding a dedicated technical TA, and coordinating accommodation, we agreed on the following:
Strengthen the social science content: more talks and workshops led by people from social science backgrounds, and a research design workshop covering the range of research methodologies participants might draw on for their projects.
Consider a pre-institute remote bootcamp: we are still weighing a short online bootcamp in the weeks before the institute, to bring participants to a common baseline before the program formally begins.
Ouagadougou, Burkina Faso — August 10-21, 2026
Istanbul, Turkiye — July 6-18, 2026
University of Johannesburg, South Africa — June 28 - July 10, 2026
Nairobi, Kenya — June 15-26, 2026
Melbourne, Australia — June 22 - July 3, 2026
Shanghai, China — June 24-30, 2026
Rotterdam, the Netherlands — June 8-19, 2026
University of Pennsylvania, Philadelphia — June 15-26, 2026
Rochester, NY — May 11-24, 2026
New Brunswick, NJ — June 1-12, 2026
Online — May 25-29, 2026
Singapore — July 13-20, 2026
Stanford University — August 10-21, 2026
University of California, Los Angeles — June 8-12, 2026
University of the District of Columbia, Washington, D.C. — June 14-28, 2026
University of Florida (CJC) — June 15-19, 2026
University of the Republic (Udelar), Montevideo — TBD, 2026
Seattle, WA — July 6-17, 2026
Uyo, Nigeria — July 20-31, 2026
Johannesburg, South Africa — August 17-21, 2026