The Graduate Workshop in Computational Social Science (GWCSS) has been a core feature of summers at SFI for a quarter-century. This year, in response to the ongoing pandemic, the 26th GWCSS was condensed into two intensive and productive days online.
The assembled Ph.D. students heard lectures from the faculty on computational approaches to social science research rooted in the complex systems methods pioneered at SFI. Students presented their dissertation research and engaged in discussions with the faculty and fellow participants. The virtual workshop culminated in presentations of the homework problem, an essential feature of the GWCSS experience.
The homework, a hackathon-like challenge, presents students with an open-ended question involving complex systems. Participants work in small teams, which form in the afternoon and must present a solution early the next day. “The ability of the students to generate new, productive research ideas in such a short time is remarkable, both to us and, more often than not, to the students as well,” says SFI External Professor John Miller, co-founded GWCSS with External Professor Scott Page, and this year co-directed the program with Page, SFI Professor Mirta Galesic, and SFI External Professor Henrik Olsson.
This year’s homework problem centered around a question of contemporary significance: students were asked to model whether or not individuals decide to get vaccinated during a pandemic. “In under 24 hours, the teams implemented their ideas in fully working models,” says Olsson. “The models captured and explored how individual beliefs about vaccines, the social influence of other people, and government interventions affect vaccine decisions.”
“The variety of models that the groups produced revealed the complexity of vaccination as a complex social system,” adds Page. “You think ‘I know the central modeling assumptions’ then you get in your group and find that others advocate for different assumptions. Your team reaches a consensus and constructs what seems like an ideal model. Then you see the other groups' models and find they've identified a different set of factors and made different assumptions. The collection of models ends up having value independent of each model."
“Modeling these and other factors can easily result in models that are either too simple, in the sense that they do not capture enough of the complexities of the real world, or too complex with too many factors that make it difficult to interpret the results,” says Galesic. “The models produced by the students managed to walk this tightrope very well.”
Arianna Salazar Miranda, a Ph.D. student in urban information systems at MIT and Workshop attendee sums up the experience: “I had the opportunity to work with people from diverse backgrounds, ranging from sociology to information science, who showed me the true value of cross-disciplinary research. I am eager to incorporate some of the modeling principles we learned in my own work. The workshop was a reminder to tackle big questions and to have fun while doing it!”