How do communities of scientists, rather than lone geniuses, produce knowledge? For social epistemology, this question also has a normative side. How should the division of scientific labour be organized? How should a community reward discoveries, share evidence, and settle disagreements? Philosophers of science and social epistemologists increasingly answer these questions with computer simulations. In these agent-based models, simple rules for individuals give surprising results for the group.
This seminar studies these models from two sides. On the philosophical side, we read the foundational papers of the social epistemology of science (Kitcher, Strevens). We also read Weisberg's Simulation and Similarity, which explains what models are and how they explain. On the practical side, we run and change the three paradigmatic frameworks of the field. These are bandit models of networks of scientists, epistemic landscapes, and bounded-confidence models of opinion dynamics. We use the Julia programming language with the Agents.jl package. O'Connor's Modeling Scientific Communities guides us through the literature.
The course has two phases. In the first five weeks, sessions on the philosophy of models alternate with the foundational papers of the social epistemology of science. During these weeks, students learn enough programming to run simple models. In the second phase, a practice session introduces each framework. A philosophical session then examines it: on robustness, on idealization, and on how models relate to their targets. In all sessions, programming serves philosophical questions. What does a model show? On which idealizations do its results depend? Do the results stay the same when the model changes? Students present two published models. In the first presentation, they reconstruct a model. In the second, they change one assumption of a model and assess the consequences. For graded credit, students also write an essay or a referee report.
No prior knowledge of programming or mathematics is necessary. We use AI coding assistants ("vibe coding") to write and change code. We also treat trust in these assistants as a question in social epistemology. We strongly recommend that students bring their own laptops.
- Kursleiter/in: Soong Hwan Yoo