This thesis explores social dilemmas in science and collective cognition more generally, using high-energy physics as a primary example. Through multiple computational case-studies, it examines how scientists in this field navigate dilemmas related to specialization, adaptation, cooperation, and coordination. The core contribution of this research is the development of an "adaptive multi-agent system'' framework for understanding and modelling collective cognition at large. This framework integrates concepts from social epistemology, game theory, and cognitive and complexity sciences in order to investigate how rationally bounded agents, operating within structured social networks, resolve social dilemmas arising in scientific contexts. Three central tenets guide this approach: (1) collective cognition gives rise to emergent social dilemmas with significant epistemic implications, (2) understanding these dilemmas requires a computational framework drawing from a wide range of disciplines, and (3) to better understand collective cognition, we need computational approaches that are both formal and empirical.Three case studies in high-energy physics illustrate different aspects of the framework. The first study examines the growing divide between phenomenologists and theorists, particularly in relation to the theory of supersymmetry. The second analyzes how physicists collectively balance specialization and adaptation in a changing scientific landscape, using Inverse Optimal Transport to assess how cognitive costs shape research trajectories. The final case study investigates the trade-offs and dilemmas involved in the diffusion of scientific conventions by applying a novel statistical physics approach to a sign convention in physics.This thesis offers new perspectives for the philosophy of science, the sociology of science, and computational social science. First, it shows how the notion of collective cognition offers new insights on longstanding issues such as scientific underdetermination and the relation between the cognitive and the social in science. In addition, this thesis contributes to fill a gap between qualitative and quantitative social studies of science, through novel implementations of qualitative insights into quantitative analyses, and by demonstrating the potential of computational models to assist the formation of new philosophical or sociological concepts. Finally, this thesis shows how inverse problems can bridge otherwise disconnected formal and empirical traditions in computational social science. It also suggests that recurrent patterns of self-organization in socio-cognitive systems could be understood functionally as solutions to social dilemmas, opening up new research opportunities in the science of complex systems.
Lucas Gautheron (Tue,) studied this question.