Proceedings of the Annual Meeting of the Cognitive Science Society

How do organizational phenomena emerge from individual cognition? We address this question using a multi-agent simulation in which LLM agents possess individual cogni- tive architectures, including private beliefs, pairwise trust, and social-risk calculation. Using a cognitive process-tracing approach, we prompt agents with universal psychological principles rather than prescribed behaviors, thereby avoiding circularity. We observe the emergence of the paradox identified by Amy Edmondson: higher interpersonal trust leads to significantly greater information disclosure, mirroring the find- ings in human organizations. This macro-level pattern arises purely from individual cognitive modeling without explicit team-level programming. Furthermore, we show that team structure shapes these dynamics: inter-faction trust dramatically reduces information suppression, and trust effects amplify with team size, yielding greater returns in larger organizations. These results demonstrate that psychologically meaningful organizational phenomena can emerge from cognitively grounded agent models, opening a path toward a systematic computational science of organizations.