Proceedings of the Annual Meeting of the Cognitive Science Society

Human agents constantly exchange information, but deciding which evidence is worth transmitting is a fundamental cognitive challenge. This paper investigates "coherence" as a heuristic for regulating evidence sharing in groups. Extending previous models of individual coherence-based filtering, we employ computational simulations to analyze agents who selectively share evidence based on its coherence with their current beliefs. We find that coherence-based selective sharing is significantly more effective than individual evidence filtering. It helps preserve priors in overly noisy epistemic environments, while still allowing updating when evidence is noisy but truth-conducive. However, this strategy becomes a liability when misleading evidence forms a coherent alternative, steering group learning away from the truth. These results illuminate the conditions under which coherence-based heuristics improve or distort collective inquiry.