Coherence And Reflective Luck

Revisiting Miščević Through Computational Modeling

Povzetek

This paper examines Miščević’s (2007) hint that coherence considerations can help reduce reflective luck, a form of epistemic luck that stems from an agent’s own cognitive fragility. To do so, we refer to the results from our recently developed computational model, in which we simulated agents who update beliefs probabilistically but may filter evidence based on its impact on coherence. The results show that coherence filtering improves stability in noisy environments by protecting agents from random misinformation, although the exact outcome depends on armchair luck (in our case, this refers to the agent’s prior beliefs). However, in biased environments, the same strategy leads agents to dismiss corrective evidence, reinforcing false beliefs. We argue that coherence is not a universal epistemic virtue, but an adaptive strategy whose success depends on the environment. Our findings extend Miščević’s proposal and highlight the importance of flexibility in belief formation.