Talk: Kevin Dorst (MIT)
Location: Geschwister-Scholl-Platz 1, Room A 017
03.06.2025 at 16:00
Title:
A Theory of Confirmation Bias
Abstract:
You exhibit confirmation bias when, on average, searching for evidence in favor of a claim leads you to become more confident of it. It’s pervasive, important, and mysterious. Standard irrationalist theories make confirmation bias inexplicable. Standard (Bayesian) rational theories make it impossible. However, those standard theories presuppose clarity—that we always know what our own subjective probabilities are. I show that once we permit ambiguity—uncertainty about what your own opinions are—confirmation bias becomes inevitable, even for rational Bayesians. I show how this theory can explain many of the empirical trends, including why confirmation bias is so hard to eliminate.