We model the dynamics of discrimination and show how its evolution can identify the underlying cause. We test these theoretical predictions in a field experiment on a large online platform where users post content that is evaluated by other users on the platform. We assign posts to accounts that exogenously vary by gender and history of evaluations. With no prior evaluations, women face significant discrimination, while following a sequence of positive evaluations, the direction of discrimination reverses: posts by women are favored over those by men. According to our theoretical predictions, this dynamic reversal implies
discrimination driven by biased beliefs.