Reputation formation and reinforcement of biases in a post-truth world

A potentially biased expert transmits information about a binary state to a decision maker over two periods. The expert is imperfectly informed and is concerned about her reputation for unbiasedness. The decision maker wants his action to match the state in every period, but the true state is never observed. So, he updates his belief about the state by considering both the expert's report and her reputation. The expert now faces two competing incentives -to improve her reputation by disavowing potential bias, and to shift the decision maker's future belief about the state. The game has several novel equilibria. I show that a biased expert may never be disciplined to be honest, while both unbiased and biased experts may lie to signal unbiasedness. A report that disavows a bias is typically seen as conforming to norms, but here it could also arise from reputational incentives to be contrarian.