Bayesian Doublespeak

Abstract

We propose a theory for why rational agents are susceptible to misinformation and use it to address applied questions about how to combat misinformation. Such agents cannot fully distinguish misinformation from truthful information due to uncertainty about a sender's motives. Thus, even overwhelming amounts of truthful information may not lead agents to learn the truth. In equilibrium, misinformation takes the form of "doublespeak," or messages consistent with the truth in a different state of the world. Restricting communication length and increasing reputational concerns may limit doublespeak, but fact-checking does not. Our results explain the powerful persistence of misinformation.

Keywords: Misinformation, Disagreement, Polarization

JEL Classification: D82, D83