.jpg)
ABSTRACT
People who spread misinformation in public debates expose others to the risk of forming false beliefs. Excluding them from participation can limit this exposure, but fact-checking takes up resources of time and money, and censorship violates social and political norms. Here, computer simulations of Bayesian learning in social networks suggest that, in some contexts anyway, the epistemic benefits of excluding sources of misinformation might be small or nonexistent, and not worth associated costs. It is shown more specifically that, under certain conditions, open-minded agents in a network can learn just as well in the presence of false ranters: information resistant agents that repeatedly broadcast falsity within the network. Relevant conditions are that the open-minded agents can keep track of their social sources and maintain appropriate levels of trust in them, and that some sufficiently reliable sources introduce truth into the network.