People's beliefs about others are often impervious to new evidence: we continue to cooperate with ingroup defectors and refuse to see outgroup enemies as rehabilitated. Resistance to updating beliefs with new information has historically been interpreted as reflecting bias or motivated cognition, but recent work in Bayesian inference suggests that belief maintenance can be compatible with procedural rationality. We propose a mentalizing account of belief maintenance, which holds that protecting strong priors by generating alternative explanations for surprising information involves more mentalizing about the target than nonrational discounting. We review the neuroscientific evidence supporting this approach, and discuss how both types of processing can lead to fitness benefits.
Keywords: Bayesian inference; Theory of Mind network; group bias; mentalizing; moral psychology; prediction error.
Copyright © 2019 Elsevier Ltd. All rights reserved.