The Psychology of Motivated versus Rational Impression Updating

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. Recent theoretical work suggests that instances of seemingly motivated belief maintenance may be compatible with Bayesian-rational inference. When we have strong prior beliefs, protecting these beliefs from revision by generating an ad hoc auxiliary hypothesis is procedurally rational.Computational neuroimaging work shows that brain regions in and outside of the ToM network are involved in probabilistic belief updating. The medial prefrontal cortex in particular is implicated in updating beliefs about social agents, and updating value representations with information from social sources.We propose that, given its role in both mental state inference and belief updating, the ToM network will be recruited for both rational belief updating and rational belief maintenance, but less so for nonrational discounting of information.