Epistemically irrational, allostatically adaptive: a predictive processing account of conspiracy theory beliefs

Abstract

Belief in conspiracy theories is often considered to be irrational. The cognitive and motivational mechanisms driving such beliefs remain poorly integrated in leading models of the mind. This paper applies the Predictive Processing (PP) framework to account for two central belief-forming biases associated with irrational conspiracy theory belief: confirmation bias and motivated reasoning. While PP readily explains confirmation bias as a result of precision-weighted Bayesian inference, accommodating motivated reasoning, where desires and emotions shape what agents come to believe, is more challenging. We argue that this gap can be addressed by incorporating insights from the Free Energy Principle (FEP) and research on allostasis (i.e. predictive stabilization in response to stressors). We distinguish between beliefs sustained by cognitive reinforcement of prior beliefs and those shaped by emotional and motivational factors. From the point of view of motivated reasoning, interoceptive and affective priors involve high precision that can affect the generative model. Such motivated reasoning downplays epistemic value and information maximisation, instead focusing on affect-regulation – suggesting that PP, when focusing on metabolic and emotional costs tied to belief formation, can accommodate the role of desires and motivations in shaping beliefs. This work advances how we can understand belief systems under the PP framework and offers a nuanced perspective on the rationality of belief in conspiracy theories. Further, it illuminates why belief in poorly supported and disproven conspiracy theories persists, and presents a refined model of rationality under cognitive and emotional constraint.