Highlights
- Delusions may be maintained by changed external input and prediction error processing.
- Perceived belief utility and high accordance with related beliefs favor maintenance.
- Systematized delusions may result from deriving auxiliary hypotheses or latent causes.
- How well conflicting evidence is integrated depends on the context in which it is encountered.
- Impaired cognitive capacity and flexibility favor belief persistence but are not sine qua non.
During Bayesian inference, the posterior belief reflects a weighted compromise between the prior belief and the sensory inputs. If the prior belief is relatively less precise (i.e., has a larger variance) than the sensory input, the updated posterior belief is influenced strongly by the input (A).
If the prior belief is relatively more precise than the input, the posterior belief remains very close to the prior belief (B).
In the face of contradictory inputs, a prior belief can be nonetheless maintained by updating an auxiliary belief that is used to explain the prediction error (C).
Similarly, a prior belief can be maintained by actively seeking out new inputs that are more compatible with the prior belief (D).
Contemporary definitions of delusions highlight their resistance to conflicting evidence as the core feature, but there has been little progress in understanding why even explicit confrontation with contradicting evidence seldom leads to belief revision. This review aims to generate new hypotheses on delusion maintenance to inform research and clinical practice.
We systematically reviewed psychological models of belief maintenance and extracted their proposed mechanisms and moderators. The electronic search identified six dissociable perspectives: Bayesian inference, associative learning, utility-informed, cognitive processing informed, dissonance-theory informed, and cross-disciplinary perspectives. The proposed mechanisms involved changing the sensory evidence to fit with one’s prior belief (e.g. shifting attention to contextual cues, changing or reinterpreting sensory input), reducing the relevance of the prediction error (e.g., finding causes of the unexpected event, increasing the abstraction of the prior belief) and reinforcement (e.g. reduction of arousal). The moderators related to prior belief (e.g. its temporal stability, utility, or self-relevance), size of the prediction error, precision of the sensory evidence (e.g. quality, order), context (e.g. typicality, complexity), and individual differences (e.g. cognitive flexibility). We discuss how each proposed mechanism and moderator and their interactions align with current knowledge on delusions and offer compelling and novel explanations for delusion maintenance.
We conclude that delusion maintenance and systematization can be convincingly explained by combinations of prior belief precision, ambiguous input and situational characteristics and may not necessarily require neurobiological deficits or generalized biases.
The hypotheses derived offer multiple new avenues for research and for optimizing the learning process in interventions for delusions.
Surtil : 