The computational relationship between reinforcement learning, social inference, and paranoia.

Area of Science:

• Cognitive Neuroscience
• Computational Psychiatry
• Social Psychology

Background:

• Heightened uncertainty about the world's state is theorized to drive aberrant belief updates, increasing persecutory delusion risk.
• The specific relationship between an agent's uncertainty and the phenomenology of paranoia, versus other belief types, remains unclear.

Purpose of the Study:

• To investigate how uncertainty influences belief updates in both social and non-social contexts.
• To examine the impact of paranoia on responses to changing contingencies in social and non-social tasks.
• To model the sociocognitive mechanisms underlying persecutory delusions.

Main Methods:

• Utilized a probabilistic reversal learning task and a modified repeated reversal Dictator game with 693 participants.
• Employed computational models to quantify rigidity across reversals and uncertainty about the environment/partner.
• Assessed the impact of pre-existing paranoia on task performance and model parameters.

Main Results:

• Paranoia correlated with uncertainty about a partner's behavior and rigidity in attributing harmful intent in social tasks.
• In non-social tasks, paranoia was associated with higher decision temperatures and suboptimal choices.
• Decision temperature in non-social tasks related to priors of harmful intent and uncertainty in social beliefs.

Conclusions:

• Paranoia is associated with a pervasive uncertainty about the world and its agents, which is slower to resolve.
• This uncertainty is expressed differently in social contexts compared to non-social ones.
• The findings offer a model for understanding persecutory delusions and suggest future causal experiments.