The effect of uncertainty on prediction error in the action perception loop

Area of Science:

• Cognitive Neuroscience
• Computational Psychiatry
• Human Agency Research

Background:

• Distinguishing self-generated sensations from external causes is fundamental for agents.
• Agency inference becomes particularly challenging under conditions of uncertainty.
• The predictive processing framework suggests active control of prediction error closes the action-perception loop.

Purpose of the Study:

• To investigate how prediction error is minimized under varying levels of uncertainty and volatility.
• To examine the impact of uncertainty on agency judgments, movement, policy selection, and hypothesis switching.
• To explore the relationship between autism traits and these processes.

Main Methods:

• Utilized a novel, temporally-sensitive behavioral proxy for prediction error.
• Manipulated conditions of uncertainty and volatility.
• Assessed agency judgments, movement characteristics, policy selection, and hypothesis switching.
• Measured autism traits in participants.

Main Results:

• Prediction error was minimized most rapidly when volatility was high and agency was reported, irrespective of judgment accuracy.
• Uncertainty broadly affected the accuracy of agency judgments, movement, policy selection, and hypothesis switching.
• Individuals with higher autism traits showed altered policy selection, uncertainty sensitivity, and hypothesis switching, but not overall accuracy.

Conclusions:

• The minimization of prediction error is influenced by perceived agency and environmental volatility.
• Uncertainty significantly modulates multiple aspects of cognitive control and decision-making.
• Autism traits are associated with distinct patterns of policy selection and uncertainty processing, highlighting potential computational differences.