Active Inference and Auditory Hallucinations.

Area of Science:

• Computational psychiatry
• Neuroscience of perception
• Cognitive modeling

Background:

• Auditory verbal hallucinations (AVH) are distressing symptoms in neuropsychiatric disorders like schizophrenia.
• Perception involves balancing prior beliefs with sensory evidence to form hypotheses.
• An imbalance where prior beliefs dominate can lead to false inferences, such as hearing voices without auditory input.

Purpose of the Study:

• To develop a novel computational model of AVH using active inference.
• To explore the mechanisms underlying false inferences in AVH.
• To formally express computational underpinnings of AVH and their potential neurobiological links to schizophrenia.

Main Methods:

• Formulation of AVH as a false (positive) inference within a Markov decision process framework.
• Modeling perception as hypothesis testing using sensory data to disambiguate world states.
• Simulations to investigate how prior beliefs about policies and sensory reliability influence perceptual inference.

Main Results:

• Demonstrated that AVH can occur when an agent expects to hear a voice amidst imprecise sensory data.
• Showed that the content and confidence of prior beliefs are influenced by beliefs about listening/talking policies and sensory data reliability.
• Identified computational mechanisms leading to an imbalance where prior beliefs dominate perceptual inference.

Conclusions:

• The proposed active inference model offers a formal account of AVH computational mechanisms.
• The model highlights the critical role of the interplay between action and perception in AVH.
• Findings suggest potential mappings between the model and neurobiological changes observed in schizophrenia.