Computational Models of Psychotic Phenomena

Computational models of psychotic phenomena seek to characterise the mechanisms by which aberrant perceptual and cognitive processes give rise to hallucinations, delusions and related symptoms. Central to this endeavour is the notion that perception and belief emerge from the integration of prior knowledge with incoming sensory data via Bayesian inference. When the precision assigned to priors or sensory evidence is misweighted, prediction errors may be amplified or attenuated, leading to maladaptive updating of beliefs. Hierarchical frameworks extend this view by positing multiple levels of inference, such that alterations at lower sensory layers need not map linearly onto higher cognitive layers. Models of circular inference describe how feedback and feedforward loops can corrupt signals, ‘seeing what we expect’ and ‘expecting what we see’. Other approaches use hidden Markov models or neural network simulations to capture transient states that resemble hallucinations or delusional conviction. Together, these computational accounts offer a unifying language for describing diverse psychotic symptoms, suggest biomarkers for early detection and point towards personalised interventions that target specific inference deficits.

Research from Nature Portfolio

Recent studies have provided empirical support for circular inference as a mechanistic account of schizophrenia. In one influential experiment, patients and healthy controls performed a forced-choice task that varied the strength of sensory cues and prior information. Behavioural responses were best explained by models incorporating ascending and descending inference loops. Increased ascending loops correlated with the severity of positive symptoms, while enhanced descending loops correlated with negative symptoms, and both loops tracked disorganised symptoms. This work establishes circular inference as a bridge between computational theory and clinical presentation, suggesting that distinct symptom domains arise from specific patterns of corrupted feedforward or feedback processing.

Computational Models of Psychotic Phenomena publication trend

The graph below shows the total number of articles in computational models of psychotic phenomena across all publications each year (not limited to Nature Index journals).

Technical terms

Predictive coding: A theoretical framework in which the brain minimises the difference between expected and actual sensory input by adjusting internal models.

Prediction error: The discrepancy between predicted and observed sensory signals that drives learning and belief updating.

Circular inference: A process whereby feedback and feedforward signals become mutually corrupted, leading to over-reliance on expectations or sensations.

Perceptual prior: An internally held expectation about the likelihood of sensory events, which influences perception under uncertainty.

Precision weighting: The assignment of confidence to sensory data or priors, determining their influence on inference.

False alarm: An instance in experimental tasks where a signal is reported in the absence of a corresponding stimulus, used as a proxy for hallucinations.

References

1. Different learning aberrations relate to delusion-like beliefs with different contents. Brain (2024).
2. Functional connectivity and glutamate levels of the medial prefrontal cortex in schizotypy are related to sensory amplification in a probabilistic reasoning task. NeuroImage (2023).
3. Robotically-induced auditory-verbal hallucinations: combining self-monitoring and strong perceptual priors.. Psychological Medicine (2023).
4. Experimental evidence for circular inference in schizophrenia. Nature Communications (2017).