Abstract
Although there have been so many studies on schizophrenia under the framework of predictive coding, works focusing on treatment are very preliminary. A model-oriented, operationalist, and comprehensive understanding of schizophrenia would promote the therapy turn of further research. We summarize predictive coding models of embodiment, co-occurrence of over- and under-weighting priors, subjective time processing, language production or comprehension, self-or-other inference, and social interaction. Corresponding impairments and clinical manifestations of schizophrenia are reviewed under these models at the same time. Finally, we discuss why and how to inaugurate a therapy turn of further research under the framework of predictive coding.
Introduction
Schizophrenia research has historically been fragmented, with neurobiological findings, clinical symptoms, and subjective patient experiences often treated as separate entities. This "split" nature hinders the development of unified therapeutic approaches. Lingyu Li and Chunbo Li propose that the Predictive Coding (PC) framework can bridge these gaps by viewing the brain as an active "prediction machine." This machine constantly balances internal models (priors) against incoming sensory data (error signals). The authors argue for a transition from purely descriptive research to a "therapy turn," where computational models directly inform clinical interventions.
The Predictive Coding Framework
The PC framework suggests that the brain minimizes "surprise" or free energy by updating internal beliefs based on prediction errors. In schizophrenia, this process is disrupted. A key insight is the Paradox of Priors: symptoms like hallucinations and sensory overload suggest both over-weighted and under-weighted priors. The authors resolve this by proposing that these imbalances coexist at different levels of the neural hierarchy. For instance, weak low-level sensory priors might lead to sensory flooding, which the brain attempts to compensate for with overly rigid high-level inferences.
Sensorimotor Processing and Embodiment
Schizophrenia often involves motor dysfunction and a blurred sense of self. Under the PC model, the brain predicts the sensory consequences of its own movements. When these forward models fail, the brain relies on retrospective "backpropagation" to explain actions, potentially leading to the feeling that one's limbs are controlled by external forces.
Patients also show increased sensitivity to the Rubber Hand Illusion (RHI), attributed to decreased weights for high-level priors regarding body representation. This makes them more susceptible to misidentifying external objects as part of themselves.
Language and Communication
Language disorders, such as Formal Thought Disorder, are mapped onto three hierarchies: phonetics, semantics, and pragmatics. The schizophrenic brain is described as "rigidly Bayesian," where an inability to use context (pragmatics) to constrain meaning (semantics) leads to "semantic disinhibition."
Auditory Verbal Hallucinations (AVH) are viewed as a combination of strong phonetic priors and a failure to access contextual memory, causing the brain to misattribute internal speech to an external source.
Toward a Therapy Turn: The "Digital Twin"
The ultimate goal of this framework is to move beyond description toward personalized medicine. The authors propose the concept of a "Digital Twin"—a computational model of a specific patient's predictive coding parameters. Clinicians could use these twins to simulate the effects of antipsychotics or psychotherapy, optimizing treatment before implementation. This operationalist approach aims to unify the "triad" of body, language, and social interaction into a single, treatable model.
Summary updated