Computational Mechanism for the Effect of Psychosis Community Treatment: A Conceptual Review From Neurobiology to Social Interaction

Abstract

The computational underpinnings of positive psychotic symptoms have recently received significant attention. Candidate mechanisms include some combination of maladaptive priors and reduced updating of these priors during perception. A potential benefit of models with such mechanisms is their ability to link multiple levels of explanation, from the neurobiological to the social, allowing us to provide an information processing-based account of how specific alterations in self-self and self-environment interactions result in the experience of positive symptoms. This is key to improving how we understand the experience of psychosis. Moreover, it points us toward more comprehensive avenues for therapeutic research by providing a putative mechanism that could allow for the generation of new treatments from first principles. In order to demonstrate this, our conceptual paper will discuss the application of the insights from previous computational models to an important and complex set of evidence-based clinical interventions with strong social elements, such as coordinated specialty care clinics (CSC) in early psychosis and assertive community treatment (ACT). These interventions may include but also go beyond psychopharmacology, providing, we argue, structure and predictability for patients experiencing psychosis. We develop the argument that this structure and predictability directly counteract the relatively low precision afforded to sensory information in psychosis, while also providing the patient more access to external cognitive resources in the form of providers and the structure of the programs themselves. We discuss how computational models explain the resulting reduction in symptoms, as well as the predictions these models make about potential responses of patients to modifications or to different variations of these interventions. We also link, via the framework of computational models, the patient's experiences and response to interventions to putative neurobiology.

Keywords: assertive community treatment; computational modeling; coordinated speciality care; hallucination; positive symptoms; schizophrenia.

Figure 1

The perception of an agent that does not hallucinate is shown on the left. The incoming sensory evidence about the environment has a normal confidence ascribed to it and the agent has an expansive policy space without bias toward a specific string of actions. The agent does, however, have a strong prior belief that it will perceive a singing bird. On the right, when precision of incoming sensory evidence is reduced and the forest scene looks blurry, the agent is unable to use the environment to correct for their strong prior belief and hallucinates the singing bird.

Figure 2

On the left, the agent experiences reduced precision of incoming sensory evidence, but it has an expansive policy space without bias toward any string of actions and has equally weighted priors. On the right, the agent experiences an increase in prior precision over policies, and their policy space is such that this results in having a bias toward actions that heavily involve listening and therefore the expectation of sound. This results in the hallucinated songbird because the agent is over-confident in these policies and less certain about other policies which may have implied other possible states of the world.

Figure 3

On the left, the agent experiences incoming sensory evidence to which it ascribes low confidence, and it has equally weighted priors and policies in an expansive policy space. The result is an agent that does not hallucinate. On the right, the agent experiences degraded precision of incoming sensory evidence and a decreased policy space. The decreased policy space now only includes a couple of policies that rely heavily on listening. The result is an agent that does not have a full range of policies adaptable to the world it finds itself in and which imply that a singing bird is present and thus the agent hallucinates the songbird.

Figure 4

On the left, the agent experiences degraded precision of incoming sensory evidence, a strong prior belief, a decreased policy space, and strong confidence in policy that relies exclusively on listening. This agent hallucinates the songbird as it would in the cases presented in Figures 1–3. On the right are four panels, each with a single change from the hallucinatory perception on the left. Each change represents an effect that treatment could have on a hallucinating agent (note that in almost all cases multiple parameters change with treatment, but for simplicity we demonstrate the possible changes independently). The first is improved sensory precision: with clearer incoming sensory evidence, the agent can use its perception of the environment to correct for an over-weighted prior/policy or a maladapted policy space. We hypothesize that the approach of CSCs and ACT teams may help to increase the quality of and confidence in the external environment and as such improve the sensory precision. The second is expanded policy space: with additional actions that the agent could take, it can make further observations about its environment and potentially choose a policy that less heavily implies the presence of the songbird, allowing it to correct for an over-weighted prior/policy and degraded sensory information. We hypothesize that this could occur as a result of cognitive training or therapy and social interactions and external support provided through ACT teams, or that policy space could be preserved via the early intervention provided by CSCs. The third is reduced prior precision over policies: the agent now has equal confidence in the listening action as well as the listening, looking closely, and walking closer action, and as such a weaker prior favoring the presence of the bird. This is the effect hypothesized to be produced by dopamine blockade and helps the agent disengage. The fourth is reduced influence of sensory priors: when the prior beliefs are weighted equally, the agent is not biased toward perceiving the songbird despite the limited policy space and strong prior precision over policies. We hypothesize that this could occur as a result of cognitive training or therapy and social interactions and external support provided through ACT teams, or that the formation of fixed maladaptive priors could be prevented via the early intervention provided by CSCs.