As with many other areas of psychology and philosophy, the study of delusions is taking a social turn. This has different manifestations. For example, in an extremely interesting article, Sam Wilkinson argues that the very attribution of delusional status to certain beliefs is wrapped up in social practices of folk-epistemic approval and disapproval. To call something a delusion, Wilkinson argues, is not to describe – or at least not only to describe – reality but to express a certain kind of disapproval at a violation of social-epistemic norms.
Other work focuses on the centrality of social cognition – and disturbances to social cognition – for understanding the kinds of beliefs that get characterised as delusions in psychiatry. This idea is central to Kengo Miyazono and Allesandro Salice’s fascinating “testimonial theory of delusion,” according to which a combination of testimonial isolation and testimonial discount play important and underappreciated roles in the formation, maintenance, and elaboration of delusions in schizophrenia.
Vaughan Bell, Nichola Raihani, and – again – Sam Wilkinson similarly emphasise this reduced sensitivity to social context in their important manifesto for the importance of coalitional cognition for understanding delusions. Further, they speculate that evolved social mechanisms for managing social influence, affiliation, and strategic social behaviour are central for determining the overwhelmingly social themes of delusions, an idea that also plays a role in Joel and Ian Gold’s “social theory of delusions.”
Our recent preprint, “Bayesian Psychiatry and the Social Focus of Delusions,” is a speculative attempt to connect this social turn in the study of delusions to influential work in computational psychiatry that draws on a conception of the brain as a hierarchically structured statistical inference mechanism.
As we note in our article, we are convinced of the importance and explanatory power of this research programme, which we refer to as “Bayesian psychiatry.” Not only is a conception of the brain as a predictive modelling engine utilising sophisticated forms of statistical inference increasingly well-established in cognitive and computational neuroscience, but this perspective offers a battery of important and illuminating conceptual, theoretical, and methodological tools for understanding the dysfunctions and aberrations that underlie psychiatric disorders.
Despite such attractions, we also argue that Bayesian psychiatry is sometimes tacitly aligned with a conception of the brain as a content-neutral, domain-general learning mechanism that is likely to obscure many of the distinctive ways in which the human mind can break down and malfunction. To illustrate this, we explore some of the most influential attempts to understand psychosis within this research programme, such as those that postulate aberrations in uncertainty-weighted prediction error minimisation and volatility estimation.
We argue that explanations of psychosis that rely on such domain-general learning differences are unlikely to be able to capture aspects of its highly domain specific phenomenology. For example, the overwhelmingly social themes of clinical delusions cluster in a tiny region of the vast space of possible themes that abnormal beliefs could represent, and it is difficult to see how generic dysfunctions in statistical inference could explain this.
To address this, we suggest that Bayesian psychiatry might benefit from accommodating the evolved functional specialisations of the human brain. Of course, such functional specialisations are not realised in discrete self-contained anatomical modules at the macroscopic level of brain structures. Nevertheless, we speculate that Bayesian psychiatry will only be successful to the extent that it recognises that the brain’s statistical algorithms operate in the control centre of a unique primate that evolved to navigate a distinct world of opportunities and risks.