The Bayesian equation and psychosis

Abstract

Sir, Chambon et al . (2011) use the Bayesian equation to design an exacting experiment to differentiate the responses of psychotic subjects and controls in a variety of ‘mentalizing’ tasks. Psychotic subjects correctly inferred basic non-social intentions, but in the non-social superordinate condition they relied on prior expectations over sensory evidence, a finding that correlated with positive symptoms. Social paradigms showed less dependence on prior conditions and more reliance on sensory evidence, which correlated with negative symptoms. What conclusions might be drawn if one were to apply their results and the Bayesian equation directly to psychosis in vivo as typically observed in clinical practice? Clinical experience suggests that the errors in judgement in psychosis involve particular combinations and exaggerations of biases found in non-psychotic individuals (Bentall et al ., 2001), coupled with departures from the Bayesian equation, which may be unique to psychosis. In the simplest form of the Bayesian equation, the probability that a belief ‘B’ is true\the probability it is not true, after receiving information ‘D’, is a function of four independent variables arrayed as two ratios, the likelihood ratio and the prior odds (PO) ratio (Fischoff and Beyth-Marom, 1983; Hemsley and Garety, 1986). The likelihood ratio is the information value of data D with respect to belief B being true (D true) divided by the information value of D for B being false (D false). The PO ratio is the likelihood of B being true in light of all experience prior to D (PO true) divided by the likelihood …