Fig. 2
From: Models of persecutory delusions: a mechanistic insight into the early stages of psychosis
Functional anatomy of social inference: this schematic is an approximation of a neural process model of social inference. The neural signatures of the computational quantities are based on the previous, reproduced fMRI results [62]. a The hidden states that the agent infers on are arranged in a hierarchy as proposed by the HGF. In this graphical notation, diamonds represent quantities that change in time (i.e., that carry a time/trial index k). Hexagons, like diamonds, represent quantities that change in time, but additionally depend on the previous state in a Markovian fashion. From top to bottom, x3 represents the current volatility of the adviser’s intentions, x2 the adviser’s fidelity or tendency to give helpful advice, and x1 represents the accuracy of the current observation (advice or cue). b The inferred states are represented by circles. Thus, based on the empirical findings, we propose the following theoretical neural model of social inference: Cue-related PEs update predictions about the visual outcome and are conveyed via projections from lingual gyrus to posterior parietal cortex, whereas advice PEs, which update the advice accuracy, are passed from low-level regions (including the VTA) to higher-level “theory of mind” regions, i.e., for example, dorsomedial PFC. High-level volatility PEs are further transmitted via the cholinergic septum to cingulate regions. The precisions (advice and volatility) modulate the impact of PEs on medial PFC activity
