Do psychiatrists have a spider-sense? Clinical impressions as violations of interpersonal Bayesian expectation

Abstract

Psychiatric diagnosis depends on the clinician’s ability to infer mental states from complex and dynamically evolving interpersonal signals. During the psychiatric interview, clinicians do not only collect explicit symptoms; they continuously sample a rich array of non-verbal and paralinguistic cues including gaze, prosody, affective reciprocity, and interactional timing. These signals are central to psychiatric assessment, yet their contribution to diagnostic reasoning remains poorly formalized.

In this Review, we propose a computational account of clinical impressions in psychiatry. We integrate evidence from phenomenological psychopathology, social cognition, and computational neuroscience. We argue that such impressions may arise when the clinician’s generative model of social interaction is persistently disconfirmed by subtle deviations in interpersonal behaviour, generating high-level social prediction errors experienced as a sense of incongruity or failed attunement.

We formalize this process as VIBE: Violation of Interpersonal Bayesian Expectation. VIBE denotes a pre-reflective epistemic signal generated when observed interpersonal cues deviate from the clinician’s expected model of social reciprocity. It is not a diagnostic criterion but orients hypothesis generation by drawing attention to disturbances in interpersonal predictability. We hypothesize that VIBE may constitute a transdiagnostic mechanism: different disorders, from schizophrenia to autism spectrum disorder and mood disorders, may disrupt distinct dimensions of interpersonal prediction. The Praecox Feeling, first described by Rümke in 1942, can be read as the earliest clinical formalization of this process and as its schizophrenia-spectrum instance, in which the profile of interpersonal violations is consistent enough to be clinically recognizable.

This framework recasts the diagnostic impression as an embodied inferential process rather than an opaque subjective faculty. Teaching clinicians to articulate, examine, and recalibrate these impressions could strengthen diagnostic reasoning, reduce bias, and improve therapeutic attunement. VIBE provides a bridge between phenomenological psychopathology and computational neuroscience, and offers a tractable, empirically testable model for studying the neurocomputational basis of psychiatric clinical judgement.