Inferential Pharmacology: A Framework for the Diagnostic Use of Pharmacological Response in Psychiatry

This is a preprint. It has not undergone peer review. Background. Pharmacological response in psychiatry is conventionally evaluated as a therapeutic outcome, while its informational dimension — what the quality, pattern and temporal dynamics of the response reveal about the structural organisation of the psychopathological system — remains largely unsystematised in formal diagnostic reasoning. Framework. This paper proposes and operationalises inferential pharmacology, which reinterprets pharmacological response as a structured evidential event updating diagnostic probability distributions over competing hypotheses, extending the ex juvantibus tradition with an explicit probabilistic foundation. The framework is formally analogous to a single-subject (N-of-1) design with explicit pre-intervention predictions, and draws on four traditions: ex juvantibus reasoning; predictive processing and the aberrant salience model; Fava’s behavioural toxicity framework; and clinical analogues from geriatric delirium and alcohol use disorder pharmacotherapy. Key constructs. A structured instrument — the Pharmacological Response Observation Grid (PROG) — organises observation along six dimensions, with a six-step protocol translating observations into explicit Bayesian updating. A four-category taxonomy classifies responses by diagnostic content: specific, non-specific, paradoxical, and absent or subthreshold. The concepts of epistemic cost of symptomatic suppression and high-entropy diagnostic prior are introduced, with alcohol use disorder as paradigmatic illustration, integrating the tripartite craving taxonomy and the contested OPRM1 A118G polymorphism into the PROG prior. Conclusions. Pharmacological response is proposed as a structured diagnostic resource with defined validity conditions: the clinician does not “try a medication” but conducts a documented inferential event with falsifiable predictions on a single subject.