Signature Over Substance: Homeopathy as a Diagnostic-Moderated Information Intervention

Abstract

Homeopathy remains scientifically contested because ultra-high dilutions challenge substance-based causal models and reported outcomes are heterogeneous across indications and study settings. We translate the controversy into preregisterable discrimination tests that separate non-specific context effects from any diagnosis-specific add-on attributable to a structured intervention signature under matched context. The framework is mechanism-open and defines a preregistered primary outcome as the change in a prespecified deviation proxy (a composite is allowed with weights fixed a priori). The key prediction is diagnostic moderation: if a diagnosis-specific add-on exists, it should strengthen as preregistered diagnostic certainty increases, where diagnostic certainty is computed independently of outcomes (preferred: a preregistered latent-class measurement model; fallback: prespecified agreement-based proxy strata). The proposed matched-context three-arm design compares a diagnosis-congruent protocol, an otherwise identical placebo, and a diagnosis-incongruent mismatch protocol; the mismatch arm is governed by an explicit equivalence margin as a falsification device for expectancy artifacts. Validation combines an in silico suite (type-I error, power, and a blinding-leakage stress test) with a secondary consistency check using published risk-of-bias reliability strata. This manuscript is a modelling and preregistration framework with simulation validation; it does not claim clinical efficacy.