Measuring Perceptual Indistinguishability: A Pre-Registered Metamerism Instrument for AI Brand Perception (PRISM-M)

When two entities with structurally distinct multi-dimensional perception profiles receive identical aggregate readouts – one health score, one ranking slot, one binary pick – the distinction the full instrument could see has been destroyed. Existing treatments assert such indistinguishability; this paper measures it. PRISM-M is a pre-registered instrument built on a dual-floor procedure: a pair counts as a measured metamer only when the full readout resolves it beyond the instrument’s own operator noise floor while the aggregate readout, with its own separately measured floor, does not. The procedure is demonstrated end-to-end on AI brand perception: 40 brands from a stratified public-index frame, read through four artifact channels by four cross-family operator pairs, against a battery of three aggregators (scalar score, ranking, binary pick), with planted-positive and same-brand-negative controls. The pre-registered campaign returns a boundary result that exhibits the discipline: one discordant renderer inflates every noise floor, and the instrument reports wholesale sub-resolution rather than manufacturing findings, while its concordance diagnostic localizes the discordance and exploratory analysis of the concordant operator triplet shows the predicted metameric structure – half of resolvable brand distinctions destroyed by a scalar score. Includes zharnikov-2026az-measuring-perceptual-indistinguishability.yaml (Paper Spec v0.1.0) — a machine-readable specification of the paper's claims, assumptions, and dependencies. The paper's full machine-first bundle (the SPINE claim/dependency graph and the ONTOLOGY term module) lives in the public repository; see https://github.com/spectralbranding/paper-spec for the standard. This PDF is generated programmatically from that machine-first source under a research-as-repository model.