After the Machine

This record publishes After the Machine — the AI Interaction Outcome Profile (AIOP), the sixth instrument (A6) in the Applied Series of the Heartbeat Framework, at v1.0. Every control currently governing conversational AI in health and care assesses the technology; none instruments the person. AIOP is a longitudinal outcome instrument for adults in 24-hour care settings who use conversational AI, designed to measure what sustained interaction does to the person over time. It is bipolar by construction — every axis runs from a harm pole to a benefit pole — and person-referenced by principle: trajectory is read against the individual's own pre-AI baseline, never against a population norm. Two parallel lanes, self-report and observer, feed a change-over-time log that makes every administration auditable; a one-time baseline module captures the person's normal before the machine; an optional collateral supplement reads the longitudinal memory that families and closest friends hold. Six domains carry the profile: displacement–supplementation (including unaided capability retention), agency and control, epistemic stance, regulation in absence, human alliance, and disclosure pattern. Disagreement between lanes is scored, not averaged away — the second reading. A rapid-triage red route interrupts the measurement machinery entirely wherever indication of harm surfaces, handing off to the statutory and clinical frameworks that already govern the setting. The instrument positions directly into the NHS assurance landscape — DCB0160/0129, DTAC v2, the NICE Evidence Standards Framework, HoNOS/MHSDS, and CQC inspection — as the standardised artefact that "clinical monitoring in place" mitigations can cite. Anonymised aggregation of profiles across a team yields the ward-level signal without a second instrument — one measure, two altitudes — completing the sibling relationship with The Damping Audit (A5): the ward and the person. AIOP launches at Tier C evidence: design rationale and internal coherence. It is not yet empirically validated, not normed, and not in routine clinical use. The paper states a pathway to Tier B in ESF-mappable terms, names its confounds — incremental validity, reverse causality, recall bias — and registers three falsifiable predictions scored against the log the instrument itself generates. The full administration forms, scoring sheet, and a worked example are included. Deposited alongside its record-keeper: The Glass Ledger v2 (DOI 10.5281/zenodo.21515861), the tamper-evident journal the change-over-time log is designed to be compatible with. Screen → AIOP → Ledger: assess, measure, record. Paul Blatherwick, RMN. CC BY 4.0.