The Damping Audit

The Damping Audit is a ward-level containment instrument for AI deployment in 24-hour care settings, published within the Heartbeat Framework Applied Series. The Cultural Safety Case (10.5281/zenodo.21443120) argued that deployment risk is the product of capability and culture, and named a "damping audit" as the missing measurement class. This document ships that instrument. It operationalises the discretion layer — the places where staff absorb, translate, correct, sense, and escalate between the organisation and the person in care — as five node types, and measures what a proposed deployment does to them. The instrument comprises four parts and a rolling panel: a ten-node discretion map drawn on the floor by the people who hold the nodes; an exhaust panel of five lines the service already holds; a nine-indicator relational-load index built around the anchor question "What has happened to the humans around this team since the system arrived?"; and a deployment overlay with three-row gate logic. Escalation rules deliberately mirror the AI Interaction Risk Screen (10.5281/zenodo.21357450), the dyad-level instrument this audit scales to team level — making the Cultural Safety Case's P1 and P4 predictions scorable rather than merely dateable. The sensing node — the instinct layer, the experienced clinician's or carer's unease that precedes measurable change — is treated as pattern recognition running ahead of articulation, grounded in the gut-feelings and nurse-worry literatures, with the deployment-specific failure mode named: systems demote unease from signal to dissent. Evidence Tier C: internally coherent, unvalidated. Each index indicator is designed as a slot a validated psychometric instrument can occupy without redesign. Runnable by an ordinary service without a research department; couples into DCB0160 clinical safety cases and existing inspection layers. A worked example (Fernlea House, pseudonymised composite) runs a full audit cycle including one fired escalation.