HR Governance × EVIDE

This document records the external domain contribution of Saly Man to the development of the EVIDE framework (External Verifiable Integrity of Decision Events) and the </AI> Protocol, produced by Informatica in Azienda under the direction of Dott. Emanuel Celano.

The contribution was developed through a structured technical exchange spanning March and April 2026, focused on the operational application of EVIDE to HR and AI governance workflows, including AI-assisted candidate screening, human override scenarios, escalation paths, and classification governance under the EU AI Act.

Saly Man operates at the intersection of AI governance, EU AI Act compliance, and HR systems architecture. His expertise in how decisions are formed, reviewed, and documented in high-impact recruitment workflows provided a real-world operational surface for testing and refining the EVIDE schema.

The following seven schema updates were directly shaped by the technical exchange with Saly Man.

The following scenario was developed jointly to map a concrete HR workflow onto the EVIDE evidentiary layer. It remains the clearest illustration of how the two layers interact in practice.

• Step 1 — AI output: AI scores the candidate. A risk signal is triggered: an employment gap is flagged as a negative indicator. A risk register entry is created.
• Step 2 — Initial human review: Reviewer 1 (HR Screener) reviews the AI output and confirms the rejection. Oversight log entry: confirmed_ai_recommendation
• Step 3 — Escalation: The candidate raises a potential bias concern. Case escalated to a Senior HR Officer. Incident log entry created. Escalation trigger: candidate_challenge_received
• Step 4 — Senior review and override: The Senior HR Officer identifies the flagged gap relates to parental leave — a protected characteristic. Override applied. Oversight log entry: override_misclassification_correction
• Step 5 — Final decision declared and closed.

{
  "evide_schema": "1.8",
  "source_system": "HR_Screening_Platform",
  "source_reference": "CANDIDATE-2026-4821",
  "source_timestamp_utc": "2026-04-21T14:30:00Z",

  "decision": {
    "type": "candidate_evaluation",
    "status": "finalized",
    "closure_timestamp_utc": "2026-04-21T14:31:00Z",
    "summary": "Candidate advanced after escalation and override of AI rejection"
  },

  "authority": {
    "id": "reviewer_047",
    "role": "Senior HR Officer",
    "verification": "dapi_certified"
  },

  "intervention": {
    "type": "override",
    "rationale_type": "misclassification_correction",
    "rationale": "AI flagged parental leave gap as risk signal. Senior reviewer identified a protected characteristic. Override applied after escalation.",
    "taxonomy_version": "HR_Taxonomy_v2.1",
    "classification_status": "stable",
    "classification_context": {
      "taxonomy_reference": "https://.../hr-evaluation-taxonomy-v2.1",
      "threshold_reference": "https://.../ai-act-art14-non-discrimination",
      "threshold_status": "met",
      "threshold_authority": {
        "attribution_status": "attributed",
        "authority_ref": "https://.../hr-escalation-policy-v3"
      }
    },
    "trace": {
      "reference": "OVERSIGHT-LOG-2026-4821",
      "access": "restricted"
    }
  },

  "chain": {
    "parent_evide_id": "EVIDE-20260421-0041",
    "chain_type": "escalation"
  }
}

"classification_context": {
  "taxonomy_reference": "https://.../hr-evaluation-taxonomy-v2.1",
  "threshold_reference": null,
  "threshold_status": "not_defined",
  "threshold_authority": {
    "attribution_status": "unknown"
  }
}

If the escalation threshold was not defined upstream, the record still anchors the decision and the override — but it now reveals that no formal escalation threshold governed when the case should be elevated. Not individual failure. A governance gap made persistent.

A recurring finding from the exchange was that most governance failures in HR systems do not come from missing logs, but from the absence of a clearly defined decision condition at the moment of closure.

The minimum anchoring contract defines the conditions under which a decision object is considered sufficiently defined to externalize as accountable. It is not a property of EVIDE itself, but of the interface between the governance layer and the evidentiary layer.

Elements consistently identified across the exchange:

• a stable and closed decision reference
• a clearly attributable authority
• a structured rationale — not only narrative text
• a retrievable connection to the underlying process or evidence
• a level of consequence that makes the decision externally accountable

This contribution is referenced in the External Expert Signals section of the EVIDE public Signals page, under the category 'External Domain Contribution — HR Governance Scenario.'

Signals: EVIDE Signals page

EVIDE JSON Schema: app.certifywebcontent.com/json

Framework: CertifyWebContent.com