This document is a machine-generated Article 86 compliance record produced by the
Friction-Led Intelligence (FLI) dual-ontology decision engine. It demonstrates, by
construction, that the credit rejection recorded above is traceable to exact,
auditable ontology conditions — not reconstructed after the fact. It is intended
for compliance officers, legal counsel, and regulators verifying EU AI Act
Article 86(1) obligations. Full technical documentation and the working prototype
are available at github.com/kadelAnurag/ontology-decision-engine
and TechRxiv DOI 10.36227/techrxiv.175099813.
Section 1
Decision Conditions Triggered
Rule evaluated: Rule_PersonalLoan_Approve
| Condition | Required | Actual Value | Result |
|---|
| Disposable Income | ≥ £40,000 | £28,000 | FAIL |
| Total Savings | ≥ £50,000 | £65,000 | PASS |
| Risk Band | Low Risk, Medium Risk | Low Risk | PASS |
Section 2
Conditions Failed — Plain Language
- Disposable income of £28,000 is £12,000 below the required threshold of £40,000.
Section 3
Applicant Counter-Evidence
Divergence: Company applied an absolute income threshold of 40000.0 against a snapshot disposable income of 28000.0. Client's agent asserts a 5-year average income of 44000, which may better reflect creditworthiness. The threshold's teleological purpose — assessing repayment capacity — may not be served by snapshot logic.
Client context: Salaried employee with 7 years continuous tenure at the same employer. Current disposable income is temporarily reduced due to a one-off salary restructuring during a departmental reorganisation. Five-year average disposable income of £44,000 consistently exceeds the Rule_PersonalLoan_Approve threshold of £40,000. The point-in-time snapshot used by the company ontology does not capture this longitudinal context.
| Counter-Evidence Field | Client-Asserted Value | Significance |
|---|
| 5-year average disposable income | £44,000 | Exceeds £40,000 threshold |
| Years in continuous employment | 7 years | Establishes longitudinal income stability |
Section 4
Ontology Property Reference
challengesAssessment — a formal machine-readable object property in the client-agent ontology. It connects a FrictionSignal node to the specific EligibilityAssessment node in the company decision ontology that produced the rejection. This cross-ontology link is the core primitive of the FLI architecture: it allows regulators to query the exact rule, thresholds, and conditions that were evaluated, without requiring access to the institution's full model or training data.
Section 5
Article 86(1) — Regulation (EU) 2024/1689
Natural persons to whom decisions based on the output of high-risk AI systems listed in Annex III are taken shall have the right to obtain from the deployer clear and meaningful explanations of the role of the AI system in the decision-making procedure and the main elements of the decision taken.
Regulation (EU) 2024/1689 of the European Parliament and of the Council
(EU AI Act), Article 86(1). Official Journal of the European Union,
12 July 2024. Enforcement date: 2 August 2026.
Section 6
Audit Metadata
| Case ID | 79839dbd |
| Timestamp | 2026-04-26 14:44 UTC |
| Product evaluated | PersonalLoan |
| Rule referenced | Rule_PersonalLoan_Approve |
| Risk band (model output) | Low Risk |
| Risk probabilities |
Low 77.8% · Medium 21.9% · High 0.4% |
| Final decision | Reject |
Section 7
ISO 42001:2023 Compliance Mapping
| ISO 42001 Clause | Requirement | FLI Implementation |
|---|
| ISO 42001 Clause 6.1.2 | AI risk assessment | ConditionEvaluation trace — each threshold tested, actual value compared, and pass/fail result recorded as OWL individuals at evaluation time. |
| ISO 42001 Clause 9.1 | Monitoring, measurement, analysis and evaluation | Audit timestamp and Case ID recorded at evaluation time and persisted to decision_feedback.csv and the decision ontology. |