RSE
Explaining the score in plain language. The companion architecture to the AI-IP risk index that provides structured explanations of risk scores for regulatory submission, board reporting, and policyholder communication.
Why this exists — the regulatory anchor
Regulators (PRA, FCA, Lloyd’s, state insurance commissioners) increasingly require explainability for algorithmic risk scoring. RSE provides the structured methodology for explaining AI-IP risk scores in formats that satisfy regulatory submission requirements.
The four-layer explanation methodology
Layer 1 — Score decomposition (which risk-index factors contributed most to the score). Layer 2 — Counterfactual analysis (what would need to change to materially alter the score). Layer 3 — Peer comparison (how this score compares to similar operators in the same sector). Layer 4 — Temporal trend (how this operator’s score has changed over time and why).
Output format
Structured JSON output with natural-language summaries at each layer. Designed for three audiences: (i) board-level summary, (ii) underwriter-level detail, (iii) regulatory-submission-level full methodology.
Integration with AI-IP risk index
RSE consumes the AI-IP risk index’s factor-level scores and produces explanations. Every risk-index recalibration trigger (legislative change, AI capability change, claims pattern change) also triggers an RSE re-explanation.
CIP RSE Architecture v3.20 (Path B expansion v3.23), https://creativeip.org/cari-rse