A geodesic exposure unit for institutional AI liability.
Actuaries need an exposure basis for AI liability where loss history is thin and model-level metrics are insufficient. The holonom measures normalized avoidable institutional distortion accumulated by an AI-enabled decision as it moves through banking, insurance, legal, audit, and evidentiary pathways — designed for rating, reserving, and portfolio accumulation.
A holonom is a normalized exposure unit for avoidable institutional distortion in AI-enabled decision pathways.
P = P₀ + λ·ΣH. A candidate additive risk charge grounded in evidence quality, defensibility, and coverage ambiguity.
R = R₀ + α·H_claims. Path-dependent claim friction feeds LAE, defense-cost load, and uncertainty margin.
ΣH rolls up by LOB, use case, severity, evidence quality, and defensibility class — an aggregate exposure view, not a governance score.
Applies wherever AI-enabled decisions transport into insurable outcomes.
Existing AI risk proxies are weak. Holonoms are stronger.
| Number of models | Ignores institutional use |
| Number of decisions | Ignores severity and evidence |
| Revenue | Too blunt |
| Users | Not liability-specific |
| API calls | Actuarially shallow |
| Governance maturity | Qualitative, not exposure-based |
Measures normalized avoidable institutional distortion per transported decision. Path-dependent. Evidence-aware. Designed for rating, reserving, and portfolio accumulation where historical loss data is sparse.
Theory note, rating worksheet, LOB mapping, claims appendix.
This web tool — score decisions, indicate premium, reserving lens.
Automated ingestion of AI decision logs and evidence bundles.
Reinsurance-grade aggregate H, correlation, capital load.
Actuarial, mathematical, legal, and insurance advisors are invited to critique the unit definition.
Holonom v0.1 is an indicative exposure model, not a filed rating plan or an actuarial standard of practice. The framework hardens through adversarial review before commercialization.
From model risk to measurable exposure units.
Open the workbench, score a sample portfolio, and see the holonom pricing and reserving indications update live.
Holonom.xyz defines the unit. Evidence-infrastructure implementation: Bankabil.