A Non-Compensatory Multimetric Framework for Technology Assessment in Regulated Environments

Abstract

Evaluating emerging technologies in regulated environments requires decision frameworks that integrate technical maturity, risk management and compliance. Traditional multi-criteria decision analysis (MCDA) relies on compensatory logic, in which strong performance on some criteria offsets deficiencies in others; outranking methods such as ELECTRE and PROMETHEE do provide veto mechanisms, but their thresholds are elicited from the decision maker and defined on pairwise preference differences, which is incompatible with requirements that are absolute, externally imposed and auditable. This paper proposes a Non-Compensatory Multimetric Framework (NCMF) whose contribution is not the veto mechanism itself but its operationalization: a unary regulatory gate whose thresholds are read from binding legislation (DORA, Solvency II) and instantiated by verifiable evidence in externally audited public disclosures, combined with three indices that provide the component-level instrumentation outranking methods lack — the Integrated Systems Maturity Index (ISMI), the Criticality-Weighted Risk Index (CWRI) and the Regulatory-Gated Mazziotta-Pareto Index (RG-MPI), all derived from a Technology Implementability Index (TII). The framework is validated on two real technology initiatives of a leading insurance group whose authorization outcomes are documented, and benchmarked against AHP, TOPSIS, PROMETHEE II, ELECTRE and the ungated MPI on the same decision matrix. The initiative that the organization actually rejected is technically superior on every technical criterion (ISMI 8.74 vs. 7.82; CWRI 0.0 vs. 54.6) and deficient only in regulatory compliance (75 % vs. 100 % of veto items); consequently AHP, TOPSIS, PROMETHEE II and the ungated MPI all rank it first, in 96.6–100 % of 10 000 Monte-Carlo parameter draws, whereas the proposed framework reproduces both documented outcomes in 100 % of draws. A component-level analysis further localizes 87.9 % of the residual technical risk of the authorized initiative in a single backend component, a diagnostic that no outranking method supplies. The pipeline runs in O(mn^2)time and scales to systems of thousands of components. Transfer to other regulated sectors —banking, medical devices, critical infrastructure— is conditional on the existence of an equivalent mandatory audited disclosure instrument and is stated as a hypothesis requiring sector-specific validation. All code and data are publicly available.