TRACE: A Metrologically-Grounded Engineering Framework for Trustworthy Agentic AI Systems in Operationally Critical Domains

TL;DR

TRACE is a comprehensive engineering framework for trustworthy agentic AI in critical domains, integrating a layered architecture, trust metrics, and model parsimony principles across diverse applications.

Contribution

It introduces a novel, metrologically grounded framework with explicit architecture, trust metrics, and a focus on model parsimony, applicable across various operational domains.

Findings

Validated across clinical, industrial, and judicial AI applications.

Introduced the Computational Parsimony Ratio (CPR) as a measure of model efficiency.

Demonstrated the importance of explicit LLM validation separation in architecture.

Abstract

We introduce TRACE, a cross-domain engineering framework for trustworthy agentic AI in operationally critical domains. TRACE combines a four-layer reference architecture with an explicit classical-ML vs. LLM-validator split (L2a/L2b), a stateful orchestration-and-escalation policy (L3), and bounded human supervision (L4); a metrologically grounded trust-metric suite mapped to GUM/VIM/ISO 17025; and a Model-Parsimony principle quantified by the Computational Parsimony Ratio (CPR). Three instantiations--clinical decision support, industrial multi-domain operations, and a judicial AI assistant--transfer the samearchitecture and metrics across principally different governance contexts. The L2a/L2b separation makes the use of large language models a deliberate design decision rather than an architectural default, with parsimony quantified through CPR. TRACE introduces CPR as a first-class design principle in trustworthy-AI engineering.