Decision Legitimacy Evidence Architecture (DLEA)

Overview

The Decision Legitimacy Evidence Architecture (DLEA) is a recognition-level governance construct that identifies the minimum categories of evidence required for an AI-enabled decision or action to be considered legitimate at the moment it occurs.

It is intended for use in environments where automated or agentic systems act under delegated authority and where post-incident explanation, auditability, or accountability is required. DLEA is concerned with evidentiary legitimacy at time-of-action, not model design, training, or performance evaluation.

⸻

The problem it addresses

In post-incident reviews and audits, organisations often find that although logs, metrics, or model outputs exist, they lack clear contemporaneous evidence of: • who authorised the action, • whether the system was ready to act at that time, • what justified the specific action taken, • and how accountability can later be reconstructed.

This gap is not primarily a tooling or logging issue. It is an evidentiary and governance issue that appears after an action has already occurred.

⸻

What DLEA is

DLEA provides a neutral evidentiary lens for reasoning about AI-enabled actions at time-of-action, organised around four categories: • Authority evidence • Readiness evidence • Action legitimacy evidence • Accountability and record evidence

It helps teams recognise whether sufficient evidence exists to justify an action when it happened, rather than reconstructing justification later.

⸻

What DLEA is not • It is not a framework, standard, or methodology • It does not prescribe controls, tooling, workflows, or implementations • It does not assert regulatory compliance • It does not provide operational or technical guidance

DLEA is intentionally implementation-neutral.

⸻

Intended use

DLEA is designed to support: • internal governance and review discussions • post-incident and audit reasoning • cross-functional alignment between technical, risk, legal, and policy teams

It is particularly relevant for automated or agentic systems operating in regulated or high-impact contexts.

⸻

Status

This repository documents the recognition-level architecture only. Instantiation, conformance artefacts, or licensing details are not included here.

⸻

Authorship

DLEA is authored as an independent governance construct focused on evidentiary legitimacy at time-of-action.

⸻

Optional (only if you want a final line)

This repository exists to make the problem recognisable, not to propose a solution.