R0021/2026-03-25/Q007 — Query Definition

Research	R0021 — Prompt engineering definitions
Run	2026-03-25
Query	Q007

Query as Received

What published research exists on AI decision auditing, explainability requirements, or judgment logging in automated AI systems?

Query as Clarified

• Subject: Published research on making AI decisions auditable, explainable, and traceable
• Scope: Academic papers, government programs (DARPA XAI), regulatory requirements (EU AI Act, GDPR Art 22)
• Evidence basis: Peer-reviewed papers, government reports, regulatory texts

Ambiguities Identified

1. "Decision auditing" could mean post-hoc explanation or real-time logging. This research covers both.
2. "Judgment logging" is not a standard academic term. Closest concepts are explainability, interpretability, and audit trails.
3. The field is vast — this research focuses on key programs and regulatory requirements rather than comprehensive literature review.

Sub-Questions

1. What was the DARPA XAI program and what did it achieve?
2. What do GDPR Article 22 and the EU AI Act require for explainability?
3. What is the state of XAI research as of 2024-2025?
4. Are there practical AI audit frameworks in use?

Hypotheses

ID	Hypothesis	Description
H1	Substantial research exists with practical frameworks	Published research on AI auditing and explainability is mature, with deployable frameworks
H2	Research is minimal or purely theoretical	Little published work exists, or existing work is not practically applicable
H3	Research is active and growing but challenges remain	Significant body of work exists, but practical deployment of AI auditing faces unresolved challenges