Algorithmic Impact Assessments Across Government Agencies

Abstract

Background: Recent developments involving algorithmic impact assessment have made implementation quality, oversight, and administrative capacity central policy concerns.

Objective: The analysis examined how policy relating to algorithmic impact assessment operated in practice and whether differences in capacity and oversight were associated with outcomes in relation to government.

Methods: A structured review covered 612 statutes, regulations, policies, contracts, guidance documents, administrative records, or public disclosures. A predefined two-stage coding framework assessed legal clarity, transparency, procedural safeguards, oversight, and implementation requirements, with discrepancies resolved through independent review.

Results: Although 77% of reviewed materials contained formal provisions on algorithmic impact assessment, only 57% translated those provisions into sufficiently detailed implementation or oversight requirements. Transparency and review were deficient in 22% of cases, while more comprehensive instruments tended to assign responsibility more clearly and leave fewer gaps.

Conclusion: The results shift attention from whether rules on algorithmic impact assessment exist to how those rules are administered in practice. This means that AI governance in the public sector needs enforceable controls, capable staff, documented accountability, and meaningful human review.

Study design: agency document audit.

Publication history: Originally published in print on 15 June 2024, Volume 4, Article 6. Online abstract record added on 16 September 2026. Full-text PDF will be added when the digital galley is ready.