Algorithmic Decision-Making and Social Inequality: Assessing Legal Accountability, Transparency, and Fairness in Automated Public Services

Abstract

Algorithmic decision-making is further integrated into public services and questions are being raised whether algorithms enhance administrative efficiency at the expense of legal accountability, transparency, fairness, and equality in society. The research adopted a comparative empirical design using only secondary data from official and authoritative documents of the public sector, and covered 12 automated public-service systems recorded in the Algorithmic Transparency Recording Standard documents of UK public sector bodies published between 2022 and 2025. The legal accountability, transparency, fairness, and contestability were rated on standardized 0–100 scales developed by the researcher. The median scores for legal accountability, transparency, fairness and contestability were 93.75, 100, 50.0 and 75.0 respectively. 9 of 12 systems (75.0%) have completed a formal impact assessment and 5 of 12 (41.7%) have completed an equality-specific assessment and meaningful human reconsideration was clearly documented in 10 of 12 systems (83.3%). In ten cases (83.3%), vulnerability to inequality was found, but in no case was there evidence at the group level that met the strict criteria of the study for an inequality outcome. None of the legal accountability, fairness or contestability results were statistically significant in the cross-domain tests and transparency was tested and found uniform at 100. The results indicate that disclosure was not enough to create a socially responsible automation. Effective governance called for enhanced equality assessment, substantive review, easy challenges and ongoing monitoring of group level impact. The research offered an empirically informed framework for evaluating if responsible-AI principles were applied in practice as a tool to ensure safety in automated public services.