Closing the Street-Level Discretion Gap: A Bureaucratic Calibration Framework for Algorithmic Decision Support in Means-Tested Benefit Administration

Abstract

Algorithmic decision-support systems (ADSS) increasingly mediate frontline caseworker determinations in means-tested public benefit programs, yet institutional frameworks governing street-level discretion calibration remain underdeveloped. This study examines the structural misalignment between automated eligibility scoring and human administrative judgment in welfare administration across three OECD jurisdictions. Drawing on a mixed-methods design combining administrative microdata analysis (n = 84,630 casefiles) and semi-structured interviews with 62 frontline administrators, we identify systematic discretion compression—a condition in which ADSS output constrains caseworker agency below legislatively mandated thresholds. Our findings demonstrate that procedural legitimacy deficits arise when algorithmic outputs are treated as binding rather than advisory. We propose a Bureaucratic Calibration Framework (BCF) operationalizing tiered override protocols, auditability benchmarks, and discretion-preservation mandates, offering actionable governance pathways for public administrators navigating automation-accountability tensions.