Human Endorsement in Automated Government

Abstract

Claims that human endorsement legitimises failed automation need evidence connecting an act of authorisation to the process it is said to legitimate. This comparison of England's Ofqual grading arrangements and Australia's Robodebt scheme tests those connections in Someyo Kamal Utsho's account, building on existing pairedcase and Robodebt scholarship. Ofqual reports model evaluation and student-level outcome analysis, but the inspected sources neither establish nor disprove the alleged ministerial endorsement and publication without commissioning a substantive equity audit. The Royal Commission into the Robodebt Scheme reconstructs an automated route that contradicts uniform officer-confirmed processing. Colleen Taylor's testimony separately identifies her team's calculation-focused task; the Commission's own adverse finding concerning Online Compliance Intervention calculation and notification remains. Branch and task differences qualify the evidence for endorsement without establishing the absence of every formal duty. The earlier proposal provides a constructive comparison: cumulative evidence supports named knowledge findings and an instruction that advanced the proposal, while later failures to inquire and inform Cabinet remain. Such consequential participation is compatible with Utsho's concern about upstream authority, but does not itself establish deference or downstream reliance on a signature. The contribution is a source-specific explanation of what supports an endorsement claim and what documented participation leaves unresolved.

How the inquiry is constructed

A source-specific comparison of England's Ofqual grading arrangements and Australia's Robodebt scheme tests endorsement claims in Someyo Kamal Utsho's account against the records examined, including Royal Commission findings and testimony. It builds on existing paired-case and Robodebt scholarship and distinguishes branches, tasks, knowledge and actions. The analysis does not establish that every formal duty was absent, nor that documented participation proves deference or downstream reliance on a signature.

What the analysis establishes

• An endorsement claim requires evidence connecting an act of authorisation to the process it is said to legitimate.
• The inspected Ofqual sources report model evaluation and student-level analysis but neither establish nor disprove the specific alleged endorsement and publication without a substantive equity audit.
• The Robodebt Royal Commission reconstructs an automated route inconsistent with uniform officer-confirmed processing; calculation-focused work and the Commission's adverse finding require separate treatment.
• Differences between branches and tasks qualify endorsement claims without showing that every formal duty was absent.
• Evidence of knowledge and participation in an earlier proposal does not by itself prove deference or downstream reliance on a signature; later failures to inquire and inform Cabinet remain relevant.