Sun Young Kim (Yeungnam University) has posted Reconstructing Administrative Accountability and Control Authority in AI Government: Designing a Dual-Control Structure with AI-Instigated Human Oversight within a HITL–HIC Framework on SSRN. Here is the abstract:
The proliferation of artificial intelligence (AI)-based automated decision-making (ADM) systems in public administration has significantly enhanced operational efficiency, yet it has simultaneously engendered structural deficits in control authority. Traditional accountability frameworks—predicated on human agency and centered on answerability, evaluation, and enforcement—falter in intelligent governance environments where algorithms execute substantive judgments. This study reframes the core problem not as a lack of accountability but as an ambiguity of control authority, and undertakes normative theoretical research to derive institutional design principles for addressing it. It examines why the Human-in-the-Loop (HITL) approach fails to guarantee substantive control, owing to automation bias and formalistic intervention, and why Human-in-Command (HIC) alone proves insufficient in practice. Drawing on Kant’s categorical imperative, Rawls’s theory of justice, and administrative accountability theory, the study proposes a multilayered dual-control structure—[HOTL×(HITL+HIC)]+AIHO—built around three reconceptualized administrative constructs: Administrative HITL, which institutionalizes human intervention within administrative procedures and responsibility structures; Administrative HIC, which vests final control authority over AI systems in humans; and Administrative AIHO, which enables AI systems to detect anomalies in advance and trigger human intervention as an institutional mechanism. The framework shifts the governance of human–AI interaction from a collaborative model toward a human-centered one, and expands the concept of accountability beyond answerability and enforcement to encompass controllability. Its contribution lies in offering normative and institutional implications for redesigning automated decision-making in AI-based government.
To receive new posts from Legal Theory Blog by email, get a free subscription to Legal Theory Stack.
Lawrence Solum