When Should Algorithms Resign? A Proposal for AI Governance
Umang Bhatt, Holli Sargeant
TL;DR
This paper proposes algorithmic resignation, a governance paradigm that embeds selective disengagement of AI systems into design and deployment. It defines resignation across three factors: System Performance, User Expertise, and Sociotechnical Information, outlining concrete mechanisms like deferral, access control, and contextual policies. The authors argue that internal governance, rather than ex-post enforcement, can reduce risks, improve decision quality, and aid regulatory compliance, by preventing overreliance and ensuring human oversight. They discuss incentives, nudging, and engagement considerations, and illustrate how organizations can tailor AI use to task, user, and context to balance efficiency with safety.
Abstract
Algorithmic resignation is a strategic approach for managing the use of artificial intelligence (AI) by embedding governance directly into AI systems. It involves deliberate and informed disengagement from AI, such as restricting access AI outputs or displaying performance disclaimers, in specific scenarios to aid the appropriate and effective use of AI. By integrating algorithmic resignation as a governance mechanism, organizations can better control when and how AI is used, balancing the benefits of automation with the need for human oversight.