The Boundaries of Automation: A Theory of Persistent Human Participation

Theory paper argues human participation in AI systems is structurally persistent, not merely a capability gap

A preprint from arXiv challenges the assumption that human-in-the-loop arrangements are temporary workarounds for insufficient AI capability. The authors identify three structural reasons human participation may persist even with highly capable AI: complementarity (humans contribute unique capabilities), normative value (participation matters for agency and learning), and 'target emergence' (the goal itself is constituted through the human-AI interaction rather than specified in advance). The paper argues human-AI co-construction is a permanent feature of certain activity classes, with implications for automation limits, system design, and AI ethics.