Human-AI Collaboration Needs More Than a Human in the Loop: Checkpoint-Based Governance (CBG)

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

This paper asks what Human-AI Collaboration should be and who answers for what it produces. It defines the collaboration as human direction, AI execution, and a named human who decides at a checkpoint and answers for the result. It traces the idea of human oversight and accountability from the cybernetics literature of 1948, and the formal requirements that followed in engineering, audit, law, and policy, to Article 14 of the EU AI Act. It then separates Responsible AI, where the checking is done by machines or by a human who only watches, from AI Governance, where a named human holds binding authority. Current practice falls short of that line, from workplace surveys that report agents acting without real-time human involvement to research that tests AI as a tool without recording the method of use. The author's practice is offered as one answer, not the only one. It begins with a 2012 method of pairing every fact with a tactic and a measurable outcome, which the author calls Factics, carried into AI work, and adds Checkpoint-Based Governance (CBG), which differs from human-in-the-loop review by requiring binding authority, a decision record, and a measured check on rubber-stamping. The paper proposes studies the author's searches have not found in the literature. It argues that the future of work is an operating model that leads with growth beyond efficiency, which the author calls the Growth OS, practiced as Augmented Intelligence, where people use AI under CBG as an amplifier, not a replacement. It closes by presenting governance as both a safety net and a cognitive tool. CBG has not been independently validated, and the author discloses his conflict as the developer of the frameworks described.