The Illusion of Control: Real Human Agency in the Age of AI

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).

The usual test for human control over artificial intelligence is whether a human remains in the loop. That is no longer enough. A human may retain legal responsibility, an override and the final decision. Yet AI may increasingly select the evidence, frame the problem, generate the alternatives and recommend the answer. Recent research on automation bias, cognitive offloading and human oversight shows that this is already a practical issue, not a theoretical one. I call this Effective Authority Drift: decision-making influence moves towards AI while formal authority remains human. The claim is narrow. AI does not need to take the decision away from us to weaken our control over it. It need only shape enough of the process from which the decision emerges. The question is therefore not simply whether a human makes the final choice. It is whether the human retains enough independent capacity to understand the decision, challenge its assumptions and recognise when the choices themselves should be questioned. That is the practical test of real human agency.