Human Magic in the Loop

"Human in the loop" has become a reassurance. Put a person at the end of an automated process, give them a button, and the system can claim human oversight. The reassurance is thin. A person who approves what they no longer understand is a formality, not a safeguard. The phrase describes where the human sits. It says nothing about whether the human is thinking.

The real question is whether people heed what a machine produces or simply defer to it. Heeding means engaging the output, testing it against judgment, and overriding it when judgment says so. Deferring means signing off because the machine is probably right and the meeting is long. As work moves to AI agents, deferring becomes the path of least resistance, and the human in the loop quietly becomes a human in name only.

Human Magic in the Loop is the design principle that answers this. It treats human judgment as something to engineer into a system on purpose, at the moments that matter, rather than something that will appear because a person happens to be present. Where does a decision need a human standard? Where would deferring be expensive? Those points get designed for heeding, with the information, the time, and the authority that real judgment requires.

I developed this idea with Bart Victor, my collaborator on LEGO® Serious Play®, in our 2026 paper "How It Continues." It applies wherever a team now includes machines. The materials here work through where human judgment belongs by design, and how to tell heeding from deferring before the difference costs you something.

Read the paper →