WTF is Human-in-the-loop Theater?
A human approval step that looks like oversight but is really rubber-stamping things they can't meaningfully check.
Also called: hitl theater, approval theater, accountability sink
TL;DR
"A human approves every change" sounds safe. But if the approver can't actually evaluate a 2,000-line diff, or clicks yes 200 times a day, the step adds friction without safety, and blurs who's really accountable.
How it works
Thoughtworks (Caution, Radar Vol. 35) calls it keeping a human in an agent's loop as an empty approval step rather than a real check, which turns approval into an accountability sink. Recommendations: risk-tier what humans review, and where review exceeds human capacity, shift to technical validation such as mutation testing, property-based testing, formal models (e.g. TLA+) and AI-assisted semantic review.
Origin
- Coined / introduced by
- Thoughtworks Technology Radar
- When
- Nov 2026
- Story
- Flagged as Caution in Radar Vol. 35.