Margaret Mitchell's new paper: AI agents are pushing humans out of the loop
mmitchell_ai · x · 2026-08-28
Margaret Mitchell, Avijit Ghosh, and Samir Passi published a position paper, "AI Agents Push Humans Out of the Loop" (arXiv:2608.23642), arguing that the popular "human in the loop" remedy is failing in the age of AI agents.
Key points:
- Current agent system designs impede effective human oversight, while extended automation use degrades the very cognitive capacities oversight requires — the classic "irony of automation."
- The authors argue overseers' situated goals and cognitive needs should be treated as a first-class design priority, on par with agent capability.
- Drawing on automation and HCI research, the paper outlines design-level affordances and organizational protocols to support critical judgement and counteract skill atrophy, urging developers and deployers to adopt them.
More from Safety
- NIST AI RMF Deep Dive: An Engineer’s Blueprint for Trustworthy AI — iamKierraD · 2026-08-28
- InfoSec's real issue isn't AI, it's institutional memory loss — dyn___ · 2026-08-28
- EU AI Act is now enforceable: What it actually means for compliance — Miles_Brundage · 2026-08-28
- 72% of Americans Say Doctors Must Disclose AI Use, Pew Survey Finds — HealthcareAIGuy · 2026-08-28
- Prompt injection joke targeting humans in original language — dioscuri · 2026-08-28
- AI Security Crosswalk maps OWASP risks to 25 major industry frameworks — brucemacv · 2026-08-28