Artificial Stupidity Guards Human Judgment

This is less a plea to make AI dumber than a reminder that safety is a workflow problem: if humans stop practicing judgment, autonomy becomes theater. Deliberate friction sounds wasteful until the cost of one unchecked edge case dwarfs the saved seconds.

• –Air France Flight 447 is the right cautionary tale: automation can leave people unable to recover when the machine hands control back
• –The piece maps cleanly to scalable oversight and verification debt, where the bottleneck is trustworthy review, not raw output generation
• –Built-in pauses, second-opinion prompts, and occasional forced review can preserve judgment, but only if the friction is targeted and not just noisy latency
• –The incentive problem is real: vendors sell speed and autonomy, while medicine, law, and defense need systems that keep people mentally engaged