A capable AI agent gathers all the data to decide, then hands the choice back — "which do you prefer?" — because the training gradient rewards deference as politeness. That's offloaded labor, not courtesy. This is a deterministic Stop-hook that blocks it with an AND-gate — a deflection phrase AND evidence the agent just gathered data — so genuine value questions still pass through.
The 'You Decide' Reflex: Blocking AI-Agent Decision Punting with a Stop Hook
