Conjecture

Standing and harm-internalisation are not rival constraints on correction but two legs of one loop, and they run in series rather than in parallel. The loop closes only if the cost of an error reaches a party who can impose cost back on whoever decides. Internalisation is the proximate mechanism: it shortens the loop to zero length by making the decision maker the harmed party, which is why correction becomes routine where it holds. Standing is the upstream mechanism: it is how the internalisation channel gets installed in the first place, because liability, licensing and disclosure rules are themselves won by parties able to protest. The AI-as-normal-technology record supports both halves and neither alone: garment fire safety, meatpacking food and worker safety, steamboats, mining and early aviation all raced to the bottom while victims existed and were harmed, because "it is hard for consumers to assess product safety", and the race went away "once regulation forces companies to internalize the costs of their safety practices". Prediction: ranking the two constraints is a question about time horizon, not about which is real. Over a decision cycle internalisation predicts where correction capacity is built; over an institutional cycle standing predicts where internalisation gets imposed.

Institutional design