Institutional safeguards and revisability

Compare appeal, independent review, sunset and exit mechanisms for an AI-assisted public service.

Imagine a public agency adopting an AI system that recommends who receives a scarce service. Today’s designers disagree about fairness and future circumstances will change. What would let affected people notice mistakes and obtain a meaningful remedy?

Viatopia is an orientation toward excellent futures, rather than one fully specified destination. Here we explore four possible safeguards. They require capacity, legitimate authority and evidence; checking a box does not make a deployment safe.

Revision has costs. Constant reopening can obstruct reliable services, and review bodies can be captured too. Ask who can trigger change, under what evidence and with what protections.

Which safeguard would fail first in your context, and what would make it work in practice?

Interactive tools load with JavaScript. Original sources retain their language and attribution.