Giving the governed a vote

A system can be accurate and still have no right to decide. Legitimacy is the missing word in most AI conversations: not is it correct, but on whose authority does it impose its values, and do the people affected have any say.

What does legitimacy mean for AI?

Legitimacy for AI means the people governed by a system's values have a genuine say in what those values are. We argue endlessly about whether models are accurate and almost never about whether they are legitimate, which is a different and older question: by what right does this system impose this judgment, and can the people it affects contest it. The values layer for AI is where that question has to be answered, and answering it well is part of building trustworthy systems rather than merely capable ones.

Why legitimacy, not just accuracy

Because an accurate system enforcing values nobody agreed to is still an imposition. Accuracy is about getting the answer right within a frame. Legitimacy is about who got to set the frame. A loan model can be statistically excellent and still encode a definition of risk that the people it rejects never had a voice in. Getting the math right does not earn the right to decide.

What a vote could mean

Not a literal ballot for every parameter, but real mechanisms for the governed to shape and contest the values a system applies to them. It is genuinely hard, and I do not pretend it is solved. But a system that makes its values explicit and contestable is already more legitimate than one that hides them and calls the result neutral.