Can a Machine Be Made to Doubt?

The Turing test asks a machine to imitate confidence convincingly. Doubt is the harder performance, because doubt is not the absence of an answer — it is the active maintenance of two answers at once, and the refusal to collapse them prematurely.

A model that hedges is not doubting; it is averaging. Real doubt has a cost. It stalls the pipeline, it declines the reward, it says not yet to the thing that would score well. We have built systems that are exquisitely good at producing the shape of an answer and almost incapable of withholding one. That asymmetry is not a technical detail. It is a description of what we decided to optimize.