Training AI For When Humans Will Use It
· PDF version dated 12 August 2026; the author’s research index labels the draft July 2026.
Summary
Bryan and Gans model how AI should be trained when people decide whether to use and verify its predictions. The preferred balance between accuracy and coverage depends on verification costs, the consequences of errors, and other available information. Changes in human decision strategies can change which model is best.
Why it matters
The economic value of an AI model depends on the whole human-AI workflow, including review and error costs. A model that answers more questions can be preferable when checking is cheap, while selective prediction can be preferable when checking is expensive.
What to keep in mind
This is a theoretical decision model, not a deployment experiment. Its conclusions depend on assumptions about information, verification and user behavior; it does not establish one optimal training strategy for all real-world uses.