In most software engineering applications, the human is in the loop, and perfect outputs from AI are not strictly necessary for the latter to be useful. All current scoring methods—whether functional ...