Okay — but is the AI overconfident in choices more or less often than a human?