Confidence Can Be Used to Discriminate Between Accurate and Inaccurate Lie Decisions

Abstract

There is a long-standing belief that confidence is not useful at discriminating between accurate and inaccurate deception decisions. Historically, this position made sense because people showed little ability to discriminate lie-tellers from truth-tellers. But, it is now widely accepted that, under certain conditions, people can discriminate between lie-tellers and truth-tellers. Nevertheless, belief that confidence does not discriminate between accurate and inaccurate responses persists. This belief is somewhat paradoxical because, to the extent that people can discriminate between lie-tellers and truth-tellers, signal detection theory naturally predicts a positive relationship between confidence and accuracy. In line with our signal-detection-based predictions, we show that, among decisions about whether someone is lying, those made with high confidence are more accurate than those made with low confidence. This important relationship has gone unnoticed in past work because of a reliance on inappropriate measures. Past research examining the confidence–accuracy relationship in deception research relied on correlating average confidence with proportion of correctly identified lies. These correlations provide information on whether more confident judges tend to be more accurate but remain silent on the arguably more important question of whether higher confidence decisions are more accurate than lower confidence decisions. We show that confidence–accuracy characteristic analyses are uniquely suited to measuring the confidence–accuracy relationship in deception research.