Ulric B. and Evelyn L. Bray Social Sciences Seminar

Abstract: This paper investigates mistakes and overconfidence in detecting lies in the context of news events, without the motivation to believe them. Participants in our experiments fail to detect lies by believing videos in which senders lie and not believing videos in which senders are truthful, thereby making both type I and II errors. We use a facial-expression-recognition software to measure emotions and facial movements of senders during their videos, and collect other measures of their speech (e.g., word count). A number of emotions, facial movements, and speech characteristics of senders affected receivers' beliefs. However, the effects were often in the wrong direction or concentrated on indicators that did not predict a video's truthfulness. Yet, participants are overconfident about their ability—both in absolute and relative terms. Connecting these findings with the literature on fake news, we show that participants are more likely to share videos in which senders lie but fail to be skeptical when learning that videos have been shared frequently. These results suggest that, even without motivation to be biased, individuals are unaware of their limited ability to detect lies in the context of news events, and lies may spread, wrongly shaping other individuals' beliefs.
Written with Uri Gneezy.