This dependence can influence our behavior significantly. Typically, we tend to assume others are truthful, a pattern confirmed by this study: even though volunteers knew half the statements were lies, they identified only 19% of them as such. However, this changed dramatically when people opted to utilize AI tools, raising the accusation rate to 58%.
In some respects, this development is beneficial as these tools can help identify more lies in everyday situations, such as combating misinformation on social media. However, there are concerns. It could erode trust, a crucial element of human relationships. If accurate judgments come at the cost of weakening social bonds, is it a trade-off worth making?
Then there’s the issue of accuracy. In their research, von Schenk and her team aimed solely to create a tool more effective than humans in detecting lies—an admittedly low bar. Yet, envisioning such tools routinely assessing the truthfulness of social media posts or scrutinizing job applicants’ resumes raises broader questions. Simply being “better than human” isn’t sufficient if it means increasing accusations.
Would we accept an 80% accuracy rate, where one in five statements might be misinterpreted? Even 99% accuracy leaves room for doubt. The fallibility of traditional lie detection methods, like the polygraph, illustrates this challenge. Designed to detect physiological signs of stress assumed unique to liars, polygraphs are widely discredited and inadmissible in US courts. Despite this, they persist in some contexts, causing harm, particularly in reality TV scenarios.
Imperfect AI tools, scalable and pervasive, pose even greater potential consequences. “Given the rampant spread of fake news and disinformation, there’s utility in these technologies,” says von Schenk. “However, rigorous testing is essential—they must significantly outperform humans.” If AI lie detectors generate excessive accusations, their use may be counterproductive.
AI lie detectors also explore facial movements and microgestures linked to deception, yet their perfect detection remains elusive. Conversely, AI fuels disinformation campaigns globally, underscoring its dual nature. Regulatory frameworks will dictate whether these technologies benefit or harm society, similar to the influence of social media.
