AI-Induced Skepticism in Cybersecurity: An Empirical Study on the Emergence of the Liar’s Dividend in Digital Contexts

Sammanfattning

The rise of Generative Artificial Intelligence has fundamentally changed the landscape of digital deception. AI-generated phishing emails, deepfakes and synthetic voice messages are becoming increasingly difficult to distinguish from authentic communications. In response, considerable effort has gone into raising public awareness about these threats. However, this study questions whether such awareness is always protective or whether it can sometimes produce the opposite effect. Drawing on the concept of the Liar's Dividend introduced by Chesney and Citron (2019), this study empirically investigates whether Awareness of AI-Generated Deception fosters skepticism toward authentic cybersecurity communications and whether that skepticism manifests as a Liar's Dividend effect among individuals at a societal level. The central concern is that individuals who are highly aware of AI-generated threats may begin dismissing genuine security emails and verification requests, assuming they could be artificially generated. This study terms this phenomenon the Security Skeptic's Dividend. A quantitative cross-sectional survey was conducted with 133 respondents using authentic security communications as vignette stimuli. Five hypotheses were tested using Spearman correlation, linear regression, binary logistic regression, Mann-Whitney U tests and ordinal logistic regression. The findings confirmed that Awareness of AI-Generated Deception is positively associated with skepticism toward digital communications and that this skepticism significantly predicts the dismissal of authentic security alerts. Higher skepticism also predicted verification-seeking behaviour before compliance, a behavioural pathway not previously identified within the Liar's Dividend framework. Moral neutralisation was not supported as a driver of non-compliance, suggesting the effect operates more automatically than deliberately. Situational distraction did not amplify the dismissal effect as hypothesised and a consistent directional reversal was observed with more skeptical users displaying greater caution under distraction. These findings provide the first empirical evidence that the Liar's Dividend operates as a measurable behavioural phenomenon in a cybersecurity context, with practical implications for the design of public cybersecurity awareness programmes and security communication systems.