The lie in the eyes: Ocular indicators and machine‐learning models with a cognitive load paradigm to detect lying

Abstract Purpose Behavioural and ocular markers may provide reliable indicators of lying, although their combined use remains underexplored. This study examined whether eye‐tracking could contribute to the detection of lying using an Induced fake identity (IFI) paradigm, considering the association between pupillary dynamics, eye‐blinks and behavioural indices of lying. Machine learning was also used to assess the predictive value of these measures. Methods Fifty‐seven healthy adults underwent an IFI task, randomised to lying or truth‐telling conditions. Participants responded to Control, Expected and Unexpected questions while assuming either a fake identity or their true identity, while their ocular (pupil dilation and eye‐blinks) and behavioural (reaction times and errors) responses were recorded. Results Liars exhibited significantly longer reaction times than truth‐tellers, particularly in response to Unexpected questions, consistent with greater cognitive demands during deception. Similarly, baseline‐corrected pupil responses also differed between groups: whereas truth‐tellers showed reduced pupil dilation during Unexpected questions compared with Control and Expected questions, liars exhibited relatively stable pupil responses across question types. No group differences emerged in eye‐blink rate, nor in errors. Machine‐learning models achieved up to 84% accuracy and identified reaction times to Unexpected questions as the strongest predictor of deception, whereas ocular measures provided limited additional predictive value. Conclusions Lying is characterised by slower responses to identity‐related Unexpected questions and distinct patterns of pupillary modulation. Although reaction times were the strongest predictor of deception, pupillary responses may provide complementary information about the psychophysiological processes associated with deceptive responding.