Eye Movements as Indicators of Deception: A Machine Learning Approach

2025-05-05

AI Summary

This study investigates ocular behavioral metrics—fixations, saccades, blinks, and pupil diameter—as objective physiological biomarkers for deception detection, specifically to enhance discrimination between “concealment” and “feigning” intentions in the Concealed Information Test (CIT). Leveraging two real-world experimental datasets, it presents the first systematic validation of multimodal temporal oculomotor features for CIT classification. High-precision eye-tracking data were acquired using EyeLink 1000 and Pupil Neon systems; saccade count, duration, amplitude, and peak pupil size emerged as the most discriminative features. An XGBoost classifier achieved 74% accuracy in binary classification (revealing vs. concealing) and 49% in ternary classification (truthful vs. concealing vs. feigning). Results demonstrate that oculomotor features significantly augment conventional polygraph-based paradigms, offering a novel, non-invasive, AI-enhanced approach to deception detection.

Technology Category

• Computer Vision: Multi-modal Vision
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Humans and AI: Brain-Sensing and Analysis

Application Category

• Security and Privacy: Large-scale security measurements
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Gaze may enhance the robustness of lie detectors but remains under-studied. This study evaluated the efficacy of AI models (using fixations, saccades, blinks, and pupil size) for detecting deception in Concealed Information Tests across two datasets. The first, collected with Eyelink 1000, contains gaze data from a computerized experiment where 87 participants revealed, concealed, or faked the value of a previously selected card. The second, collected with Pupil Neon, involved 36 participants performing a similar task but facing an experimenter. XGBoost achieved accuracies up to 74% in a binary classification task (Revealing vs. Concealing) and 49% in a more challenging three-classification task (Revealing vs. Concealing vs. Faking). Feature analysis identified saccade number, duration, amplitude, and maximum pupil size as the most important for deception prediction. These results demonstrate the feasibility of using gaze and AI to enhance lie detectors and encourage future research that may improve on this.

Problem

• Evaluating AI models for deception detection using eye movements
• Assessing gaze data in Concealed Information Tests across datasets
• Identifying key eye movement features for lie prediction

Innovation

• AI models analyze gaze data for deception detection
• XGBoost achieves 74% accuracy in binary classification
• Saccade and pupil features key for lie prediction