LieVis: A Visual Interactive Dashboard for Lie Detection Using Machine Learning and Deep Learning Techniques

Lying poses a significant concern in our everyday lives, impacting social interactions. The skill of detecting lies or deceitful statements is of immense value, mainly due to the elusive nature of the underlying patterns associated with lying. Numerous studies have extensively explored lie detection, employing various advanced techniques. Researchers have devoted their attention to devising more effective and efficient solutions for lie detection. However, there remain ongoing challenges, particularly in achieving higher accuracy levels. There has been relatively limited focus on developing visually interpretable and interactive systems for lie detection. Therefore, we leveraged artificial intelligence (AI), incorporating machine learning and deep learning techniques to augment the existing system. Our experiment shows that deep learning models, specifically BERT, achieve the highest accuracy, whereas machine learning models such as Random Forest (RF), and K-Nearest Neighbors (KNN) exhibit superior accuracy across both sets of features. To enhance result exploration, we introduced a dedicated visual interactive system/dashboard called LieVis, meticulously designed to facilitate comprehensive exploration of the obtained results. The evaluation of LieVis showcases its superiority over existing systems. The source code and the live dashboard with result visualization are available on our GitHub repository
¹
.