Round Table: Do Liars Blink Differently? Teaching Computers to Detect Deceptive Eye Behavior
Aaron C. Elkins, Ph.D.
The National Center for Border Security and Immigration
University of Arizona
Dr. Elkins is a candidate for the Si2 Big Data position. Lunch will be provided at 11:30, talk starts at noon.
The eyes have long been considered a window into our inner thoughts and intentions. One long standing behavioral stereotype is that when someone averts their gaze they are being dishonest or are feeling ashamed. While this heuristic has proven unreliable, research has revealed that eye behavior patterns are strongly tied to emotion and cognitive functioning.
The primary limitation to uncovering what these eye behavior patterns actually are, is the enormous task of manually and accurately coding hundreds of hours of eye behavior from video datasets. To address this challenge, this study investigates the development and use of a computer vision-based automated blink detection algorithm applied to a video dataset (N=176) where participants were randomly assigned to lie in a structured interview to questions that varied on cognitive difficulty.
The results of this research are a real-time blink detector and a statistical model for explaining some of the eye-behavior variance exhibited by speakers when lying and honestly responding to interview questions. These findings contribute to the on-going investigations into behavioral sensing technology that will allow computers to interact naturally with users as social actors—enriching human-computer interactions and leading to the development of new automated interviewing and security screening applications.
Aaron C. Elkins is a Postdoctoral Researcher in the MIS department at the University of Arizona. Aaron was previously a Research Fellow at the Intelligent Behaviour Understanding Group at Imperial College London and the National Center for Border Security and Immigration, a Department of Homeland Security Center of Excellence. Aaron conducts laboratory and field experiments that investigate how the voice, face, body, and language, reveal emotion, deception, and cognition for human-computer interaction applications. He also researches how decision makers will actually use and are affected by these artificial intelligence-based technologies.