Development and Evaluation of a Method Employed to Identify Internal State Utilizing Eye Movement Data

In the attempt to recognize and estimate human internal states, such as varying emotions, psychological and conceptual conflicts pose interesting and challenging issues. In this paper, we explore a pattern recognition technique that can detect a state of confusion and can estimate human interest, each an internal state of mind. In order to automatically detect a state of confusion from the objective data made available to us, the technique we present relies upon eye movement data. We have conducted three experiments in which subjects are confronted with a task that includes a trap intentionally designed to confuse them. We have recorded their eye movement data. We demonstrate that approximately 89% of a state of confusion can be detected from eye movement data by using a backpropagation algorithm. Moreover, for estimating human interest, we present a technique that builds upon the foundation of our confusion detection technique. As a result, we can demonstrate that approximately 60% of human interest can also be estimated through eye movement data.