Using Eye Tracking to Detect the Effects of Clutter on Visual Search in Real Time

Abstract

Display clutter causes decrements in visual search performance and can be a threat to safety and efficiency in complex, data-rich domains. Addressing the problem requires a means to detect the presence of clutter in real time, predict its effects, and then trigger countermeasures before breakdowns in information search can occur. Eye tracking is a promising technique for achieving these goals; however, to date, it has been used almost exclusively offline for display evaluation purposes. The goal of this research was instead to develop and evaluate models that combine eye tracking metrics to detect the effects of clutter early on in the search process. Participants were asked to locate targets in a simulated graphics program. Three eye tracking metrics—scanpath length, mean saccade amplitude, and mean fixation duration—were calculated over a 3-second time window. These metrics were then used as input to a set of logistic regression models to predict whether users’ response time will be relatively long or short. The accuracy of the models averaged 75% and the true positive rate was above 90%, with an ability to predict response time as early as 3.6 s into the visual search task. The results of this study confirm that eye tracking metrics can be used to predict the effects of display clutter in real time. They add to the knowledge base in attention and eye tracking, and they ultimately contribute to the design of adaptive displays that lead to improved operator performance.