Eye trackers are an important tool in cognitive science research. Advances in eye-tracking technology have reduced the invasiveness of early equipment, along with the cost of acquiring devices for research. As such, eye-tracking is becoming an essential part of research on software engineering and program comprehension. However, eye-tracking devices create challenges for researchers. Eye-tracking devices have inherent margins of error in sampling introduced by various factors, which cannot be controlled and can introduce errors in study data. As such, despite best efforts, eye-tracking data always needs some form of correction to identify and remove invalid data or adjust it as samples drift or become displaced over time due to participant movement, fatigue, or other factors. The processes used to correct eye-tracking data are time-consuming and require trial and error to correct the data reasonably while mitigating the impact of the corrections.

Additionally, eye-tracking infrastructure to support program comprehension studies with realistic stimuli in natural software development environments is limited. This work addresses these limitations. Focusing on program comprehension-based studies using eye-tracking, this work presents SrcGaze, an algorithm that automatically corrects fixation gaze event data using source code stimuli. Compared with state-of-the-art fixation correction methods, SrcGaze achieves 73% agreement with manually corrected fixation data. The algorithm also has a linear runtime, making it the most computationally efficient approach; it can correct over 44,000 fixations in only 6 seconds.

This work also contributes to iTrace, an infrastructure to support eye-tracking studies using source code. This work includes a redesign of the infrastructure, the creation of plugins to support multiple development environments and tools, implementation of eye-tracking data analysis and storage tools, and Deja Vu, a novel approach to supporting high-speed eye-tracking devices sampling at speeds >60 Hz while also supporting recording and playback of participant studies for unlimited research opportunities post-data collection.

Join us on Thursday, Nov. 19, for this exciting presentation from Drew Guarnera, assistant professor of Computer Science at the College of Wooster. The presentation will begin in Tomsich 101 at 11:10 a.m. (Please note the date, time and location change). We hope to see you there!