Leveraging Fixation Transition Patterns and Targeted Regions of Interest for Analyzing Code Comprehension

This research-to-practice full paper describes a study on specific eye metrics data and the possible correlations with students' comprehension level. There is a consensus understanding in academia that providing feedback early and often significantly helps students as problems are identified and remedied early. However, as class sizes increase, this becomes more difficult to achieve. Code comprehension techniques can be helpful for educators by identifying students needing assistance and guide them toward academic success and mastery of coding concepts. We hypothesize that by analyzing students' gaze patterns in certain regions of the code, educators can assess and address students' coding challenges. The study distinguishes between frequently and infrequently fixated code regions, using eye metrics to assess students' error perception and challenges. We examined eye-tracking during coding comprehension exercises and proposed a systematic method to identify students needing help based on fixation patterns within Targeted Regions of Interest (TROIs). The methodology uses coding exercises seeded with errors, grades on the tasks, student feedback, measured fixation counts, and average fixation durations of the students' eyes within TROIs of the code compared to the rest code. We conducted preliminary experiments using Java code with 10 students in an introductory programming course. Our initial findings from the collected eye-tracking data showed that the high-scoring students fixated more on the TROIs, frequently backtracked to the method definition, and skimmed the code from top to bottom. The absence of these patterns could be early indicators of code comprehension challenges.