Ph.D. Dissertation Defense: Kang-il Park
Monday, June 22, 2026
12:00 PM
AVH 103C
Zoom: https://unl.zoom.us/j/98239267386
"Assessing Program Comprehension and Emotional State Using Biometrics"
Developers work with many different artifacts and environments, including coding standards, how code is presented in an IDE, and the ways they communicate with one another. Understanding how these aspects affect program comprehension can provide insights into developers’ cognitive processes and mental models, and ultimately into how their productivity might be influenced. This dissertation presents three empirical studies that assess comprehension and emotional state using biometric devices while developers perform software engineering tasks.
The first study leverages eye tracking technology to replicate a prior questionnaire-based online study. This replication was used to determine the visual effort needed to read code presented in different readability rules, namely minimize nesting and avoid do-while loops. Each rule is evaluated on code snippets that are correct and incorrect with respect to a requirement. The gaze data show strong support for the minimize nesting rule, while the avoid do-while rule was less important.
The second study investigates whether the use of scope highlighting during code comprehension tasks affects how users focus their gaze and whether it affects the accuracy and speed of their responses. We found no differences between the three code presentation modes in terms of correctness or time taken to complete tasks. We did observe a difference in eye movements between the Java and Stride (frame-based) languages, with Stride showing higher cognitive load. However, there was still no difference in task performance.
The third study examines whether prompting developers to periodically self-report their emotional state during bug-fixing tasks affects performance and inferred emotional responses, using eye tracking, electrodermal activity, and facial expression data. The results do not conclusively show that self-reflection on emotions affects performance on programming tasks. However, we identified potential in using a combination of electrodermal activity and facial expression data as predictors for classifying current emotional state during programming tasks.
The results on the tasks studied suggest that developers minimize nesting and limit the use of frame-based programming languages, unless well-trained, to reduce cognitive load during program comprehension. The findings also indicate that combining electrodermal activity and facial expression data is a promising approach for monitoring developers' emotional states.
Committee:
Dr. Bonita Sharif, Advisor
Dr. Robert Dyer
Dr. Witawas Srisa-an
Dr. Michael Dodd