This paper investigates how generative AI tools contribute to Comprehension Debt in the context of an undergraduate software engineering project. The author’s starting point is that such tools have rapidly become integral to software development and are especially attractive to students because they can reduce cognitive load, while introducing a socio-cognitive risk in exchange.
The study is qualitative and is based on 621 reflective diaries written by 207 students over eight weeks. From that material it identifies four patterns by which Comprehension Debt accumulates in students’ use of generative AI tools — AI-as-black-box code acceptance, context-mismatch debt, dependency-induced atrophy, and verification-bypass — alongside one mitigating pattern, in which students use the tools as a comprehension scaffold and build a deeper understanding of the code through them.
The paper’s central conceptual argument is that Comprehension Debt is distinct from traditional technical debt, because it resides in the collective cognition of development teams rather than in the codebase itself. From that the author draws a pedagogical conclusion: software engineering education needs explicit strategies to mitigate it, with emphasis on verification practices, structured retrospectives, and active learning assessments.
Key Points
- Defines Comprehension Debt as the growing gap between what a development team knows about its codebase and what it actually needs to understand in order to maintain and modify it effectively
- Identifies four accumulation patterns: AI-as-black-box code acceptance, context-mismatch debt, dependency-induced atrophy, and verification-bypass
- Identifies one mitigating pattern, in which generative AI is used as a comprehension scaffold
- Argues Comprehension Debt is distinct from technical debt because it resides in a team’s collective cognition rather than in the codebase
- Recommends verification practices, structured retrospectives, and active learning assessments as pedagogical mitigations
Notes
The evidence base is a single undergraduate software engineering course: 621 diaries from 207 students over eight weeks. The patterns are therefore drawn from a student population working on a course project, and the paper frames its recommendations for software engineering education rather than for professional teams.
The paper is five pages and carries a journal reference to the Proceedings of the 2026 International Conference on Evaluation and Assessment in Software Engineering Companion.