AI-assisted development can leave working, passing code that developers no longer understand. The review collects practitioner-reported causes and mitigation practices to detect and reduce that gap.
Key takeaways · AI-distilled
The study is a grey-literature review of 76 practitioner sources such as blog posts and online discussions, chosen because practitioners were writing about cognitive debt before academic studies could measure it.
The authors find cognitive debt can build up unnoticed behind working code and passing tests, especially when developers accept AI-generated solutions without building their own understanding of them.
Practitioners report harder maintenance and debugging, knowledge fragmented across the team, and reduced productivity as the consequences, according to the abstract.
Proposed countermeasures include documenting design rationale, keeping changes review-sized, comprehension gates and knowledge-sharing rituals, with reliability metrics suggested as a signal for detecting the debt.
articleThe Psychological Costs of Artificial Intelligence Adoption in Software EngineeringAdam Alami, Elda Paja, Abhishek Tiwari
articleBuilding AI-Intensive Software with AI: Early Results and a Cautionary Tale on Measuring Development CostVictor Barros de Miranda Neves, Kiev Santos da Gama, Vinicius Cardoso Garcia
articleHelpful but Fallible: Developer Experiences of AI Tools Under a Coordinated Industrial Roll-outAndreas Bexell, Rushali Gupta, Lo Gullstrand Heander, Emma S\"oderberg, Per Runeson, Sigrid Eldh, Wei Shi, Konstantin Malysh