2026 ASEE Annual Conference & Exposition

The integration of generative artificial intelligence (AI) tools into undergraduate electrical and computer engineering (ECE) education is fundamentally reshaping how students solve problems, produce artifacts, and demonstrate learning. Generative systems capable of producing code, analytical solutions, and technical explanations in real time are now routinely embedded in student workflows. While these tools expand access to academic support and improve task completion efficiency, they introduce a critical challenge: correct outputs can no longer be assumed to reflect underlying understanding. Prior research in engineering and higher education reveals that AI-assisted environments can increase performance while simultaneously obscuring the reasoning processes that traditionally serve as evidence of competence [1], [2].

This study presents a multi-year, practice-based case study of AI integration within an ABET-accredited ECE program, incorporating survey-informed exploratory data from undergraduate ECE students across multiple course contexts collected through April 2026 (N = 180). Results show that over 90% of students report frequent use of AI for programming, debugging, and analytical tasks, yet reported confidence in independently explaining and adapting AI-assisted solutions remains substantially lower than reported task completion success. These patterns suggest a persistent misalignment between artifact production and demonstrable competence, defined here as the ability to explain, adapt, verify, and justify solutions independently of AI assistance, consistent with prior work on automation bias and cognitive offloading in AI-assisted systems [3].

We argue that these findings reflect a deeper structural issue in engineering education, the artifact–competence divergence problem, in which students can generate correct solutions without demonstrating the capacity to evaluate, verify, or justify those solutions under changing conditions. To address this challenge, the paper introduces and utilizes an assessment-centered framework grounded in engineering adequacy judgment, which characterizes competence as the ability to determine when available evidence is sufficient to justify action under uncertainty. This framing aligns with established perspectives in engineering practice that emphasize verification, reasoning, and responsible decision-making under incomplete information [4], [5].

Further, the findings demonstrate that assessment design, not the presence of AI, is the decisive factor in determining whether AI functions as a learning support or a substitute for understanding. In artifact-based assessment environments, AI enables correct outputs without requiring demonstrable competence; in explanation- and verification-centered environments, AI use is constrained by the need to justify reasoning. This result reframes AI integration as an evidentiary validity problem, shifting the central educational question from whether students can produce correct answers to whether those answers can still serve as credible evidence of competence.