This study examines whether a pedagogy-guided, artificial intelligence (AI)-assisted learning environment can improve learning performance and student perceptions in undergraduate engineering education. Rather than positioning generative AI as an answer-generating tool, the study frames it as a scaffold embedded within a structured teaching process.
A quasi-experimental design was implemented in an undergraduate engineering course in Taiwan. Eighty-two students participated across an AI-assisted group and a comparison group receiving conventional instruction. The intervention followed four recurring stages: problem introduction, AI-supported exploration, critical reflection and discussion with feedback. Quantitative evidence was drawn from five rubric-assessed problem-solving assignments and a post-course questionnaire administered to the AI-assisted cohort, while qualitative evidence came from students' reflective writing. Because the questionnaire was administered only to the AI-assisted cohort, its results are reported descriptively rather than as between-group effects.
The AI-assisted cohort showed higher composite assignment performance and reported strong engagement, perceived usefulness and problem-solving confidence. Qualitative findings showed that structured reflection helped students question AI-generated responses, identify limitations and refine their own reasoning rather than accept outputs uncritically.
As the comparison group came from a previous semester, assignments were scored by the course instructor and prior AI exposure was not measured systematically, the findings should be interpreted as promising exploratory evidence rather than definitive causal proof.
The findings suggest that the educational value of generative AI depends on pedagogical orchestration, guided prompting and reflection rather than simple access to the tool.
The paper offers a classroom-tested model for integrating generative AI into higher engineering education in a pedagogically meaningful and ethically responsible way.
