2026 ASEE Annual Conference & Exposition

Generative Artificial Intelligence (GAI) is reshaping professional education, yet it poses significant challenges for academic implementation. These include fear of academic dishonesty, misconceptions about its functioning, and ethical risks arising from uncritical reliance on automated outputs. In Engineering Education, there has been growing recognition of GAI as a co-pilot, with its use through “prompt engineering”, a process analogous to the engineering design cycle. In contrast, in other STEM areas, such as foundational Medical Education, where ethical and critical thinking are also fundamental learning objectives, traditional pedagogical paradigms often lack the analytical frameworks to evaluate these pedagogical interventions. In addition, given the increasing use of GAI in STEM, it remains unclear whether the mechanisms through which students-GAI interactions develop technical agency and ethical judgment, or whether these can spontaneously emerge in non-technical cohorts through unguided “productive failure”. This study addresses this gap by exploring how unguided interaction with GAI triggers iterative design behaviors and epistemic validation strategies in STEM medical students. Using the Productive Failure Engineering Education framework as a theoretical lens, we examined the procedural experiences of five medical students (semesters I–VI) tasked with generating visual conceptual representations using GAI without prior technical training. Using an exploratory qualitative thematic analysis, we identified how the initial lack of guidance forced students to confront the algorithm’s hidden nature. Participants initially treated GAI as an authoritative source, which led to failure and annoyance, inducing them to try again. This Productive Failure catalyzed a shift in cognitive strategy: students spontaneously adopted iterative debugging cycles, refined constraints (prompt specificity), and developed verification protocols to check GAI outputs against domain truth. The results suggest that unguided GAI interaction fosters a “designer mindset” in medical students, mirroring iterative behaviors typical of engineering undergraduates. Our study contributes to Engineering Education by demonstrating how multidisciplinary frameworks can support AI literacy and critical thinking across professional fields. Lastly, we provide insight into how engineering constructs such as Productive Failure and Iteration operate outside formal engineering curricula, revealing which elements of technical agency in STEM undergraduate students emerge spontaneously.