2026 ASEE Annual Conference & Exposition

Educators are at a crossroads with AI. Possible paths forward include active adoption, attempts to prohibit AI in the classroom, or anything in between. While it is important to consider the negative effects of easy access to generative text-based AI on academic integrity and learning, AI is here to stay. As such, it is paramount for engineering educators to thoughtfully explore how AI technology, particularly off-the shelf tools already being adopted by students, can advance domain-specific engineering education. With limited experimental evidence available on measured AI benefits vs challenges, current best practices are scarce. The lack of data is particularly acute for specific engineering tasks, since most publications to date center on generalized tasks such as writing and coding. For this study, broad research questions are used: “how and to what extent does embedding text-based AI into structural design assignments, do students find success or struggles, and what impact does AI have on their resulting performance?”
In response, this research explored leveraging text-based AI in two structural design courses. The study focused on structural steel design due to its naturally varying levels of design complexity, as well as current efforts in industry to supplement this process with AI. Activities and demos were developed for four design situations of increasing open-endedness, in which AI could inform design activities, clarify processes, and promote understanding. The study compares two separate cohorts at our institution: Cohort 1 includes future structural engineers, while Cohort 2 includes students pursuing other architectural engineering sub-disciplines (mechanical, electrical, and construction). Data was collected directly on student AI use and performance using rubrics. A pre-post survey was also conducted with questions on student perceptions on various AI attributes.
The results indicate that student perceptions and performance are mixed: some recognized how they must actively curate input prompts and carefully check outputs for correctness, while others overly relied on AI to provide answers with limited results checking or demonstration of underlying conceptual understanding. Enthusiasm towards learning with AI was equally mixed. Cohort 1 students were more engaged on average than Cohort 2 in the topic. In aggregate, both cohorts trusted AI results across the four design situations beyond what would be expected of a professional engineer, as demonstrated by Pitts et al. [1] and Amoozadeh [2]. Cohort 1 frequently mentioned that it would have been faster to do the activities strictly by hand over leveraging AI, an observation that may change over time with increased familiarity, multimodality, and accuracy for next-generation AI tools. That said, Cohort 2 often performed numerically better on many problems.