2026 ASEE Annual Conference & Exposition

This Work-in-Progress paper presents a three-phase longitudinal study on the impact of Large Language Models (LLMs) on the professional development and learning of chemical engineering students. This paper evaluates the integration of ChatGPT/Copilot into three assignments across a cohort of senior chemical engineering students (n ≈ 65) enrolled in Chemical Engineering Lab I, Process Safety and Ethics, and Chemical Engineering Lab II at a public R1 university. Assignments in Fall 2025 positioned ChatGPT as (1) a teamwork coaching tool to help student teams convert vaguely defined collaboration challenges into actionable improvement plans and (2) a domain scaffold supporting process safety tasks including HAZOP development, identification of relevant engineering standards, construction of a LOPA, and preparation of a neutral “risk gap” memo addressing tensions between budget constraints and safety recommendations. Students were explicitly reminded that AI output may be incomplete or incorrect and that engineering judgment remained their responsibility. A qualitative synthesis of reflections and submitted work identified themes related to student adoption of AI suggested practices, strategies for constraining prompts and outputs, evidence of safety and standards literacy within ChatGPT responses, and approaches to ethical reasoning under realistic professional pressures. A follow up assignment in Spring 2026 (Chemical Engineering Lab II) analyzed AI usage patterns after these instructor-led experiences. Data from 17 student teams indicate that students transitioned to an independent "human-in-the-loop" model. Results highlight a significant increase in AI literacy: Students shifted from using AI for "answers" to using it for "overhead" tasks (i.e. concept clarification) while acknowledging skepticism towards AI responses. This work-in-progress study suggests that LLMs are most effective in engineering education when students are tasked with assessing the performance of the LLM. Instructors should design assignments that push AI to fail (i.e. complex calculations) while probing students to assess the strengths and weaknesses of LLMs to enhance learning outcomes.