AI-Assisted Ideation in Mission Critical Engineering: A Pilot Study

This pilot study examines how access to a large language model (LLM) influences professional engineers’ ideation on a mission-critical aerospace design task. Participants generated concepts for landing a fixed-wing aircraft on Mars, either with or without LLM support. Outcomes were assessed using idea counts, NASA SME ratings of desirability, viability, and feasibility (DVF), design-space coverage, process timeline coding, and post-session surveys. Across 8 participants and 66 ideas, AI-assisted ideas scored higher on viability, feasibility, and composite DVF at the idea level, while the manual group produced more ideas and broader category coverage. Process coding showed that AI-assisted participants shifted time away from ideating and toward prompting and reviewing. Perception differences were limited, though qualitative responses suggested text-modality friction and lower process enjoyment in the AI assisted condition. Given the small sample and nested data structure, findings are preliminary and motivate a larger follow-on study