Work-in-Progress: Students' Prompting Strategies When Solving an Engineering Design Task

This work-in-progress research paper investigates how students prompt generative AI while tackling an engineering design challenge. As tools like ChatGPT become common in education, we must understand how to incorporate them into our teaching and guide students on their proper use. Although the existing literature focuses on understanding what students use generative AI tools for, less work has been done to examine students' prompting strategies when engaging with these systems. We leveraged a new brainstorming assignment in a first-year engineering course at a large Midwest public university, where students ideated collaboratively using ChatGPT for the design of their term project - a semiautonomous robot (n = 97 teams, 589 prompts). Most prompts (~50%) ranged between 55 and 95 characters, or approximately 8 to 19 words. Our initial qualitative findings suggest that their approaches center on seeking information, much like how they would use a search engine. Others directly ask the chatbot to provide alternatives for their design without providing the appropriate criteria or constraints. A subset of prompts had an evaluative component, asking ChatGPT to weigh ideas against one another. We found that 11 % of prompts included instructing ChatGPT to produce the “best” solution instead of generating multiple ideas, suggesting students focus on using the chatbot in a more convergent design process - unlike divergent thinking in brainstorming.