Location
Hilton Hawaiian Village, Honolulu, Hawaii
Event Website
https://hicss.hawaii.edu/
Start Date
3-1-2024 12:00 AM
End Date
6-1-2024 12:00 AM
Description
With the widespread adoption of generative large language models (GLMs) such as GPT-3 or ChatGPT for human-AI problem solving, understanding the effect on performance becomes important. Brainstorming is an established approach for generating ideas to solve problems. In this study, we investigate how AI ideas affect the brainstorming performance metric ‘flexibility’, which refers to the breadth of exploration or coverage of the topic. The foundation for our analysis is the data from an experiment (n=52) in which individual participants brainstormed in two conditions: (1) human-only (baseline) and (2) human+AI (treatment). The treatment condition had access to ideas generated via the GLM OpenAI GPT-3.5. Results show significantly higher flexibility for the human+AI as compared to the human-only condition with a large effect size. With our study, we contribute to the literature of electronic brainstorming, brainstorming with GLMs, as well as to the research challenge of human-AI collaboration.
Recommended Citation
Memmert, Lucas and Bittner, Eva, "Human-AI Collaboration for Brainstorming: Effect of the Presence of AI Ideas on Breadth of Exploration" (2024). Hawaii International Conference on System Sciences 2024 (HICSS-57). 6.
https://aisel.aisnet.org/hicss-57/cl/machines_as_teammates/6
Human-AI Collaboration for Brainstorming: Effect of the Presence of AI Ideas on Breadth of Exploration
Hilton Hawaiian Village, Honolulu, Hawaii
With the widespread adoption of generative large language models (GLMs) such as GPT-3 or ChatGPT for human-AI problem solving, understanding the effect on performance becomes important. Brainstorming is an established approach for generating ideas to solve problems. In this study, we investigate how AI ideas affect the brainstorming performance metric ‘flexibility’, which refers to the breadth of exploration or coverage of the topic. The foundation for our analysis is the data from an experiment (n=52) in which individual participants brainstormed in two conditions: (1) human-only (baseline) and (2) human+AI (treatment). The treatment condition had access to ideas generated via the GLM OpenAI GPT-3.5. Results show significantly higher flexibility for the human+AI as compared to the human-only condition with a large effect size. With our study, we contribute to the literature of electronic brainstorming, brainstorming with GLMs, as well as to the research challenge of human-AI collaboration.
https://aisel.aisnet.org/hicss-57/cl/machines_as_teammates/6