How Do Users Organize Prompting Workflows in Human–AI Collaboration? A Process Mining Analysis of AI Literacy-Based Strategy Patterns
Growing use of large language models (LLMs) in creative and knowledge work has increased interest in how users collaborate with AI beyond output quality. This study examines how AI literacy and task type relate to prompting strategy organization, workflow, and perceptions of collaboration in ChatGPT-supported writing. Forty university students completed a divergent brand-naming and a convergent writing task. Prompting logs were analyzed using sequential pattern metrics and process mining, complemented by interviews. Higher AI literacy was associated with more structured coordination among prompting strategies, whereas lower literacy showed more repetitive, generation-centered pathways. These differences appeared in specific transition and loop configurations rather than overall transition diversity. Participants described AI as assistive, yet behavioral traces showed substantial AI contribution to outputs. These findings provide process-oriented insights into how AI literacy and task type relate to strategic adaptability in human–AI collaboration.
Authors
- Gyewon Jeon (ORCID: https://orcid.org/0000-0002-4794-0046)
- Sangwon Lee (ORCID: https://orcid.org/0000-0002-9825-0854)
- Dongmin Shin (ORCID: https://orcid.org/0000-0003-1785-4603)
- Younghyun Kim (ORCID: https://orcid.org/0009-0008-0608-7887)
Institutions
- Korea University (KR)
- Hanyang University (KR)
Publication Details
- Journal
- International Journal of Human-Computer Interaction
- Published
- 2026-11-17
- DOI
- https://doi.org/10.1080/10447318.2026.2758472
- Primary Topic
- Usability and User Interface Design
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- National Research Foundation of Korea
- Ministry of Science and ICT, South Korea
- Institute for Information and Communications Technology Promotion