An Agent-Based Model of GenAI in Creative Writing: Novelty Gains and Diversity Declines
Generative artificial intelligence (GenAI) is being rapidly adopted in education. Its effects on creativity unfold at both individual and collective levels. We develop an agent-based model that represents students’ creative outputs in a two-dimensional idea space to examine how GenAI assistance in creative writing shapes individual creativity and classroom-level idea diversity. Simulations show that GenAI increases individual creativity, with disproportionately larger gains for less creative students, while reducing collective diversity through idea homogenization. Critically, the model reveals a more nuanced, non-linear pattern of collective diversity than empirical accounts of homogenization alone: collective diversity can increase at low-to-moderate adoption rates before reaching a tipping point and declining at high adoption rates. These findings highlight the need to evaluate GenAI in education using both individual performance metrics and collective system outcomes.
Authors
- Shashank Mehrotra (ORCID: https://orcid.org/0000-0002-6749-3773)
- Xiaoxuan Cheng (ORCID: https://orcid.org/0000-0001-8892-3683)
- John D Lee
Institutions
- University of Wisconsin–Madison (US)
- Honda Research Institute, USA (US)
Publication Details
- Journal
- Proceedings of the Human Factors and Ergonomics Society Annual Meeting
- Published
- 2026-11-12
- DOI
- https://doi.org/10.1177/10711813261498805
- Primary Topic
- Creativity in Education and Neuroscience
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- Honda Research Institute, USA