Publication

Exploring human agency in an AI-supported simulation environment for complex systems education

Man Su, Amit Nair, Tomohiro Nagashima

Studies in Higher Education · 2026

@inproceedings{su2026exploring,
  title     = {Exploring human agency in an AI-supported simulation environment for complex systems education},
  author    = {Man Su and Amit Nair and Tomohiro Nagashima},
  booktitle = {Studies in Higher Education},
  year      = {2026},
  doi       = {10.1080/03075079.2026.2733930},
}

Abstract

This study investigates how university students enact learner agency in an AI-supported simulation environment designed to teach complex systems through animal flocking phenomena. Using a 2 × 2 factorial design, we manipulated parameter-control (access to simulation slider change) and questioning AI agent (access to an optional LLM-based conversational agent). Fifty-five students’ learning gains and interaction logs were included in the final analysis. Although all conditions demonstrated significant pre–post improvement, no main or interaction effects of choice configurations emerged after controlling for prior knowledge. Log analyses revealed that learning gains were associated with how choice was enacted rather than whether it was available. Sustained and focused parameter manipulation in the most conceptually complex lesson predicted higher gains, whereas extensive interaction in simpler lessons did not. Engagement with the conversational AI agent varied widely and was not directly associated with learning outcomes. Findings suggest that productive learner agency means more than simply having choices in AI-based digital learning environments, the capacity to make decisions and enact choices when learning becomes more difficult over time need to be supported.