Understanding Mental Models of AI through Player-AI Interaction

Games provide a valuable research setting for studying how people understand AI: The paper argues that AI-based games, particularly through player-AI interaction, offer an ideal domain to observe how mental models of AI develop in real-time, suggesting this interactive context reveals insights that traditional research methods may miss.

Mental models of AI are actively constructed through interaction, not passive: By studying player-AI interaction, researchers can observe that people don't simply accept AI behavior at face value—they actively build and refine their understanding through repeated engagement, which has implications for how AI systems should be designed to support accurate mental models.

Explainable AI benefits from understanding player mental models: The case study demonstrates that insights about how players understand AI can inform the design of more effective explainable AI systems, suggesting that responsible AI design should account for how people actually develop understanding rather than assuming how they will interpret AI behavior.

Current tools and knowledge gaps limit our ability to design human-centered AI: The paper identifies that we have "limited tools to study" how mental models form, highlighting a gap in responsible AI practice—organizations may be deploying AI systems without adequate understanding of how users will interpret and trust them.