Who Tells the Better Math Story? Exploring Gen AI in a Digital Algebra Game

This study examines how primary school students aged 9 to 11 evaluate AI-generated versus self-generated stories within a digital algebra game. Using an embedded GPT-4 system called Algy, the research asked 136 students across two schools to compare an AI-produced narrative against their own, with 58 students preferring the AI version and 64 favoring their own. The work situates itself within game-based learning and generative AI for reflective education, asking how bounded AI use can support learning while preserving student agency.

Students who chose AI-generated stories typically valued perceived competence, structural clarity, and reliability, while those who preferred their own stories emphasized ownership, emotional engagement, and cognitive ease. Students viewed the AI positively as a helper and teacher, yet they also demonstrated critical awareness of its shortcomings, including confusing content and cultural mismatches. Teachers similarly found the structured integration of generative AI more controllable and trustworthy than unstructured use.

A notable finding is the tendency to treat AI as the "safer" choice under uncertainty, which the authors interpret as a possible sign of automation bias. This interpretation rests on self-reported interview and survey data from two schools, so it should be regarded as suggestive rather than conclusive. The authors nonetheless conclude that carefully scaffolded AI can prompt reflection while maintaining student ownership of mathematical narratives, though the study's limited scope constrains broader generalization across contexts and age groups.