Children's Mental Models of AI Reasoning: Implications for AI Literacy Education

Children's understanding of AI reasoning varies significantly by age: Younger children (grades 3-5) tend to view AI as possessing inherent intelligence, while older children (grades 6-8) are more likely to understand AI as a pattern recognizer. This suggests that AI literacy curricula should be developmentally tailored rather than one-size-fits-all.

Three distinct mental models emerged from the research: Children conceptualize AI reasoning through Deductive, Inductive, and Inherent models. The prevalence of the "Inherent" model among younger students indicates they may not yet grasp that AI reasoning is fundamentally algorithmic and learned from data, which has implications for how we teach AI concepts.

Current AI education may not adequately address reasoning processes: Although reasoning algorithms are identified as one of the "Five Big Ideas" in AI education, the study found that children lack robust mental models in this area. This gap suggests that existing curricula may need strengthening in explaining how AI actually makes decisions.

Explainable AI tools need to account for children's misconceptions: The identified tensions in children's understanding point to the need for better-designed explainable AI systems that can help children move from misconceptions (like inherent intelligence) toward accurate mental models of algorithmic reasoning.

Age-appropriate scaffolding is essential for AI literacy: The research demonstrates that children's reasoning about AI develops with age, suggesting educators should use scaffolding techniques that progressively build from simpler concepts toward more sophisticated understanding of pattern recognition and algorithmic decision-making.