On the Role of Domain Experts in Creating Effective Tutoring Systems

Domain experts should be central to AI education systems, not peripheral: The paper argues that the AI for education community often overlooks the value of highly curated knowledge from domain experts, suggesting that effective tutoring systems require this specialized human input rather than relying solely on automated AI approaches.

Expert knowledge can make AI systems more transparent and educational: By using explainable AI techniques combined with expert-specified rules, tutoring systems can automatically generate lessons that show learners how problems are solved, moving beyond traditional AI debugging to create genuinely educational content.

Expert-designed curricula improve both learning outcomes and efficiency: When domain experts structure the learning pathway for a concept, adaptive tutoring systems can provide better educational experiences while simultaneously using more computationally efficient algorithms—demonstrating that human expertise and AI efficiency aren't competing goals but complementary ones.

Expert knowledge is practical and accessible across domains: The paper's case study on pollinator identification demonstrates that domain expert knowledge can be readily elicited and applied across different fields, suggesting this approach is scalable beyond just one educational context.

This approach addresses responsible AI concerns: By centering human expertise and transparency in educational AI systems, this method helps ensure tutoring systems are interpretable, aligned with pedagogical best practices, and accountable to domain knowledge rather than operating as "black boxes."