Best practices for RL in education: evaluate live with actual users at scale, look beyond STEM, and target under-served learner groups including children¶

Strategy · All strategies
Evidence · 1 claim (1 for) · 1 study (1 review), q2 · 0 of 1 report an effect size · 1 claim rests on one study

Description¶

Drawing on the systematic review, the authors distill best practices for implementing RL in education. Central recommendations are the "need for broader and more large-scale deployments and evaluations with actual users relative to only using simulated data", extending research beyond the STEM umbrella to test transferability of concepts, and expanding applications targeting children to support their overall learning process. They also caution that education poses unique challenges — limited interaction log data, partially unobservable psychological processes, multiple optimization goals, immediate reaction requirements, and interpretability of RL actions.

Design Implications¶

Context¶

Requirements¶

• Evaluations should move beyond simulated users toward live interaction with actual learners
• Applications must address RL's substantial data requirements, e.g., environments that can record learner data in large quantities

Constraints¶

• The review notes methodological issues and the lack of simulations that adequately reflect the respective learning context as current limitations

Target Learners¶

• adult and higher-education learners (current majority)
• children and school learners (recommended expansion)

Target Learning Goals¶

• enhancing learning outcomes and student engagement through adaptive, personalized learning environments

Claims¶

• RL-based educational applications concentrate on STEM subjects (n = 39 of 62 papers specifying a subject) and on college/higher-education learners (n = 30), with 36 papers not specifying a target group [+W]

Related Strategies¶

• Future improvements: real-world user studies, advanced student modeling, and spaced repetition

Examples¶

-

Key Sources¶

• Anna Riedmann, Philipp Schaper, Birgit Lugrin. (2025). Reinforcement Learning in Education: A Systematic Literature Review. International Journal of Artificial Intelligence in Education. https://doi.org/10.1007/s40593-025-00494-6