Reinforcement learning-based adaptive learning: Rewards improvement considering learning duration.

Area of Science:

• Artificial Intelligence
• Educational Technology
• Machine Learning

Background:

• Adaptive learning systems use reinforcement learning (RL) to personalize educational content.
• Current RL systems primarily focus on learning effectiveness, potentially increasing study time.
• Learner prior knowledge significantly impacts the time needed to master new material.

Purpose of the Study:

• To develop an RL-based adaptive learning system that optimizes both learning effectiveness and time efficiency.
• To incorporate learner prior knowledge into the RL reward function.
• To investigate the impact of an improved reward mechanism on learning duration and efficiency.

Main Methods:

• Proposed a novel RL-based adaptive learning system with an enhanced reward function.
• Integrated factors of material usefulness and individual learner prior knowledge into the reward.
• Utilized Monte Carlo simulation studies to evaluate the system's performance and recommendation strategies.

Main Results:

• The enhanced RL reward function significantly reduced learners' overall learning duration.
• The system demonstrated interpretable recommendation strategies that contributed to efficiency gains.
• Learning efficiency improved across learners with diverse levels of prior knowledge.

Conclusions:

• Optimizing RL rewards to include time efficiency alongside learning effectiveness enhances adaptive learning systems.
• Considering individual prior knowledge in RL is crucial for reducing learning time and increasing efficiency.
• The proposed RL approach offers a more balanced and effective adaptive learning solution.