Online Pedagogical Tutorial Tactics Optimization Using Genetic-Based Reinforcement Learning.

Area of Science:

• Artificial Intelligence
• Educational Technology
• Machine Learning

Background:

• Intelligent Tutoring Systems (ITS) utilize tutorial tactics to optimize student learning pathways.
• Previous research indicates tutorial tactics significantly impact learning gains, even with identical content.
• Existing Reinforcement Learning (RL) methods for tactic induction face scalability limitations and often rely on offline learning.

Purpose of the Study:

• To introduce and evaluate a novel Genetic-Based Reinforcement Learning (GBML) approach for inducing tutorial tactics.
• To enable online learning of tutorial tactics without requiring pre-existing datasets.
• To enhance the scalability of RL for complex ITS problems.

Main Methods:

• Developed a Genetic-Based Reinforcement Learning (GBML) framework for online tutorial tactic induction.
• Integrated a genetic-based optimizer for efficient rule discovery and generation.
• Applied the GBML method to learn adaptive policies directly from the learning environment.

Main Results:

• The GBML method successfully induced effective tutorial tactics in an online learning setting.
• The approach demonstrated enhanced scalability compared to traditional RL methods for larger problems.
• Experimental results validated the GBML method's capability in learning optimal tutorial strategies.

Conclusions:

• The GBML approach is a viable and scalable method for inducing tutorial tactics in Intelligent Tutoring Systems.
• This research supports the use of GBML for developing real-world ITS applications.
• Online learning with GBML offers a promising direction for adaptive educational technology.