Supervisor-Worker Problems with an Application in Education.

Area of Science:

• Artificial Intelligence
• Educational Technology
• Machine Learning

Background:

• Student motivation in e-learning is crucial but challenged by skill practice choices.
• Students may avoid practicing difficult skills, hindering learning.
• Forcing practice can decrease motivation and increase dropout rates.

Purpose of the Study:

• To model the tradeoff between student choice and skill mastery in e-learning.
• To develop a multi-agent planning framework for personalized e-learning.
• To enhance student engagement and learning outcomes.

Main Methods:

• Formalized the Supervisor-Worker Problem with Partially Overlapping goals (SWOPP).
• Developed two algorithms for the supervisor agent to guide the worker (student).
• Deployed SWOPP in a real-world study for personalized math questions in K5 e-learning.

Main Results:

• SWOPP effectively guided students to practice necessary skills.
• The system successfully personalized math questions for K5 students.
• Student motivation was maintained during the e-learning sessions.

Conclusions:

• The SWOPP model provides an effective approach to balancing student autonomy and directed learning.
• Personalized e-learning using AI planning can improve skill acquisition without compromising motivation.
• This framework has practical applications in educational software design.