An Educational System for Personalized Teacher Recommendation in K-12 Online Classrooms
Project Overview
The document outlines the application of generative AI in education, specifically through a teacher recommendation system tailored for K-12 online one-on-one classrooms. This innovative system generates pseudo matching scores to evaluate and rank potential student-teacher pairs based on a variety of features, enhancing the visibility of new teachers while fostering diversity in the recommendations. Real-world tests of the system have shown notable improvements, resulting in significant reductions in the number of attempts required to successfully match students with suitable teachers. Overall, the findings suggest that generative AI can effectively streamline the process of teacher-student pairing, making it more efficient and responsive to individual needs in the educational landscape.
Key Applications
Teacher recommendation system for online one-on-one classes
Context: K-12 online education, targeting students and teachers
Implementation: Developed a four-component framework that includes pseudo matching scores, a ranking model, novelty boosting for new teachers, and a diversity metric.
Outcomes: Reduced student-teacher matching attempts from 7.22 to 3.09, improved recommendation accuracy and diversity.
Challenges: Limited teacher availability, lack of ground truth for matching success, cold-start problem for new teachers, and high demand for diverse recommendations.
Implementation Barriers
Operational challenge
Limited sizes of demand (students) and supply (teachers) make it difficult to match students with available teachers.
Proposed Solutions: Develop a system that gives extra ranking incentives to new teachers to boost their visibility.
Data quality issue
Lack of reliable ground truth data for matching success due to noisy ratings from K-12 students.
Proposed Solutions: Generate pseudo matching scores based on student dropouts and course completions.
Cold-start problem
New teachers have difficulty being recommended due to lack of prior performance data.
Proposed Solutions: Implement a novelty boosting module to enhance the ranking of new teachers.
Recommendation diversity
High demand for diverse recommendations to prevent multiple students from booking the same teacher.
Proposed Solutions: Introduce a diversity metric that guards against recommending the same teachers to different students.
Project Team
Jiahao Chen
Researcher
Hang Li
Researcher
Wenbiao Ding
Researcher
Zitao Liu
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Jiahao Chen, Hang Li, Wenbiao Ding, Zitao Liu
Source Publication: View Original PaperLink opens in a new window
Project Contact: Dr. Jianhua Yang
LLM Model Version: gpt-4o-mini-2024-07-18
Analysis Provider: Openai