Students' Perceived Roles, Opportunities, and Challenges of a Generative AI-powered Teachable Agent: A Case of Middle School Math Class
Project Overview
The document explores the use of generative AI (GenAI) in educational settings, particularly through the implementation of Teachable Agents (TAs) in middle school math classes. It emphasizes how students perceive these AI-powered agents as valuable learning companions that foster a deeper understanding of mathematical concepts through a 'learning-by-teaching' strategy. Key applications of GenAI include enhancing student engagement and facilitating personalized learning experiences. However, the findings also reveal notable challenges, such as the TAs' presentation skills, their limited adaptability to diverse learning styles, and the essential role of teacher involvement in maximizing effectiveness. These barriers highlight the need for careful consideration and improvement in the design and deployment of GenAI tools in education to fully harness their potential for supporting student learning.
Key Applications
GenAI-powered Teachable Agent for middle school math learning
Context: Middle school math classroom, targeting 6th-grade students
Implementation: Developed a GenAI-powered TA named [ANONYMOUS] that engages students in math problem-solving through conversation, utilizing a knowledge graph and large language models.
Outcomes: Increased student engagement, improved understanding of math concepts, enhanced metacognitive skills, and positive attitudes towards math and AI.
Challenges: Lack of problem presentation skills, inability to accommodate different learning styles, need for teacher support, and technical issues.
Implementation Barriers
Technical Barrier
Technical errors such as server loading times and internet connection issues hinder natural conversations with the TA.
Proposed Solutions: Improving server capacity and providing a more robust internet connection.
Pedagogical Barrier
The TA's inability to present problems clearly and its surface-level questioning limited the depth of student engagement. Additionally, students expressed a need for teacher engagement during interactions with the TA for support and guidance.
Proposed Solutions: Enhancing the TA's question generation to include higher-order thinking prompts. Encouraging teachers to participate actively in the learning process alongside the TA.
Learning Style Barrier
The TA did not accommodate various learning styles, which made it difficult for some students to engage effectively.
Proposed Solutions: Developing multimodal communication strategies that cater to different learning preferences.
Social-Emotional Barrier
The TA lacked social-emotional skills, making it less relatable for students who wished to engage in informal interactions.
Proposed Solutions: Integrating more human-like conversational capabilities and understanding of colloquial language.
Project Team
Yukyeong Song
Researcher
Jinhee Kim
Researcher
Zifeng Liu
Researcher
Chenglu Li
Researcher
Wanli Xing
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Yukyeong Song, Jinhee Kim, Zifeng Liu, Chenglu Li, Wanli Xing
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