Teach AI How to Code: Using Large Language Models as Teachable Agents for Programming Education
Project Overview
The document examines the application of generative AI, specifically large language models (LLMs), in educational settings, focusing on programming education through the development of TeachYou, a system featuring a teachable agent named AlgoBo. This system promotes Learning by Teaching (LBT) and addresses advantages such as scalability and decreased psychological barriers for learners, while also recognizing the challenges of creating knowledge models. Key features of TeachYou include a Reflect-Respond prompting pipeline to simulate knowledge acquisition, Mode-shifting to foster knowledge-building dialogues, and a Teaching Helper that provides metacognitive feedback. Evaluations indicate that the system enhances the quality of dialogue and knowledge development among learners. Moreover, the document highlights various applications of generative AI in education, such as interactive tutoring systems that leverage LLMs for teaching programming concepts, offering metacognitive scaffolding, and utilizing teachable agents that adapt based on student interactions. Overall, these AI-driven approaches significantly boost student engagement, improve learning outcomes, and encourage reflective learning strategies, demonstrating the transformative potential of generative AI in educational contexts.
Key Applications
AI-Assisted Teaching and Learning Agents
Context: Programming education for novices and undergraduate students, as well as K-12 students, focusing on algorithm learning, teaching assistant training, and factual knowledge assessment.
Implementation: The use of AI agents and chatbots, including GPT-based models and interactive dialogue systems, to facilitate teaching, learning, and assessment. These systems engage students through dynamic conversations, teaching simulations, and adaptive quizzes tailored to individual responses.
Outcomes: Enhanced understanding of programming concepts, improved teaching skills among students, increased student engagement, and personalized learning experiences. Teaching assistants are better prepared and more effective in real classroom settings.
Challenges: Participants may experience cognitive overload, and managing the chatbot’s responses can be challenging. There is also a need for effective design to ensure engagement and maintaining the accuracy of adaptive algorithms while ensuring high-quality feedback.
Implementation Barriers
Technological/Technical Barrier
High programming skills required for authoring knowledge models for teachable agents limits accessibility for educators. Additionally, there is difficulty in integrating AI systems with existing educational technologies.
Proposed Solutions: Use of LLMs to reduce manual effort in creating teachable agents, providing a more user-friendly interface for customization. Developing standardized APIs and frameworks for easier integration.
Psychological Barrier
Learners may feel discouraged to teach due to the perceived high competence of LLMs, leading to knowledge-telling instead of knowledge-building.
Proposed Solutions: Implementing prompting techniques to constrain the knowledge of the agents and encourage more inquiry-based interactions.
Cognitive Barrier
Students may experience cognitive overload when interacting with AI systems.
Proposed Solutions: Implementing user-friendly interfaces and providing adequate training for students.
Ethical Barrier
Concerns about data privacy and the ethical use of AI in educational contexts.
Proposed Solutions: Establishing clear guidelines and policies for data usage and ensuring transparency.
Project Team
Hyoungwook Jin
Researcher
Seonghee Lee
Researcher
Hyungyu Shin
Researcher
Juho Kim
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Hyoungwook Jin, Seonghee Lee, Hyungyu Shin, Juho Kim
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