How Do Teachers Create Pedagogical Chatbots?: Current Practices and Challenges
Project Overview
The document explores the integration of generative AI, particularly AI chatbots, into the educational landscape, emphasizing the role of teachers in developing pedagogical chatbots using large language models (LLMs). Based on semi-structured interviews with K-12 educators, it reveals that teachers are motivated to customize these chatbots to align with their lesson plans, ensuring relevance and effectiveness in their classrooms. However, they encounter several challenges, including a limited understanding of LLMs, difficulties in the technical implementation of chatbots, and the necessity for robust monitoring and feedback systems to evaluate the chatbots' impact on learning. The findings underscore the importance of teacher involvement in the design and deployment process, as well as the need for support and resources to overcome barriers, ultimately illustrating the potential of generative AI to enhance educational experiences when appropriately implemented.
Key Applications
Pedagogical Chatbots
Context: K-12 education, targeting both novice and experienced teachers
Implementation: Teachers create chatbots tailored to specific class activities using commercial platforms like Mizou.
Outcomes: Enhanced personalized learning experiences; teachers gain insights into chatbot capabilities and integration into lessons.
Challenges: Novice teachers struggle with initial design and technical implementation; experienced teachers face challenges with analyzing data and managing interactions.
Implementation Barriers
Technical Barrier
Limited understanding of LLMs leads to over- or under-estimation of chatbot capabilities.
Proposed Solutions: Providing detailed references to lesson plans and examples to help teachers adapt chatbots.
Time Constraint
Prompt engineering and debugging require significant time investment from teachers.
Proposed Solutions: Implementing prompt templates and collaborative design processes to reduce workload.
Cognitive Load
Teachers struggle to monitor and intervene in real-time during chatbot interactions due to cognitive burden.
Proposed Solutions: Developing interfaces that support real-time monitoring and intervention during lessons.
Feedback Limitations
The manual review of open-ended conversations hinders effective feedback post-class.
Proposed Solutions: Creating analytical tools to help teachers review and assess student-chatbot interactions efficiently.
Project Team
Minju Yoo
Researcher
Hyoungwook Jin
Researcher
Juho Kim
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Minju Yoo, Hyoungwook Jin, 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