VTutor: An Animated Pedagogical Agent SDK that Provide Real Time Multi-Model Feedback
Project Overview
The document discusses VTutor, an innovative open-source software development kit designed for creating animated pedagogical agents (PAs) that utilize large language models (LLMs) to deliver personalized feedback in educational environments. VTutor overcomes the shortcomings of traditional PAs by enabling real-time, adaptive interactions through a streamlined architecture suitable for web platforms. Its engaging anime-style characters effectively reduce the uncanny valley effect, thereby boosting learner engagement and motivation. Evaluation findings demonstrate that VTutor significantly surpasses conventional talking-head models in multiple dimensions, underscoring its effectiveness in enhancing AI-driven educational experiences. Overall, the implementation of VTutor showcases the transformative potential of generative AI in education, paving the way for more interactive and personalized learning opportunities.
Key Applications
VTutor - Animated Pedagogical Agent SDK
Context: Educational settings, targeting learners of various ages and backgrounds
Implementation: VTutor integrates LLMs for text generation and uses text-to-speech technology for audio output, with real-time lip synchronization and animations rendered via WebGL.
Outcomes: Improved engagement and learning outcomes, personalized feedback, and user preference for VTutor over traditional systems.
Challenges: Initial loading times and limitations in facial expression realism.
Implementation Barriers
Technical Barrier
Existing PAs often rely on scripted dialogues, which limit real-time adaptability and engagement.
Proposed Solutions: VTutor integrates LLMs to provide real-time, adaptive interactions.
User Experience Barrier
The uncanny valley effect can discourage learners from interacting with highly realistic agents. VTutor uses anime-style characters to maintain engagement and comfort.
Proposed Solutions: VTutor employs visually engaging and comfortable character designs to enhance learner interaction.
Project Team
Eason Chen
Researcher
Chenyu Lin
Researcher
Yu-Kai Huang
Researcher
Xinyi Tang
Researcher
Aprille Xi
Researcher
Jionghao Lin
Researcher
Kenneth Koedinger
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Eason Chen, Chenyu Lin, Yu-Kai Huang, Xinyi Tang, Aprille Xi, Jionghao Lin, Kenneth Koedinger
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