Interaction Analysis by Humans and AI: A Comparative Perspective
Project Overview
The document explores the integration of generative AI in education, particularly through the use of Mixed Reality (MR) and Large Language Models (LLMs) to enhance communication and collaboration among children during gameplay. The study finds that MR significantly boosts emotional expression and engagement among participants, indicating its potential to enrich educational experiences. However, challenges such as transcription inaccuracies hinder its effectiveness compared to traditional video conferencing tools like Zoom, which offer simpler interfaces. The research suggests that while MR can transform learning environments by fostering better communication and interaction, the current technical limitations must be addressed to maximize its potential in educational settings. Overall, the findings underscore the promise of generative AI technologies in facilitating collaborative learning, while also highlighting the need for further development to overcome existing barriers.
Key Applications
Large Language Models for transcription, correction, and emotion detection
Context: Educational context with children participating in a gesture-based guessing game
Implementation: Using LLMs (like GPT-3.5 and GPT-4) to automate transcription and emotion analysis from video and audio recordings
Outcomes: Significantly reduced processing time, improved ability for non-Finnish speakers to analyze data, enhanced emotional insights during gameplay
Challenges: Limitations in annotation accuracy, technical issues with MR systems affecting engagement
Implementation Barriers
Technical Barrier
Challenges related to the accuracy of transcriptions and the technical performance of the MR system
Proposed Solutions: Utilizing advanced LLMs for improved transcription and emotion detection, while addressing bugs and improving the technical setup
Project Team
Maryam Teimouri
Researcher
Filip Ginter
Researcher
Tomi "bgt" Suovuo
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Maryam Teimouri, Filip Ginter, Tomi "bgt" Suovuo
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