Trust and ethical considerations in a multi-modal, explainable AI-driven chatbot tutoring system: The case of collaboratively solving Rubik's Cube
Project Overview
The document explores the implementation of a multi-modal AI-driven chatbot tutoring system called ALLURE, designed to assist high school students in collaboratively solving Rubik's Cube challenges. It highlights the system's focus on ethical considerations, including data privacy, conversation appropriateness, and the prevention of information leakage, thereby ensuring a safe and reliable interaction between students and the AI. By utilizing explainable AI techniques, ALLURE delivers tailored feedback and detailed explanations, promoting an engaging and effective learning experience. The findings indicate that such generative AI applications not only enhance student collaboration and problem-solving skills but also address potential risks like abusive language and unauthorized data sharing, ultimately creating a supportive educational environment.
Key Applications
ALLURE chatbot
Context: High school students learning to solve Rubik's Cube
Implementation: Developed using the RASA framework with components for natural language processing, sentiment analysis, and explainable AI.
Outcomes: Improved student engagement, personalized learning experiences, and enhanced understanding of problem-solving through collaborative AI interaction.
Challenges: Ensuring data privacy, preventing information leakage, and maintaining a safe conversational environment for children.
Implementation Barriers
Ethical and Trust Issues
Concerns about data privacy, abusive language, and fairness in AI responses.
Proposed Solutions: Implementing strict data management protocols, using sentiment analysis to detect abusive language, and ensuring transparency in AI processes.
Technical Limitations
The black-box nature of AI and challenges in explainability.
Proposed Solutions: Utilizing explainable AI methods to improve transparency and ensure students can understand AI's reasoning.
Project Team
Kausik Lakkaraju
Researcher
Vedant Khandelwal
Researcher
Biplav Srivastava
Researcher
Forest Agostinelli
Researcher
Hengtao Tang
Researcher
Prathamjeet Singh
Researcher
Dezhi Wu
Researcher
Matt Irvin
Researcher
Ashish Kundu
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Kausik Lakkaraju, Vedant Khandelwal, Biplav Srivastava, Forest Agostinelli, Hengtao Tang, Prathamjeet Singh, Dezhi Wu, Matt Irvin, Ashish Kundu
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