Skip to main content Skip to navigation

Combining Cognitive and Generative AI for Self-explanation in Interactive AI Agents

Project Overview

The document explores the application of generative AI in education through the development of the Virtual Experimental Research Assistant (VERA), an interactive platform designed for modeling ecological systems. It emphasizes the integration of cognitive AI to enhance self-explanation capabilities, utilizing a Task-Method-Knowledge (TMK) model to articulate its reasoning process, which aims to improve user comprehension and trust in the system. The preliminary evaluation of VERA's self-explanation module, Ask-TMK, reveals positive outcomes, demonstrating its effectiveness in producing accurate and pertinent explanations in response to user inquiries. By leveraging advanced technologies such as ChatGPT and LangChain, VERA showcases the potential of generative AI to facilitate a deeper understanding of complex ecological concepts, thereby enhancing the educational experience. Overall, the findings indicate that generative AI can play a significant role in promoting interactive and effective learning environments in educational settings.

Key Applications

Virtual Experimental Research Assistant (VERA)

Context: Inquiry-based learning for students exploring ecological systems.

Implementation: VERA integrates cognitive AI with generative AI methods, specifically using a TMK model and tools like ChatGPT and LangChain to provide explanations for user queries.

Outcomes: High recall, precision, and accuracy in generating relevant explanations, enhancing user understanding and trust.

Challenges: Potential biases in the training data and the need for real-world deployment to validate effectiveness.

Implementation Barriers

Data Bias

Potential biases in the training data used for developing AI explanations. Future work will focus on deploying VERA in diverse classroom settings to gather real-world data and assess equity and bias aspects.

Proposed Solutions: Deploy VERA in diverse classroom settings to gather real-world data and assess equity and bias aspects.

Implementation Challenges

The system has not yet been deployed in classroom environments, limiting comprehensive evaluation. Conduct user-centered studies to evaluate comprehensibility and refine VERA’s self-explanation method.

Proposed Solutions: Conduct user-centered studies to evaluate comprehensibility and refine VERA’s self-explanation method.

Project Team

Shalini Sushri

Researcher

Rahul Dass

Researcher

Rhea Basappa

Researcher

Hong Lu

Researcher

Ashok Goel

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Shalini Sushri, Rahul Dass, Rhea Basappa, Hong Lu, Ashok Goel

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

Let us know you agree to cookies