Skip to main content Skip to navigation

AI-University: An LLM-based platform for instructional alignment to scientific classrooms

Project Overview

The document presents AI University (AI-U), an innovative framework aimed at revolutionizing educational content delivery through the utilization of large language models (LLMs). By employing retrieval-augmented generation (RAG), AI-U facilitates the development of personalized AI assistants that cater to individual instructional styles and specific course materials. This approach seeks to enhance instructional efficiency and boost student engagement, particularly within higher education settings. Key applications include the customization of learning experiences that adapt to students' needs, thereby fostering more interactive and effective learning environments. The findings suggest that leveraging generative AI in this manner not only enhances the quality of education but also promotes a more engaging and tailored educational journey for learners, ultimately aiming to improve academic outcomes and prepare students for future challenges. Through the integration of advanced AI technologies, the document posits that AI-U can significantly transform the landscape of education, making it more responsive and aligned with the diverse needs of today's learners.

Key Applications

AI-University (AI-U)

Context: Graduate-level course on the Finite Element Method (FEM)

Implementation: Fine-tuning a large language model (LLM) with retrieval-augmented generation (RAG) to create an AI assistant that adapts to course materials and teaching styles.

Outcomes: Demonstrated strong alignment with course materials, increased student engagement, and enhanced instructional efficiency.

Challenges: Dependence on the availability of course-specific data and potential limitations in the adaptability of the system.

Implementation Barriers

Data Privacy

Concerns regarding the handling of student data and proprietary teaching materials when using commercial AI tools.

Proposed Solutions: Utilizing open-source resources for local hosting to ensure data privacy and equitable access.

Equitable Access

Students from different socioeconomic backgrounds may have varying levels of access to AI tools and technologies.

Proposed Solutions: Tailoring AI assistants to course-specific terminology and materials to enhance accessibility for all students.

Technical Limitations

RAG-based approaches may have limited context windows that affect response quality and relevance.

Proposed Solutions: Using knowledge graphs to improve scalability and performance.

Project Team

Mostafa Faghih Shojaei

Researcher

Rahul Gulati

Researcher

Benjamin A. Jasperson

Researcher

Shangshang Wang

Researcher

Simone Cimolato

Researcher

Dangli Cao

Researcher

Willie Neiswanger

Researcher

Krishna Garikipati

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Mostafa Faghih Shojaei, Rahul Gulati, Benjamin A. Jasperson, Shangshang Wang, Simone Cimolato, Dangli Cao, Willie Neiswanger, Krishna Garikipati

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