Skip to main content Skip to navigation

The StudyChat Dataset: Student Dialogues With ChatGPT in an Artificial Intelligence Course

Project Overview

The document explores the integration of generative AI, particularly large language models (LLMs) like ChatGPT, in university-level education, focusing on an artificial intelligence course. It presents the StudyChat dataset, which records student interactions with LLMs during programming assignments, revealing both the potential advantages and challenges of using these technologies. Key applications of LLMs include personalized learning and enhanced student engagement, with findings indicating that students who engage with LLMs by asking conceptual questions tend to achieve better course performance. However, the study also raises concerns about the risks of misuse and over-reliance on LLMs, which may negatively impact learning outcomes. Overall, the document emphasizes the dual-edged nature of generative AI in education, highlighting the importance of fostering effective engagement strategies while mitigating the risks associated with dependency on AI tools.

Key Applications

StudyChat dataset and LLM-powered tutoring chatbot

Context: Upper-division university-level AI course, targeting Computer Science students

Implementation: Developed a web application mimicking ChatGPT functionalities for students to use during programming assignments without restrictions.

Outcomes: Insights into student interaction patterns with LLMs, correlations between usage patterns and course performance.

Challenges: Concerns about student misuse of LLMs, including over-reliance on AI for assignments.

Implementation Barriers

Ethical concerns

Potential misuse of LLMs by unmotivated students to complete assignments instead of learning.

Proposed Solutions: Understanding student behavior with LLMs and developing Intelligent Tutoring Systems (ITS) that adapt to encourage positive usage.

Reliance on technology

Students may become over-reliant on LLMs, which could undermine their learning and understanding of core concepts.

Proposed Solutions: Monitoring student interactions with LLMs to detect and address over-reliance in real-time.

Project Team

Hunter McNichols

Researcher

Andrew Lan

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Hunter McNichols, Andrew Lan

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