Efficacy of a Computer Tutor that Models Expert Human Tutors
Project Overview
The document explores the application of generative AI in education, focusing on a study of an intelligent tutoring system (ITS) named Guru, designed to replicate the effectiveness of expert human tutors in teaching high school biology. Over a 9-week period, the study compared the learning outcomes of students using the ITS with those receiving instruction from human tutors and those in a traditional classroom setting. The findings revealed that both the ITS and human tutors significantly enhanced student learning compared to the classroom-only condition, demonstrating that generative AI can effectively personalize education and adapt to individual learning needs. This study underscores the promise of generative AI technologies in creating innovative educational tools, such as adaptive assessment instruments and tutoring systems that emulate expert instructional strategies, ultimately aiming to improve educational outcomes and accessibility.
Key Applications
Guru - an intelligent tutoring system for high school biology
Context: High school biology education for tenth graders in an urban school setting.
Implementation: A 9-week study where students interacted with the ITS and human tutors in a controlled environment, receiving tutoring sessions that followed the state curriculum.
Outcomes: Significant positive effects on immediate and delayed tests for both the ITS and human tutors compared to classroom instruction.
Challenges: The study did not include a novice tutoring system for comparison, and the classroom instruction was limited to one teacher, which could affect results.
Implementation Barriers
Research Design Limitations
The study lacked a direct comparison with a novice tutoring system and was limited by the constraints of a single classroom teacher.
Proposed Solutions: Future studies could include a novice ITS and expert human tutors with defined experience criteria for more robust comparisons.
Assessment Construction Challenges
Creating effective assessments for dynamic tutoring was difficult, as expert tutors do not use pre-planned agendas.
Proposed Solutions: Generative AI could be used to dynamically produce assessment items that match the quality of human-generated questions.
Project Team
["Andrew M. Olney", "Sidney K. D"Mello", "Natalie Person", "Whitney Cade", "Patrick Hays", "Claire W. Dempsey", "Blair Lehman", "Betsy Williams", "Art Graesser"]
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: ["Andrew M. Olney", "Sidney K. D"Mello", "Natalie Person", "Whitney Cade", "Patrick Hays", "Claire W. Dempsey", "Blair Lehman", "Betsy Williams", "Art Graesser"]
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