A Network Connecting School Leaders From Around The Globe
4 Studies About AI Tutors Every Teacher Should Know
By Erik Ofgang
Tech&Learning
Summary of “4 Studies About AI Tutors Every Teacher Should Know”
By Erik Ofgang
Published December 16, 2024
AI tutors hold significant promise for individualized student learning, but current research presents a mixed picture. While some studies reveal potential benefits, others highlight pitfalls when AI tutors are improperly used. Erik Ofgang examines four key studies to help educators understand the complexities of AI tutoring, its successes, and challenges.
A study conducted at the University of Pennsylvania involving nearly 1,000 students revealed that AI tutors might hinder learning outcomes when misused. Students who relied on GPT-4-powered AI tutors for math performed worse than those without AI assistance. Researchers found that students often used the AI as an “answer machine,” bypassing opportunities to engage with the material and learn problem-solving strategies.
Lead researcher Alp Sungu clarified that AI itself isn’t inherently harmful; the issue arises when students misuse it for immediate answers rather than as a tool for learning. This study highlights a key takeaway: AI integration requires structured implementation and teacher oversight to ensure it fosters learning rather than encouraging shortcuts.
A meta-analysis by Yu Zhonggen from the Beijing Language and Culture University reviewed 24 studies on AI chatbots and learning. The findings showed that AI tutors significantly improved college students’ learning outcomes by enhancing motivation, self-efficacy, interest, and perceived value of education. Additionally, AI chatbots helped reduce student anxiety, further aiding academic performance.
However, the positive results were not universal. Younger students did not experience statistically significant benefits, and for college students, the motivational impact diminished over time. This suggests that while AI tutors can provide a short-term boost, long-term engagement strategies are needed to sustain their effectiveness.
In collaboration with PBS Kids, Professor Ying Xu from the Harvard Graduate School of Education conducted a study on AI-assisted learning with children aged 4-7. Students were divided into three groups:
The group that interacted with the AI-assisted Elinor outperformed others by answering the most questions correctly. Notably, this study did not rely on generative AI but instead used pre-selected, vetted responses, ensuring reliable feedback and avoiding AI “hallucinations” or misleading answers.
The study demonstrates that AI-assisted tools can significantly enhance learning outcomes for young children when used in structured environments. It also suggests that AI-based tools don’t necessarily need generative AI to be effective, especially for foundational learning.
Tutor CoPilot, a tool designed at Stanford University, illustrates how AI can support rather than replace human tutors. In a study involving nearly 1,000 students and 900 tutors, AI-assisted tutoring demonstrated clear benefits: students who worked with tutors using Tutor CoPilot were 4% more likely to master a topic after a session. Notably, students working with the lowest-rated tutors showed the greatest improvement, as the AI provided suggestions for better questioning and teaching strategies.
This study stands out because it reinforces the value of human connection in learning while showcasing AI’s role in improving tutor efficiency and instructional quality. Instead of competing with educators, AI tools like Tutor CoPilot augment their capabilities, especially benefiting less experienced tutors.
The research into AI tutors reveals a nuanced story: while AI has the potential to improve learning outcomes, its success depends on appropriate implementation. Misuse, such as relying on AI for immediate answers, can undermine learning, as seen in math studies. Conversely, AI tools that focus on interaction, vetted feedback, and tutor support have shown measurable benefits, particularly for college students and young learners.
Educators exploring AI tutors should prioritize structured usage, balancing AI’s capabilities with human guidance. Tools like Tutor CoPilot highlight creative ways AI can enhance teaching effectiveness without replacing essential human connections. As research evolves, the future holds promise for more innovative and impactful AI applications in education.
Source: Erik Ofgang, “4 Studies About AI Tutors Every Teacher Should Know”, December 16, 2024.
------------------------------
Prepared with the assistance of AI software
OpenAI. (2024). ChatGPT (4) [Large language model]. https://chat.openai.com
Tags:
SUBSCRIBE TO
SCHOOL LEADERSHIP 2.0
School Leadership 2.0 is the premier virtual learning community for school leaders from around the globe. Our community is a subscription based paid service ($19.95/year or only $1.99 per month for a trial membership) which will provide school leaders with outstanding resources. Learn more about membership to this service by clicking one our links below.
Click HERE to subscribe as an individual.
Click HERE to learn about group membership (i.e. association, leadership teams)
__________________
CREATE AN EMPLOYER PROFILE AND GET JOB ALERTS AT
SCHOOLLEADERSHIPJOBS.COM