Generative AI, including Language Models (LMs), holds the promise to reshape key sectors like education, healthcare, and law, which rely heavily on skilled professionals to navigate complex responsibilities. In education, for instance, effective teacher training with expert feedback is crucial yet costly, limiting opportunities to enhance educational quality on a larger scale.
In a new paper Tutor CoPilot: A Human-AI Approach for Scaling Real-Time Expertise, a Stanford University research team presents Tutor CoPilot, a new model that offers expert-level guidance to tutors in real time. This study is the first of its kind—a randomized controlled trial testing a Human-AI system in live tutoring scenarios.
Tutor CoPilot aims to enhance K-12 education by providing immediate, actionable guidance to tutors, ultimately improving the live learning experience for students.
In collaboration with FEV Tutor, a virtual tutoring provider, and a Southern U.S. school district, the researchers conducted a large-scale intervention involving 900 tutors and 1,800 K-12 students from Title I schools. This in-school, virtual math tutoring program allows tutors to access Tutor CoPilot during sessions by pressing a button for real-time assistance. Tutor CoPilot ensures user safety and privacy by de-identifying names and limiting information shared with external services. The AI-generated guidance draws on the Bridge method, which models expert thinking by capturing reasoning patterns, and also allows for user customization.
The researchers summarize their main foundlings as follows:
- Enhanced Student Outcomes: Tutor CoPilot significantly improves learning results, with students of tutors who use the tool being 4% more likely to master lesson topics, as shown through an intent-to-treat analysis.
- Greater Benefits for Less-Experienced Tutors: Less-experienced or lower-rated tutors saw the most significant improvement, with their students’ mastery rates increasing by up to 9% over the control group, according to an analysis of tutor ratings.
- Increased Use of High-Quality Strategies: Tutors with access to Tutor CoPilot are more likely to employ effective strategies that promote student understanding, as identified by classifiers assessing the quality of instructional approaches.
- Positive Tutor Feedback with Suggestions for Refinement: Tutors found Tutor CoPilot helpful but suggested improvements, particularly around tailoring the guidance to appropriate grade-level language.
With an estimated cost of $20 per tutor per year, Tutor CoPilot offers an affordable, scalable alternative to traditional, resource-intensive training methods.
Overall, this study demonstrates Tutor CoPilot’s potential as an effective Human-AI solution that integrates LMs with expert insights for tangible, positive outcomes in real-world settings. By supporting under-served student communities, Tutor CoPilot not only enhances educational quality but also paves the way for AI-driven expertise to transform other critical domains.
The paper Tutor CoPilot: A Human-AI Approach for Scaling Real-Time Expertise is on arXiv.
Author: Hecate He | Editor: Chain Zhang
We know you don’t want to miss any news or research breakthroughs. Subscribe to our popular newsletter Synced Global AI Weekly to get weekly AI updates.