Personalized Learning Squared (PLUS)
Hybrid Human-AI Tutoring in Middle-School Mathematics
Embedded Files
Challenge
Challenge
The efficacy of both human tutors and AI tutors (e.g., Intelligent Tutoring Systems and Adaptive Tutoring Systems) is well established. However, human tutoring can be difficult to scale, and AI tutors lack the accountability structures and implementation guardrails needed to ensure consistent student use, as highlighted by Holt's recent "5% problem" article.
Might combining AI tutor use with small, targeted portions of human tutoring produce synergistic benefits beyond what either approach could achieve alone?
Solution
Solution
In collaboration with Ken Koedinger, the project PI, our lab is developing PLUS, a hybrid human-AI tutoring program. PLUS trains college students to provide supplemental motivational and content support to school students as they practice math using AI tutors, such as IXL, MATHia, and i-Ready. As of Spring 2026, PLUS supports about 3,500 students from 16 schools across 6 U.S. states each week.
Highlights
Highlights
Our classroom evaluation studies show that hybrid human–AI tutoring almost doubles academic growth (+61% on MAP) while also improving student time on task (+25%) and skill proficiency (+36%) in the AI tutor (IXL), compared to AI-only tutoring (Gurung et al., 2026).
Virtual Human-AI tutoring ($397-$700) is significantly less costly than in-person human tutoring ($3,500-$4,300) and in-person human-AI tutoring ($2,450-$3,000) (Gurung et al., 2026; Gurung et al., 2025; Thomas et al., 2023).
Key Translation Results: From In-Person to Remote Settings
Key Translation Results: From In-Person to Remote Settings
Human-AI tutoring Improves AI-tutor Progress
Human-AI tutoring Improves AI-tutor Progress
Access to supplemental human tutor support during AI tutor practice over an academic year improved students’ grade-level progression by an additional 0.34 grade levels compared with propensity-matched students who used AI tutors only (Gurung et al., 2025).
Remote Human-AI tutoring
Remote Human-AI tutoring
Access to supplemental human tutor support during AI tutor practice increased MAP growth by +0.61 relative to AI-only support, with Human–AI students achieving nearly double (1.9x) expected annual growth within the locally comparable IK bandwidth (Gurung et al., 2026).
On-going Research
On-going Research
As PLUS has matured across four years of field implementation, we are now conducting an RCT to rigorously examine program impact and extend prior formative quasi-experimental evaluations.
We are also conducting simulation studies and implementation evaluations to identify sources of variation in implementation, including variance at the student, teacher, and school context, with implications for how mitigating such variance at scale would further enhance the benefits of human–AI tutoring.
Discoveries
Discoveries
New paradigm for scaling Human-AI tutoring: Remote human-AI tutoring provides empirical support for a scalable alternative that improves student learning by providing supplemental human support during AI tutor practice.
Adaptive human tutor allocation benefits all: Proactively supporting lower-performing students while providing reactive, on-demand support to higher-performing students improves the learning outcomes of the entire cohort (Gurung et al., 2026).
Selected Publications
Selected Publications
Gurung, A., Gao, G., Gutterman, J., Guo, B., Houk, A., Gatz, E., Thomas, D. R., Gupta, S., Branstetter, L., Brunskill, E., Aleven, V., & Koedinger, K. (2026, July). Improving Hybrid Human-AI Tutoring by Differentiating Human Tutor Roles Based on Student Needs. In Proceedings of the 19th International Conference on Educational Data Mining.
Gurung, A., Lin, J., Gutterman, J., Thomas, D. R., Houk, A., Gupta, S., Brunskill, E., Branstetter, L., Aleven, V., & Koedinger, K. (2025, July). Human Tutoring Improves the Impact of AI Tutor Use on Learning Outcomes. In International Conference on Artificial Intelligence in Education (pp. 393-407). Cham: Springer Nature Switzerland. https://doi.org/10.1007/978-3-031-98459-4_28
Thomas, D. R., Lin, J., Gatz, E., Gurung, A., Gupta, S., Norberg, K., Fancsali, S. E., Aleven, V., Branstetter, L., Brunskill, E., & Koedinger, K. R. (2024, March). Improving student learning with hybrid human-AI tutoring: A three-study quasi-experimental investigation. In Proceedings of the 14th learning analytics and knowledge conference.
(TODO: Ashish) Add the Chine et al paper.
Page updated
Report abuse