What is it about?
Many students struggle with mathematics because timely, curriculum-aligned, and personalized help is hard to access. Popular AI chatbots can give quick answers but often fail to diagnose why a student is stuck or to align with local school curricula. To address this, we developed and tested an AI Math Tutor for 9th-grade students in Vietnam. The system accepts text, photos of homework, and voice messages, and processes inputs with privacy-preserving measures. A core innovation is that the tutor analyzes conversations to detect specific misconceptions. When it finds an error pattern, it automatically generates tailored practice problems and recommends curriculum-aligned videos targeted to that mistake. Classroom trials showed that students found this adaptive feedback much more helpful and supportive than generic answers. This work offers a practical roadmap for creating AI tutors that deliver targeted, context-aware educational support and can assist teachers or be extended to other subjects and educational systems. Read the full paper: DOI 10.1145/3746274.3760396.
Featured Image
Photo by Compare Fibre on Unsplash
Why is it important?
This research is important because it offers a practical roadmap for designing AI educational tools that go beyond simply providing answers. By creating a system that can proactively diagnose student weaknesses and adapt its teaching strategies, we take a step toward truly scalable and personalized education. Beyond the technical innovation, this work addresses a broader educational equity challenge. Many AI tools are designed primarily for English-speaking learners, which limits their usefulness in local contexts. By developing a culturally aware and curriculum-aligned tutoring system, our research helps bridge that gap - offering a model for how AI can serve students in diverse linguistic and cultural environments. Ultimately, this approach contributes to a more inclusive vision of AI in education, where intelligent systems don’t just assist learning, but truly understand and support each student’s unique journey.
Perspectives
My motivation for this project was deeply personal. Growing up in Vietnam, I saw how many students struggled to find personalized math support that truly matched our local curriculum. While AI tools could provide answers, they often felt like passive search engines - they couldn’t observe, understand, or adapt like a real tutor. I wanted to change that. Our goal was to build an AI system that doesn’t just answer questions, but learns from each interaction and actively supports the student’s progress. The most rewarding moment came during user testing, when the chatbot recognized a student’s specific misconception and automatically generated tailored exercises. Seeing students say they felt “understood” by the AI was incredibly validating - it showed that empathy and adaptiveness are just as important as accuracy in educational AI. This project strengthened my belief that the future of AI in education lies in creating culturally aware, empathetic, and adaptive systems that align with local learning contexts. I hope our work serves as a meaningful example and inspires others to design AI tutors that truly understand students - wherever they are in the world.
Quy Minh Le
ISY Labs
Read the Original
This page is a summary of: Adaptive Multi-Agent Tutoring AI for Multimodal Mathematics Conversational Learning, October 2025, ACM (Association for Computing Machinery),
DOI: 10.1145/3746274.3760396.
You can read the full text:
Contributors
The following have contributed to this page
