Research & Collaboration
We work with universities, schools, and research institutions to study how AI tutors and LLMs can enhance learning, improve student outcomes, and reduce educator workload.
Inspired by Bloom's Two Sigma Problem
In 1984, educational psychologist Benjamin Bloom published groundbreaking research showing that students who received one-on-one tutoring performed two standard deviations better than students in conventional classrooms. This places the average tutored student above 98% of traditionally taught students.
Bloom called this the "2 Sigma Problem": how can we provide the benefits of personalised tutoring to all students, not just those who can afford it?
Our research explores whether AI can help solve this challenge—providing scalable, personalised learning support that guides students through understanding using evidence-based pedagogical techniques.
"The tutoring process demonstrates that most of the students do have the potential to reach this high level of learning."
Benjamin Bloom
Educational Psychologist, 1984
Our Research Focus
Our research spans the full spectrum of AI-powered learning.
Architectures for AI Learning Systems
Developing novel AI architectures specifically designed for educational contexts, including pedagogical fine-tuning, retrieval-augmented generation, and intent classification for learning support.
Human-AI Learning Experience Design
Studying how students and educators interact with AI tutors, optimising interfaces and interactions that promote genuine understanding rather than surface-level answers.
Learning Outcomes & Impact
Measuring the real-world impact of AI tutoring on student learning outcomes, engagement, and long-term knowledge retention across diverse educational settings.
Pedagogical AI Techniques
Implementing and evaluating evidence-based teaching methods in AI systems, including Socratic questioning, scaffolding, and formative assessment strategies.
Our Research Overviews
We distill key findings from the latest academic research in AI and education.
AI tutoring outperforms active learning
A Harvard RCT shows AI tutoring more than doubled learning gains versus active learning.
Generative AI can harm learning?
A UPenn study on GPT-4 in high school math: tutor safeguards are essential.
Decoding Google's LearnLM technical paper
A detailed analysis of Google's approach to fine-tuning Gemini for education.
Research Partnerships
We collaborate with leading universities and educational institutions worldwide to advance the science of AI-powered learning.
Partner With Us
We're always looking for new partners in the academic space. If your institution is interested in collaborating on research regarding AI in education, we'd love to hear from you.