Intelligent Tutoring Systems

The 2 Sigma Problem

In 1984, educational psychologist Benjamin Bloom published a landmark finding: students who received one-on-one tutoring with mastery learning techniques performed two standard deviations better than students in conventional classroom instruction.

"The average tutored student was above 98% of the students in the control class."

— Benjamin Bloom, "The 2 Sigma Problem" (1984)

Two standard deviations is roughly the difference between a C student and an A student. Bloom's research showed that this gap wasn't about innate ability—it was about the method of instruction. Given the right conditions, most students could achieve mastery.

The problem? One-on-one tutoring is prohibitively expensive. Bloom called this "the 2 sigma problem": finding methods of group instruction as effective as personal tutoring.

AI as the Solution

For decades, the 2 sigma problem seemed unsolvable. But advances in AI—particularly large language models and adaptive learning systems—offer a new possibility: delivering personalized, one-on-one instruction at scale.

An intelligent tutoring system (ITS) can:

Adapt in real-time to each student's level of understanding
Identify knowledge gaps and address misconceptions immediately
Provide unlimited patience for students who need more time
Offer Socratic dialogue that guides students to discover answers themselves
Scale infinitely without the cost constraints of human tutors

Our Research Focus

Adaptive Assessment

Building systems that can accurately diagnose what a student knows and doesn't know—not through standardized tests, but through natural conversation and problem-solving.

Personalized Learning Paths

Developing algorithms that create individualized curricula, adjusting difficulty, pacing, and teaching style based on each student's needs and preferences.

Socratic AI Tutors

Training AI systems to guide students through discovery rather than simply providing answers—asking the right questions at the right time to build deep understanding.

Knowledge State Modeling

Creating representations of student knowledge that can track mastery across interconnected concepts, identifying prerequisites that need reinforcement.

Challenges We're Addressing

Beyond Pattern Matching

Current AI can be confidently wrong. We're working on systems that know the boundaries of their knowledge and escalate appropriately when they're uncertain.

Maintaining Engagement

Effective tutoring isn't just about correct information—it's about motivation, encouragement, and building confidence. We're exploring how AI can support the emotional aspects of learning.

Equitable Access

The promise of AI tutoring is democratizing access to high-quality education. We're committed to building systems that work for all students, not just those with resources.

The Canadian Context

Canada's diverse, multilingual population and strong public education system make it an ideal environment for developing and testing intelligent tutoring systems. We're particularly interested in:

Supporting French-English bilingual education
Indigenous language preservation and education
Rural and remote learning access
Privacy-preserving approaches that comply with Canadian regulations

Get Involved

We're seeking partnerships with educators, school boards, educational technology companies, and researchers who share our vision for AI-powered personalized learning.