An intelligent tutoring system, or ITS, is software that tailors instruction to an individual learner by maintaining a model of what that learner does and does not yet know, then choosing problems, hints, and feedback accordingly. Unlike a fixed lesson or a simple quiz, an ITS reacts step by step to a student’s work, aiming to approximate the responsiveness of a one-to-one human tutor.
The best-known example is the Cognitive Tutor, developed by psychologist John R. Anderson and colleagues at Carnegie Mellon University and commercialized by the spin-off company Carnegie Learning. Carnegie Mellon’s own announcement of Anderson’s 2016 Atkinson Prize, cited here, credits him with “revolutionizing how we learn” and ties his work to the ACT-R theory of cognition. The Cognitive Tutor uses two techniques rooted in that theory: model tracing, which checks each student action against an internal model of the skill, and knowledge tracing, which estimates the probability that the student has mastered each underlying skill and updates it as they work.
ITS research predates modern machine learning and grew out of the symbolic, cognitive-science tradition, where a learner’s knowledge is represented as explicit rules. Today’s large-language-model tutors, such as Khan Academy’s Khanmigo, pursue the same goal - personalized, responsive instruction - but reach it through statistical models of language rather than hand-built cognitive models.
Why business readers should care: intelligent tutoring systems are a decades-old proof that adaptive, per-user software can measurably improve outcomes, and that the hard part is modeling the user well. That lesson - personalization depends on a good user model, not just a good interface - carries directly into recommendation, training, and AI-assistant products.