The Cognitive Tutor was an intelligent tutoring system developed at Carnegie Mellon University by the psychologist John R. Anderson and colleagues, including Albert Corbett and Kenneth Koedinger. Their paper “Cognitive Tutors: Lessons Learned” summarized roughly a decade of work, much of it carried out in Pittsburgh public-school classrooms where the team installed dedicated computer labs to teach mathematics.
What made the system distinctive was its grounding in a formal theory of human cognition. It was built on Anderson’s ACT theory (later ACT-R), which represents skill as a set of “production rules” - if-then steps a learner must acquire. The tutor ran a computational model that could solve each problem the way a student was expected to, tracked which rules a learner had mastered through a technique called knowledge tracing, and gave immediate, step-by-step feedback. The paper laid out eight design principles, including representing competence as production rules, giving immediate feedback, and minimizing working-memory load.
In 1998 the research spun out as Carnegie Learning, a company that sold Cognitive Tutor algebra and geometry curricula to schools nationwide; the US Department of Education later named its algebra course an exemplary curriculum.
Why business readers should care: long before chatbots, Cognitive Tutors showed that effective educational software depends less on the interface than on a defensible model of what the learner knows and what they should learn next. That “student model” is exactly what today’s LLM tutors still struggle to build reliably.