Cognitive tutor

Content sourced from Wikipedia, licensed under CC BY-SA 3.0.

A cognitive tutor is a type of intelligent tutoring system that uses a cognitive model to help students as they work on problems. It gives immediate feedback about whether actions are correct and can offer context-specific hints to guide the next steps.

Carnegie Learning’s Cognitive Tutor is a well-known version for high school math, based on John Anderson’s ACT-R theory of human thinking. But cognitive tutors started in the 1980s for research and have been built for other subjects like computer programming and science. They can be used in classrooms as part of blended learning, combining textbooks and software.

How it works
- Model tracing checks every student action (like entering a value or clicking a button) to see if it follows the expected steps.
- Knowledge tracing estimates which skills a student has learned and uses a progress indicator called the Skillometer.
- Together, model tracing and knowledge tracing monitor learning, guide students along the right problem-solving path, and provide feedback and help when needed.

Four-part architecture of intelligent tutoring systems
- Domain model: the rules, concepts, and knowledge of the subject.
- Student model: information about the student’s knowledge and learning progress.
- Tutoring model: decides when and how to intervene and what to teach next.
- Interface: how the tutor presents feedback, hints, and problems to the student.

The cognitive model and how tutors help
- A cognitive model mimics expert or advanced student thinking in the domain.
- The tutor uses this model to generate explanations, step-by-step feedback, and hints.
- Cognitive tutors provide a rich problem-solving environment, targeted error messages, and options for choosing the next problem.

Origins and history
- Early real-world uses in the 1980s included a geometry tutor for high school students and a LISP programming tutor for college students at Carnegie Mellon University (CMU).
- The PACT Center at CMU developed Cognitive Tutor Algebra, used in many schools by the late 1990s. Carnegie Learning later offered tutors to thousands of schools.
- Projects expanded to other courses and later products like Bridge to Algebra for middle schools.

Effectiveness and research
- Studies on Carnegie Cognitive Tutor show mixed results. Some high school math studies report positive effects, while others show little or negative effects.
- What Works Clearinghouse standards require large, well-randomized studies, which are hard to conduct in schools. So not all research meets these standards, even when results can be useful.

Limitations and challenges
- It is not practical to create cognitive tutors for every subject; most work has focused on algebra, geometry, programming, and related areas.
- Developing a cognitive tutor takes a lot of time and effort to build the domain knowledge, student models, and teaching strategies, which can limit flexibility.
- The system’s prompts and hints are designed for specific teaching approaches and may not fit every learner or classroom.
- Some critics worry that learners might use prompts to get answers too quickly, bypassing deeper learning.

What this means for use in schools
- Cognitive Tutors can be used in computer labs or during class as part of a blended learning approach.
- They can tailor practice to individual needs and provide immediate feedback.
- While they show potential to improve learning, their effectiveness varies by subject, student group, and implementation.