Search CTAT

Timeline of Cognitive Tutor History

1982

Anderson, J. R. (1983). The Architecture of Cognition. Cambridge, MA: Harvard University Press.
John R Anderson head shot

John R Anderson completes the ACT* theory of learning and problem solving. The theory holds that a cognitive skill consists of numerous rules that each relate task goals and task states to actions and consequences. These rules are formalized as production rules that store goal-oriented knowledge. As opposed to declarative knowledge, procedural knowledge is stored in production rules.

Anderson's The Architecture of Cognition is published a year later.

1982

Anderson, J. R., Corbett, A. T., Koedinger, K., & Pelletier, R. (1995). Cognitive tutors: Lessons learned. The Journal of Learning Sciences, 4, 167-207.

Carnegie Mellon researchers begin work on LISP and geometry tutors. Anderson, Corbett, Koedinger, and Pelletier (1995, p 171) write: "These two tutors embodied a number of key ideas about how computer-based instruction should be realized".

1984

Albert Corbett head shot

Anderson and colleagues complete two cognitive tutors--the Lisp Programming Tutor and the Geometry Proof Tutor (GPT)--to validate the ACT* theory of cognition in a computer tutor.

The Lisp tutor is used in a mini-course taught by its developers at Carnegie Mellon University. Besides being a useful tool for learners, the tutor also aids Anderson and Corbett as they research the effectiveness of intelligent tutoring and the validity of the ACT* theory. In a study, students that worked with the Lisp Tutor completed problems as little as one third of the time required by students working in programming environment without the tutor. Students that used the tutor also score 25% higher on subsequent tests (Corbett & Anderson, 1991).

The Lisp tutor remains in use in the self-paced course until 1999.

1985

Anderson, J. R., Boyle, C. F., & Reiser, B. J. (1985). Intelligent tutoring systems. Science, 228, 456-462. (PDF)

More than 10,000 examples of educational software exist, but most are instances of computer-aided instruction (CAI). Intelligent computer-aided instruction (ICAI), or 'programs that simulate understanding of the domain they teach and that can respond specifically to the student's problem-solving strategies', are rare. John R Anderson estimates that 200 hours of development are required to create 1 CAI hour; estimates for ICAI are unknown.

John R Anderson and colleagues introduce a more interdisciplinary approach to ICAI that combines the discipline of cognitive psychology with that of artificial intelligence (Anderson, Boyle, and Reiser, 1985).

The Geometry Proof Tutor (GPT), developed to support students in completing Euclidean proofs, is piloted in Peabody High School in Pittsburgh from the Fall of 1985 until 1987. The Geometry Proof Tutor, as well as the Lisp Tutor, are considered 'hothouse' successes. Results show that the tutors are roughly 1/2 as effective as a human tutor, but 2-3 times more effective than conventional computer-aided instruction.

The GPT requires 'powerful computer workstations'--10 XEROX D-Machines costing roughly $20,000 each--and contains ~200 production rules; the LISP tutor contains 325 production rules and 475 'buggy' rules.

1986

koedinger headshot

For his dissertation, Kenneth Koedinger embarks on the ANGLE (A New Geometry Learning Environment) Geometry Tutor Project to test novel assumptions about learning. The new tutor will then be tested in the classroom. Like GPT, ANGLE is a problem-solving environment where students construct graphical representations of Euclidean proofs. Unlike GPT, ANGLE is a based on a cognitive model that models--and therefore facilitates thinking with--an expert-like representation of target knowledge.

By the time ANGLE is piloted at Langley High School in 1991, Pittsburgh city high schools have adopted a new curriculum that de-emphasizes proof; this makes long-term use of the tutor in the classroom unrealizable.

1987

An evaluation of an early algebra tutor is performed during the 1987-8 school year in Peabody High School. While there is no significant difference between the experimental and control classes, the researchers correctly hypothesize that the difference between the tutor interface and class interface (paper and pencil) prevented adequate transfer.

The researchers create a word algebra tutor and test it in the lab, but it is never tested in a classroom.

1989

Anderson, J. R., Conrad, F. G., & Corbett, A. T. (1989). Skill acquisition and the LISP Tutor. Cognitive Science, 13, 467-506.(PDF)

Using cognitive tutors they've developed, John Anderson and Al Corbett teach Lisp, Pascal, Prolog in a self-paced course at Carnegie Mellon University. The course is later modified to include Lisp and Prolog, and remains in use until 1999.

The journal Cognitive Science publishes Anderson, Conrad, and Corbett's paper titled "Skill Acquisition and the LISP Tutor" (1989), in which the authors report on experiments completed using the LISP tutor to study cognition. Their results confirm theories of the production rule as "the right unit of analysis in understanding the acquisition of a skill" (p 493). The researchers also conclude that when the structure of the domain is factored out, skill acquisition is simple: "Once the units of knowledge are identified, acquisition of the skill can be predicted by composing simple learning functions for these units" (p 493). This initial work on the topic of what would be known as "knowledge tracing" leads to empirical validation of the concept and incorporation of knowledge tracing in subsequent cognitive tutors.

1990

NCTM logo

An education crisis in mathematics is recognized across the United States. The National Council of Teachers of Mathematics (NCTM) produce a series of reports recommending algebra and geometry for all students. The NCTM specifies that new material should emphasize problem-solving, reasoning among multiple representations (e.g., tables, graphs, natural language), and communication.

1991

With the help of Pittsburgh public school teacher Bill Hadley, the Carnegie Mellon research team sets out to develop a mathematics curriculum into which their tutors would be positioned.

Around this time, the Pittsburgh Urban Mathematics Project, or PUMP, is formed. PUMP is a collaboration between the ACT Research Group of Carnegie Mellon University's Psychology Department, researchers from Carnegie Mellon's Human-Computer Interaction Institute, and teachers in the Pittsburgh Public Schools. PUMP produces a new algebra curriculum that is consistent with the recommendations of the NCTM.

Headed by Kenneth R. Koedinger, PUMP creates an intelligent tutoring system called PAT--the PUMP Algebra Tutor, or Practical Algebra Tutor--for use with the new curriculum.

1992

The PUMP Algebra I curriculum and tutor (PAT) are piloted at Pittsburgh's Langley High School.

1993

Koedinger, K. R., Anderson, J. R., Hadley, W. H., & Mark, M. A. (1997). Intelligent tutoring goes to school in the big city. International Journal of Artificial Intelligence in Education, 8, 30-43. (PDF)

Ritter, S., Anderson, J. R., Cytrynowicz, M., & Medvedeva, O. (1998). Authoring Content in the PAT Algebra Tutor. Journal of Interactive Media in Education, 98(9). (PDF)
Langley High School

In the 93-94 school year, the Carnegie Mellon researchers extend their Algebra I course two more schools for a total of three Pittsburgh Public High Schools (Langley, Brashear, and Carrick). Based on the new PUMP curriculum, the course is set up to serve as a large-scale classroom experiment. The experiment will determine whether the PUMP curriculum and PAT as a whole are more effective than the traditional curriculum without an intelligent tutoring system.

The Algebra I course leads to 'large and consistent' gains on the NCTM standards-oriented assessments (approximately 100% or double improvement over control subjects) and 'moderate and statistically reliable' gains on the standardized assessments (~15% improvement)(Corbett, Koedinger, and Hadley 2002). The results are replicated in 1994-1995 with two more high schools.

pSAT

At Carnegie Mellon, Dr. Steven Ritter begins work on pSAT, the Problem Situation Authoring Tool, an early user-friendly interface for authoring and encoding problems. Particularly useful for encoding word problems, pSAT allows a problem author to enter problem text along with a computer-understandable representation of the problem. The author then verifies the software's encoding of the problem, and saves it. In creating the problem representation, pSat removes the need for natural language parsing at runtime.

Later with an interface in Java, pSAT still runs in 2005. Problem authors at Carnegie Learning use pSAT to work on problems for their new 'Bridge to Algebra' software, an intervention product for middle school students.

1995

Two colleges pilot the Algebra I Cognitive Tutor--an adaptation of the Practical Algebra Tutor, or PAT--in developmental mathematics courses. Results show that at the two colleges, PAT students performed more than 50% better than students in regular classes on a performance-based assessment (Koedinger and Sueker 1996).

The total number of sites using the program is now eight.

The Carnegie Mellon researchers receive funding from five local foundations (Heinz, Buhl, Grable, Mellon and Pittsburgh Foundations) for creating full courses in geometry and algebra II.

The Carnegie Mellon researchers create the Pittsburgh Advanced Cognitive Tutor (PACT) Center to continue research, development, and piloting of cognitive tutor technology and courses.

Anderson et al estimate that creating a single production rule requires 10 or more hours.

1996

The Carnegie Mellon researchers offer their cognitive tutor to high schools outside the Pittsburgh region. Participating schools include four Department of Defense Education Activity (DoDEA) schools in Europe and a third college. By the end of 1997, the total number of sites is 24.

Ritter, S. & Koedinger, K. R. (1996). An architecture for plug-in tutor agents. Journal of Artificial Intelligence in Education, 7(3-4), 315-347. (PDF)

Steven Ritter and Kenneth R. Koedinger publish their paper 'An Architecture For Plug-In Tutor Agents', which establishes architectural principles for adding tutoring to a program.

1998

Carnegie Learning, Inc. logo

In July of 1998, Carnegie Mellon University founds Carnegie Learning, Inc. to assume market-driven research, development, and dissemination activities.

Cognitive Tutor® Algebra I is in use in 75 schools.

1999

In a joint effort with the University of Pittsburgh, the Carnegie Mellon researchers receive a major National Science Foundation grant to form a research center--this becomes known as the Center for Interdisciplinary Research on Constructive Learning Environments (CIRCLE)--to study human tutoring and to develop a more effective next generation of computer-based tutors.

The Cognitive Tutor Algebra I curriculum receives accolades from the US Department of Education, which gives it the highest "exemplary" designation awarded to only five curricula.

With the help of recently founded Carnegie Learning, Inc., the CMU researchers begin developing full Cognitive Tutor courses for middle school mathematics. Albert Corbett leads development for 7th and 8th grade and pre-algebra courses; Kenneth Koedinger leads the 6th grade course development.

Cognitive Tutor Algebra I is in use in 150 public and private schools, in 14 different states. More than 300 sites are predicted for the 2000-2001 academic year.

2000

Carnegie Learning, Inc. (2002, May). Results from Moore, OK (Cognitive Tutor Research Report OK-01-01). Pittsburgh, PA. (PDF)

Kenneth Koedinger tests the Cognitive Tutor Math 6 course in two Pittsburgh schools during the 2000-2001 school year. The Cognitive Tutor classes consistently outperform students in comparison classes on two assessments in both schools (Koedinger 2002)

Moore Independent School District

During the 2000-2001 school year, Pat Morgan, mathematics coordinator for Oklahoma's Moore Independent School District, evaluates the Cognitive Tutor Algebra I curriculum using a rigorous experimental design involving five junior high schools in the district.

At the end of the school year, students that took the Cognitive Tutor Algebra I course score higher on the ETS Algebra end-of-course exam than students who took a traditional Algebra course. The same result holds true when comparing between students who had the same teacher. Cognitive Tutor students also show higher course grades and more positive attitudes towards mathematics.

The study receives accolades from the Deparment of Education's What Works Clearinghouse, who in 2004, designate the study as one of only two studies--out of over 800 submissions--examining middle school math that had a strong research design and showed positive results.


Carnegie Learning implements LMS, a learner management system that allows for storing student data over the web.

2002

Aleven, V., & Koedinger, K.R. (2002). An effective metacognitive strategy: learning by doing and explaining with a computer-based Cognitive Tutor. Cognitive Science, 26(2), 147-79. (PDF)
Vincent Aleven head-shot

Carnegie Mellon researchers Vincent Aleven and Kenneth Koedinger publish the results of an experiment on the effects of self-explanation on learning in a computer-based Cognitive Tutor. They discover that supporting self-explanation in a Cognitive Tutor during problem solving practice leads to more effective learning that avoids 'shallow heuristics' and promotes 'deeper understanding' of material. The researchers describe the greater understanding in terms of 'more integrated visual and verbal declarative knowledge, used more reflectively, and less shallow procedural knowledge.'

The study is one of the first to compare a tutor that reinforces self-explanation with one that does not. The results of the in vivo experiment are published in the journal Cognitive Science.

The success of the study prompts Carnegie Learning, Inc. to include support for self-explanation in the current version of the Cognitive Tutor for geometry problem-solving.


CTAT logo

Carnegie Mellon researchers start work on the Cognitive Tutor Authoring Tools (CTAT), a suite of authoring tools that aims to make tutor development more affordable by leveraging human-computer interaction (HCI), machine learning, and data mining techniques. In addition, the software will facilitate cognitive modeling for experts and enable cognitive modeling for novices in cognitive science.

2003

Microscope view

Awarded a grant by the Fund for the Improvement of Postsecondary Education (FIPSE), the Genetics Cognitive Tutor project at Carnegie Mellon University completes six tutor units on human pedigree analysis, recombination, and gene mapping. The researchers begin preliminary evaluations of tutor effectiveness and hold a genetics tutor summer workshop with instructors from four universities.

The Genetics Cognitive Tutor is developed primarily by Albert Corbett (Primary Investigator, HCII), Benjamin MacLaren (HCII) and Linda Kauffman (Dept. of Biological Sciences). The group plans to implement 18 lessons so that approximately half of the course taught at CMU is covered at some level.

Learn more about the Genetics Cognitive Tutor at the project's web site.

2004

The Cognitive Tutor Algebra I course is in use in roughly 2000 schools across the United States. The data show that students in the Cognitive Tutor Algebra I course score twice as high on end-of-course open-ended problem-solving tests, and 15% higher on objective tests as students enrolled in a traditional Algebra course. Most participating schools are urban or rural high schools. The researchers estimate that a half million students who have used the tutor, for a total usage time of 20 million hours.

In its second project year, the Genetics Cognitive Tutor project completes evaluations of their six tutor units at four pilot schools. The Project is awarded grants from both the National Science Foundation and the Howard Hughes Medical Institute, funding that enables it to develop tutor curriculum for a full genetics course. A total of 12 lessons are completed.

2005

LearnLab logo

The Pittsburgh Science of Learning Center (PSLC) begins on October 1, 2005. The proposal for the PSLC was submitted to the National Science Foundation's Science of Learning Center program in September of 2003 with Project Investigator Ken Koedinger and Co Project Investigators Kurt VanLehn, David Klahr, and Charles Perfetti. The grant is awarded for $25 million and 5 years. The PSLC Directors are Ken Koedinger at Carnegie Mellon and Kurt VanLehn at the University of Pittsburgh.

The PSLC brought together long-standing strengths at Carnegie Mellon in the fields of Cognitive Science and Human-Computer Interaction, including the Cognitive Tutor project, and at University of Pittsburgh's Learning Research and Development Center in the fields of Education and Cognitive Psychology.

PSLC member groups include Carnegie Learning, Inc.


In the Fall of 2005, Carnegie Mellon University researchers Bruce McLaren, Dave Yaron, and Kenneth Koedinger employ CTAT example-tracing tutors to investigate whether learning can be improved by personalized, informal language and worked examples in intelligent tutoring. The large-scale, web-based Pittsburgh Science of Learning Center (PSLC) study, conducted at the University of British Columbia, tests and expands the capabilities of the Cognitive Tutor Authoring Tools.


Carnegie Learning is awarded a three-year, multi-million dollar contract by the Los Angeles Unified School District (LAUSD) to provide algebra readiness and Algebra I curricula to more than 25,000 middle and high school students.

2006

Carnegie Learning's Bridge to Algebra

Building on its history of user interface innovation, Carnegie Learning releases Bridge to Algebra, an intervention product for middle school students that prominently features usability constraints and novel interaction mechanisms. The product was piloted in ~100 schools in 2005.

Carnegie Learning embarks on a project to create tools for teachers to build curricula. The project is funded by a subaward with "Science Mission to Planet Earth (SMPE): IT-Integrated Coastal Education" project, a grant to Southern University from the National Science Foundation's 'Information Technology Experiences for Students and Teachers' (ITEST) program.

® Carnegie Learning and Cognitive Tutor are registered trademarks of Carnegie Learning, Inc.

Reference List

  1. Aleven, V., & Koedinger, K.R. (2002). An effective metacognitive strategy: learning by doing and explaining with a computer-based Cognitive Tutor. Cognitive Science 26(2), 147-79. (PDF)
  2. Anderson, J. R. (1983). The Architecture of Cognition. Cambridge, MA: Harvard University Press.
  3. Anderson, J. R. (1993). Rules of the Mind. Hillsdale, NJ: Erlbaum.
  4. Anderson, J. R., Boyle, C. F., & Reiser, B. J. (1985). Intelligent tutoring systems. Science, 228, 456-462. (PDF)
  5. Anderson, J. R., Conrad, F. G., & Corbett, A. T. (1989). Skill acquisition and the LISP Tutor. Cognitive Science, 13, 467-506.(PDF)
  6. Anderson, J. R., Corbett, A. T., Koedinger, K., & Pelletier, R. (1995). Cognitive tutors: Lessons learned. The Journal of Learning Sciences, 4, 167-207.
  7. Anderson, J. R., Corbett, A. T., & Reiser, B. J. (1987). Essential LISP. Reading, MA: Addison-Wesley.
  8. Corbett, A.T., & Anderson, J.R. (1995). Knowledge Tracing: Modeling the Acquisition of Procedural Knowledge. User Modeling and User-Adapted Interaction, 4, 253-278.
  9. Corbett, A.T., Koedinger, K.R. & Anderson, J.R. (1997). Intelligent tutoring systems. In M.G. Helander, T.K. Landauer and P. Prabhu (Eds.) Handbook of human computer interaction, 2nd edition. Amsterdam: Elsevier Science.
  10. Corbett, A.T., Koedinger, K.R. & Hadley, W. S. (2002). Cognitive Tutors: From the research classroom to all classrooms. In P. Goodman (Ed.) Technology enhanced learning: Opportunities for change. Mahway, NJ: Lawrence Erlbaum Associates.
  11. Koedinger, K. R. (2002). Toward evidence for instructional design principles: Examples from Cognitive Tutor Math 6. Invited paper in Proceedings of PME-NA XXXIII (the North American Chapter of the International Group for the Psychology of Mathematics Education).
  12. Koedinger, K. R., Anderson, J. R., Hadley, W. H., & Mark, M. A. (1997). Intelligent tutoring goes to school in the big city. International Journal of Artificial Intelligence in Education, 8, 30-43. (PDF)
  13. Ritter, S., Anderson, J. R., Cytrynowicz, M., & Medvedeva, O. (1998). Authoring Content in the PAT Algebra Tutor. Journal of Interactive Media in Education, 98(9). (PDF)
  14. Ritter, S. & Koedinger, K. R. (1996). An architecture for plug-in tutor agents. Journal of Artificial Intelligence in Education, 7(3-4), 315-347. ( PDF)
  15. Sleeman, D. H. & Brown, J. S. (Eds.). (1982). Intelligent Tutoring Systems. New York: Academic Press.