Back More
Salem Press

Table of Contents

Principles of Computer Science

Intelligent Tutoring System

by Donald Franceschetti, PhD

Field of Study

Tutoring systems

Abstract

An intelligent tutoring system is one that modifies its interaction with the person being tutored bases on the tutee’s performance. Just as an experienced teacher will adopt his information presentation to the characteristics of the learner so an intelligent tutoring system with adjust the rate of presentation to the learners prior knowledge, level of interest, and general intelligence.

Principal Terms

  • Andes: An intelligent tutoring system for a mathematically based course in introductory physics developed for the U. S. Naval Academy.

  • AutoTutor: a program, which tutors using ordinary language and grades student responses using some form of latent semantic analysis.

  • conceptual physics: Physics taught with minimum emphasis on mathematical methods. The physics teaching community was surprised by the discovery of Eric Mazur that students could score very well on course exams and yet have very little understanding or ability to explain their understanding of the problems they had just solved.

  • cognitive load (Sweller): the number of slots in short term memory required to solve problems. When then information presented in a text exceeds the cognitive capacity of the students, little learning will occur.

  • latent semantic analysis (LSA): a method to summarize the content of a text corpus in terms of correlations between words, allowing for an evaluation of a student’s answers to questions in comparison to a subject matter corpus. Generally LSA in involves finding a 300-500 dimensional vector space representation of the text.

  • corpus: a body of text that the student is expected to be able to answer question about.

  • user modeling: a mathematical model of student behavior while learning.

The Need for Intelligent Tutoring Systems (ITS’s)

While there is no restriction of intelligent tutoring systems to topics of military importance, probably the greatest stimulus to their development comes from the armed forces. Each year several hundred thousand individuals begin advanced individual training as members of the United States’ Armed Forces. The majority of these will serve out their initial enlistment period then go on to alternative civilian employment. As a result the will be a tremendous need for basic technical training, probably a greater need than can be met with the available teaching manpower at any time in the near future.

Special Problem with Mathematics-Based Subjects

Though traditional instruction in engineering, physics, and chemistry have focused on students’ problem solving ability, students can circumvent the process by memorizing a limited set of problem templates. Indeed, the need to confirm that students are actually developing their understanding of the material is always a background consideration.

Several computer programs have been developed in recent years as intelligent tutoring systems. Among the major sponsors of research in this area are the National Science Foundation, the Institute of Education Sciences and the Office of Naval Research. These researchers are slowly converging upon a general framework for intelligent tutoring systems.

Major Components of an Intelligent Tutoring System

Here we adopt the Generalized Intelligent Framework for Tutoring (GIFT), which has grown out of recent research at the U. S. Army Research Laboratory. It is of course highly desirable that ITS’s be modular in design. So that an ITS developed for say organic chemistry could have some interchangeable parts with ITS’s for Physical Chemistry or Physics, to minimize the total effort required in in research and development and to maximize transfer of learning from one ITS to another. It is generally accepted that ITS’S will have four major components: (1) a domain model (2) a learner model (3) a pedagogical model and (4) a tutor-user interface. The GIFT model assumes in addition a sensor module, which provides a non-verbal record of the student’s psychological state while learning.

The domain model involves the structure of the knowledge to be developed in the student. It normally contains the knowledge that experts bring to the subject along with an awareness of pitfalls and popular misconceptions.

The learner model consists of the succession of psychological states that the learner is expected to go through in the course of learning. It can be viewed as an overlay of the domain model, which will change over the course of learning.

The pedagogical model takes the domain and learner models as input and selects the next move that the tutor will take. In mixed initiative systems, learners may take actions, or ask for help as well.

The tutor – user interface records and interprets the learner’s responses obtained through various inputs: mouse clicks. typed-in answers, eye tracking, and possibly postural data and keeps a record as the tutoring progresses.

Tracking Student Progress

Here are several examples of learner modeling, used in contemporary ITS’s:

Knowledge tracing: If the knowledge to be conveyed is contained in a set of production rules then skillometer can be used to visually display the students progress. Step by step knowledge tracing is incorporated in a number of tutoring programs developed in the Pittsburgh Science of Learning Center.

Constraint bases modeling: A relevant satisfied state constraint corresponds to an aspect of the problem solution. Learner modeling is tracked by which constraints are followed by students in solving problems. Successful constraint based tutors include the structured query language (SQL) tutor.

Knowledge space models underlies the very successful Assessment and Learning in Knowledge Spaces (ALEKS) mathematics tutor now widely used on college campuses to deal with students have not studied mathematics in recent semesters. The knowledge space is a large number of possible knowledge states of each mathematical field. As the topics are reviewed the student builds a coherent picture if the field as a whole.

Expectation and misconception tailored dialog. These are particularly important in fields like physics where terms from common usage, like force and energy have specific restricted meanings which students must understand before progress is possible.

Bibliography

1 

Mandl, H., and A. Lesgold, eds. Learning Issues for Intelligent Tutoring Systems. New York: Springer, 1988. Print.

2 

Stein, N. L., and S.W. Raudenbush, eds. Developmental Cognitive Science Goes to School. New York: Routledge, 2011. Print

3 

Sottilare, R., Graesser, A., Hu, X., and Holden, H. (Eds.). (2013). Design Recommendations for Intelligent Tutoring Systems: Volume 1 - Learner Modeling. Orlando, FL: U.S. Army Research Laboratory. ISBN 978-0-9893923-0-3. Available at: https://gifttutoring.org/documents/42

Citation Types

Type
Format
MLA 9th
Franceschetti, Donald. "Intelligent Tutoring System." Principles of Computer Science, edited by Donald R. Franceschetti, Salem Press, 2016. Salem Online, online.salempress.com/articleDetails.do?articleName=POCS_0070.
APA 7th
Franceschetti, D. (2016). Intelligent Tutoring System. In D. Franceschetti (Ed.), Principles of Computer Science. Salem Press. online.salempress.com.
CMOS 17th
Franceschetti, Donald. "Intelligent Tutoring System." Edited by Donald R. Franceschetti. Principles of Computer Science. Hackensack: Salem Press, 2016. Accessed September 18, 2025. online.salempress.com.