For years computers have been used as a tool in tutoring students in various subjects. Two major concepts are computer based training (CBT) and computer assisted instruction (CAI). Neither of these approaches delivers quality individualized instruction, instead the decisions about how to move a student through the material were script-like. For example “if question 21 is answered correctly, go to question 54 else go to question 32”. The abilities of the learner were not taken into account.
CBT and CAI do not provide the same kind of attention that students would receive from a live tutor. For a computer based educational learning system to provide this attention, the learning system must reason about the domain and the learner. This limitation has prompted research in the field of Intelligent Tutoring Systems (ITS) which is a subset of Artificial Intelligence.
There are five interdependent components of an ITS:
1. Student Module
2. Pedagogical module
3. Domain Knowledge Module
4. Communications Module
5. Expert Module
The student module stores information specific to and individual student. This module stores information on how well the student is doing in the exercise. It also provides information to the pedagogical module.
The pedagogical module provides a model of the teaching process. For example, this module makes the decision of when to review a topic or when to move to another topic. This decision, which reflects the differing needs of the student, is made in conjunction with the input of the student module.
The domain knowledge module contains the knowledge base of the particular subject that is being presented, Biology 103 for example. Constructing the domain knowledge module requires significant knowledge engineering to represent a domain so that the other parts of the ITS can access it.
The communications module provides for dialog between the tutor and the student. This module also controls how the information is presented on the screen. It can be set up to match the individual student’s level and proficiency.
The expert module is similar to the domain module in that it contains subject
specific data. However, the expert module also has the ability to demonstrate
how someone skilled in the subject area represents the knowledge or how he/she
might solve a problem in that domain. By using an expert model the tutor can
compare the student’s answer or problem solving method with that of the
expert.
There are two basic categories of ITS:
1. Abstraction refers to simulations of real life tasks which afford training in an area without the dangers or costs of real tasks. For example flight simulators train pilots to land a 747 without the human and equipment costs associated with this training taking place on a real airplane. Another example is the Advanced Cardiac Life Support Tutor which simulates to a student the treatment of heart attack patients.
2. Knowledge type of the instruction refers to an approach to tutoring based on the type of knowledge being taught. The most common type of ITS teaches procedural skills or in other words teaching the student how to perform a particular task. Systems thus designed are referred to as cognitive tutors. An example of a cognitive tutor is SHERLOCK which is a tutor of electronic troubleshooting.
To date, the most successful ITS have focused on the areas of mathematics and science, where it is both easier to design ITS and measure learning outcomes and improvements. Studies have been conducted in numerous high school algebra classes where some classes were exposed to an ITS and others were taught by human teachers. In several instances, students who worked with ITS scored almost a full letter grade higher than did students being taught by a live instructor.
New learning technologies in their attempt to automate learning have in some cases developed easier methods of performing the skills they attempt to teach. For example technology can now do symbol manipulation algebra and spelling correction making the learning of these skills not as important. This effect makes the learning of higher order skills more important. The evolution in required skills drives the evolution of artificial intelligence systems to teach them. One response to this trend is interactive learning environments (ILEs) which try to foster a deeper conceptual understanding of ideas that are usually taught as simple procedures.
In the short run the most effective uses of AI in education probably will not
happen quickly or without careful planning and policy making. However ITS that
focus on well defined learning goals and that can easily and cheaply be implemented
into the classroom will yield the most statistically significant improvements
in student outcomes.
Links to selected References and Resources
http://www.acm.org/crossroads/xrds3-1/aied.html
http://www.cs.berkeley.edu/~russell/ai.html