Background

 < Day Day Up > 



Before the details of intelligent tutoring systems are covered, it is important to discuss intelligent agents, because an intelligent tutoring system is considered to be an intelligent agent system. Wooldridge and Jennings (1995) defined an intelligent agent as a computer system that is capable of flexible autonomous action in order to meet its design objectives. The intelligent agent in an intelligent tutoring system performs on behalf of the tutor to help learners achieve learning outcomes and to prescribe teaching strategies based on learners’ profiles in the student model and content in the domain module. As the agent interacts with the learner, it gains more experience by learning about the learner. The expertise in the intelligent tutoring system intelligent agent should allow the agent to help learners achieve the learning outcome without human intervention. The intelligent agent should anticipate learners’ responses and respond immediately to take corrective action or to present the next learning intervention based on learners’ characteristics and styles to maximize learning benefits. In other words, the intelligent agent should form dynamic profiles of the learner and work ahead of the learner by guiding the learner in what to do next in the learning process. The intelligent agent system should behave like an expert tutor by interacting with the different components in the intelligent tutoring system to assemble the expertise required to help learners achieve the learning outcome.

An intelligent tutoring system is one type of expert system (Sowa, 1984). Boose (1986) defined an expert system as a knowledge-based reasoning system that captures and replicates the expertise of human experts. Kearsley (1987) defined an intelligent tutoring system as application of artificial intelligence techniques to teach students. Sleeman and Brown (1982) defined an intelligent tutoring system as a program that uses artificial intelligence techniques for representing knowledge and carrying on an interaction with a student. According to Sleeman and Brown (1982), an intelligent tutoring system must have its own problem-solving expertise, its own diagnostic or student modeling capabilities, and its own explanatory capabilities. It must know when to interpret a student’s problem-solving activity, what to say, and how best to say it. Hence, an intelligent tutoring system closely resembles the process when a student and teacher interact in a one-to-one situation (Tennyson & Park, 1987).

In distributed learning, students can be in any location to take courses, as long as they have access to communication technology with which to access the course. Distributed learning could be either synchronous or asynchronous. In synchronous learning, the learning is in real time, where the learner interacts and receives information as needed. In asynchronous learning, there is a delay in the interaction between the system and the student. The information in this chapter is related to synchronous learning, where students interact with the intelligent tutoring system in real time and receive feedback as they interact with the system.

An effective intelligent tutoring system should simulate what good human tutors do when tutoring in a one-to-one situation. Bloom (1984) mentioned that educators should try to replicate the same strategies used by students and teachers in a one-to-one environment to other teaching situations, because the one-to-one tutoring environment is ideal for learning. Woolf (1996) described a Cardiac and an Engineering intelligent tutor that include strategies to help achieve the two-sigma effect described by Bloom. Anderson et al. (1985) conducted a study where two groups of students were given the same lectures, but one group used an intelligent tutoring system tutor for the exercises. They found that the tutored students spent 30% less time on the problems than those working on their own. The tutored group also scored 43% better on the post- test. Anderson et al. (1985) noted that the presence of the tutor is more significant for the performance of low achievers. The goal of an intelligent tutoring system is to replicate the one-to-one interaction between a tutor and a learner. This should include all of the expertise (interface with the learner, content, a model of the student, and pedagogical) that is involved in the tutoring process.

Dede (1986) mentioned that an intelligent tutor is a stand-alone device, which can initiate interactions with its user and incorporates all the knowledge needed to teach a subject. However, to build a good intelligent tutoring system for distributed-learning systems, the expertise has to be elicited from experts in the domain. Acquiring sufficient and correct teaching expertise is a major problem for builders of intelligent tutoring systems (Woolf & Cunningham, 1987). Most expert systems projects claimed that the knowledge elicitation process is the most complex and time consuming in the development of expert systems (Berry, 1987; Olson & Reuter, 1987). Some reasons given are as follows:

  1. Experts in the field are not able to articulate and make explicit their expertise.

  2. Expertise from experts tend to be of a high level, and this cannot be used to tutor the learner.

  3. Experts obtain their expertise through an implicit learning process that cannot be made explicit (Berry, 1987).

The more experienced one becomes in a knowledge domain, the more difficult it is to make the knowledge explicit (Berry, 1987). Experts possess compiled knowledge, which exists in large chunks accumulated over the years, and this knowledge is difficult to elicit. Also, the knowledge elicitation process may influence the quality and quantity of expertise extracted.

Intelligent tutoring systems require an interdisciplinary team to design and develop. In addition to requiring domain and coding knowledge, it requires educators and psychologists to specify the instructional strategies and pedagogical model to incorporate into the system. Because conventional educational research has focused on group instruction, little is known about the same individual learning characteristics vital in developing the student model and pedagogical modules for intelligent tutoring systems (Dede, 1986). One such learner characteristic is learning style. Ally and Fahy (2002) found that students with different learning styles prefer different pedagogical support when learning in a one-to-one distance education environment. Pedagogical expertise of a tutor in a one-to-one situation is the least understood and does not exist in an explicit form to be included in an intelligent tutoring system (Ohlsson, 1987).

Some of the intelligent tutoring systems have been developed to explore the capabilities of artificial intelligence techniques in the instructional process rather than to build an effective instructional system (Park, Perez, & Seidel, 1987). The next generation of intelligent tutoring systems should be concerned with instructional and pedagogical issues rather than computer science or artificial intelligence issues, such as specific programming techniques, software architecture, etc. (Park, Perez, & Seidel, 1987). The goal of an intelligent tutoring system is to replicate the one-to-one interaction between a tutor and a learner. This should include all of the expertise (interface, content, a model of the student, and pedagogical) that is involved in the tutoring process.

An intelligent tutoring system does not act on the basis of pre-entered questions, anticipated answers, prespecified branches, and the knowledge accumulated when a student learns (Tennyson & Park, 1987). An intelligent tutoring system should have domain expertise, it should build a model of the learner, and tutor the learner based on the learner’s model. It should behave as a tutor does in a one-to-one situation. According to Woolf (1987), an intelligent tutoring system should reason about a student’s knowledge, monitor a student’s solutions, and adapt the teaching strategy to the student’s individual learning pattern. The intelligent tutoring system should be able to conduct its own learner analysis and continually improve as it interacts with learners to become a more effective tutoring system. After many learning cycles, the intelligent tutoring systems should be comprehensive enough to meet the needs of learners with different learning styles and preferences. For example, an intelligent tutoring system could monitor strategies that different learning styles use successfully and build a best practice database of effective learning strategies for different learning styles. The intelligent tutoring systems could also track common errors made by students and prescribe strategies to students to prevent them from making these errors.

Intelligent tutoring systems usually consist of a domain module, a student model, and a tutorial or pedagogical module (Ong & Ramachandran, 2003; Thomas, 2003; Park, Perez, & Seidel, 1987; Dede, 1986). Not much mention is made of an interface module for intelligent tutoring systems. An effective interface module is critical for allowing the student to interact from a distance with the different components of the intelligent tutoring systems and for getting the learner’s attention and engaging the learner in the learning process. Figure 1 shows the components of an intelligent tutoring system for distributed learning. As shown in the figure, for an intelligent tutoring system to be effective, the components must interact with each other by passing information between each other and learning from each other.

click to expand
Figure 1: Components of an intelligent tutoring system



 < Day Day Up > 



Designing Distributed Environments with Intelligent Software Agents
Designing Distributed Learning Environments with Intelligent Software Agents
ISBN: 1591405009
EAN: 2147483647
Year: 2003
Pages: 121

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net