Instructional Technology Research Paper

1. Traditional Computer-Based Instructional Systems

The first generation of computer tutoring systems were developed in the late 1960s. They were essentially a kind of programmed instructions. They mostly presented pieces of information, e.g., some element of the course content to be learned, asked short-answer questions to check whether the student had learned this element correctly, and branched to further presentations depending on the answer given. The answer provided by the student, simply marked by the program as correct or incorrect, was used to determine the next step of the student’s path along the course. These ‘traditional’ systems are often called computer-aided instruction (CAI) or computer-based instruction (CBI).

This technique already permits first steps towards the individualization of instruction, at least in principle. For example some authors try to anticipate every possible incorrect student response and to prespecify branches depending on the supposed cause of every incorrect answer. The main advantage of these traditional CAI programs is that it allows students to learn or to get trained at their own pace and in their own chosen setting (time and place). Their main drawback was the basically linear progression of the instruction. These systems were unable to explicitly take into account the learners’ knowledge, learning style, and preferences, their individual progression, and the underlying cause of their errors.

2. AI-Based Instructional Systems

The evolution beyond traditional computer-assisted instruction toward ITS has to pass three tests of intelligence. First, the subject matter, or domain, must be ‘known’ to the system well enough for it to draw inferences or solve problems by itself. Second, the system must be able to diagnose a learner’s approximation of the knowledge to be acquired. Third, the tutorial strategy or pedagogical knowledge must be intelligent in the sense that different strategies can be implemented to reduce the difference between the knowledge to be learned and the student’s actual knowledge.

The next generation of tutoring systems were based on building knowledge of the subject matter into the program (see Sleeman and Brown 1982 or Wenger 1987 for a detailed documentation and overview). There were different types of these so-called intelligent CAI or ITS. One type coached students working in open-ended learning environments to solve complex problems, such as troubleshooting an electronic circuit, writing a computer program or learning arithmetic skills. Another type attempted to have some kind of a dialog with students, e.g., in the sense of a problem monitor or as a consultant.

The first ITS-type program, Scholar, was developed by Carbonell (1970). Its subject matter was the geography of South America. What was new then was that the system’s knowledge of its domain matter was explicitly represented in a semantic network utilizing AI techniques and methods. The nodes of the network stood for geographical objects and concepts. These objects and concepts were organized in a partial hierarchy with relations like superpart, superconcept, and superattribute. Inferences could be made by propagation of inherited properties via these links. This underlying representation technique gave the system a new quality of competence.

It knew how to solve the tasks given to the student; it no longer needed a data bank of preformulated exercises together with their solutions. Such a system was capable of answering student questions as long as they were domain-related. A further consequence was that the problem-solving process itself could become part of the taught domain. Another new quality of Scholar was its capacity for mixed initiative dialogs based on a simple but robust user interface to correctly interpret the student’s questions and responses and to transmit them to the inference engine. Due to the problems of understanding and generation of natural language, only a few systems of this type were built.

2.1 The Basic Architecture Of ITS

ITS developed since 1970 usually contain four basic components or modules (Sleeman and Brown 1982, Wenger 1987).

(a) An environment (user interface) in which the student works on complex tasks. For instance, it might be a simulated piece of an electronic circuit that the student tries to troubleshoot.

(b) An expert system (domain knowledge base) that knows enough about the problems that the student works on to solve them by itself.

(c) A student modeling component (student model) that compares the student’s behavior with the expert system’s behavior in order both to recognize the student’s current knowledge for solving the problem and determine what elements of correct as well as incorrect knowledge the student is probably using.

(d) A tutorial module (teaching module) that suggests tasks to be solved, responds to the student‘s requests for help, and points out mistakes. Such responses and suggestions are based on a library of flexible usable teaching strategies as well as on the system’s model of the student’s knowledge and plans. In general, these systems try to simulate good human tutors by the integrated use of different knowledge bases. They are domain experts in the subject matter to be learned by the student, they are teaching experts using different tutorial strategies and interventions for supporting the student’s acquisition of knowledge, and they are educational psychological experts in the assessment and evaluation of the student’s domain-related knowledge. Sometimes these ITS are therefore also called knowledge-based tutors (e.g., Murray 1998).

2.2 Student Modeling

The idea of the necessity of some kind of student modeling was around from the beginning. The argumentation is well known. Efficient teaching implies that the communication of the knowledge to be learned is adapted to the cognitive prerequisites and requirements of the learners, especially to their domain knowledge. The representation of domain knowledge implemented in the ITS should reflect the student’s mental representation of this knowledge. In its simplest form, there may be only a list of what the student knows or should know. There may also be some techniques to model the learner’s errors based on incorrect and/or incomplete knowledge. Such techniques include systematic errors, often modeled as bugs, impasses, misgeneralizations or repairs in procedural skills (Brown and Burton 1978, Young and O‘Shea 1981, Sleeman 1984, VanLehn 1990), or as some kind of overlay model (Goldstein 1982). In its most complete state, the student model also contains domain-independent knowledge. In that case, the system also has to know something about knowledge acquisition based on learning processes, therefore referring to different types of students rather than to different domains. Examples of these kinds of student modeling can be found in the work of Anderson and his group (e.g., Anderson et al. 1989, Anderson and Lebiere 1998). With such information, the system is then capable of associating specific tutoring strategies to specific learning strategies.

Thanks to the complex integration of the various system modules, such an ITS is able to guide the learner as evolving to become an expert, at least in principle. Through the presentation of specific questions, tasks, or explanation requests, certain comprehension errors can be detected. Moreover, thanks to the knowledge it might thus have (or even gain) about students and their specific knowledge, problem-solving strategies, and learning mechanisms, it would be able to support a student’s learning and knowledge acquisition in quite an individualized manner.

2.3 Applications

In the last half of the 1980s many ITS were developed, despite the immense costs and the complexity involved. But they were not only in use for academic purposes. ITS have moved out of the laboratory and into classrooms and workplaces, where some have proven to be highly effective as learning aides. For example, students working with an Air Force electronics trouble-shooting tutor for only 20 hours gained proficiency equivalent to that of trainees with 48 months of on-the-job experience (Lesgold et al. 1992). In another example, students using the LISP tutor (Anderson 1990) completed programming exercises in 30 percent less time than those receiving traditional classroom instruction and scored 43 percent higher on the final exam.

Although ITS were becoming more common and proving to be increasingly effective, each one had to be built from scratch at a great expense. There was also no standardization with regard to the software and hardware used. Authoring tools for ITS have not been available until recently (Murray 1998). Although ITS show their potential effectiveness for educational purposes, there are other applications as well (cf. VanLehn 1999). A common practice is to build an open learning environment without the underlying expert system, without any student model, and without any tutorial module. The environment enables student activities that stimulate learning and may be impossible to conduct in the real world. For instance, an environment might allow students to conduct simulated physics experiments on worlds where gravity is reduced, absent, or even negative. Such environments are called interactive learning environments, micro-worlds, or virtual laboratories. Another new trend is to use networking to allow several students to work together in the same environment.

2.4 ITS From A Cognitive Science Point Of View

One of the lessons learned in the development of an ITS is that it requires the knowledge and collaboration of different disciplines. But empirical cognitive studies are nevertheless particularly important in developing ITS. Developing the expert module of a tutoring system requires studying experts as they solve problems in order to understand and formalize their knowledge. Developing an effective teaching module requires understanding how students learn so that the tutor’s comments will prompt students to construct their own understanding of the subject matter. An overly critical or didactic tutor may do more harm than good. A good first step in developing an application is to study the behavior of expert human tutors in order to see how they increase the motivation and learning of students. Developing the student model one needs to understand how students acquire new knowledge and how they apply their knowlege to solve problems. A good way to achieve this is a detailed analysis of the subject matter, its main concepts, and their relational structure and of important procedural skills to solve problems and tasks within the domain. An empirical study of students’ errors and most typical misconceptions might also be helpful.

2.5 Conclusions

Many ITS have been developed since the mid-1970s and the requirements to be solved during the development of an ITS generate many fruitful interactions between academic research and application oriented research with regard to topics like knowledge representation, computerized assessment of knowledge, learning and knowledge acquisition, expertise research. Nevertheless from a cognitive science perspective some difficult problems have not been sufficiently solved until now: First, the requirement of a quick and reliable in-time and on-line diagnosis of knowledge. Even within very complex and large behavioral spaces the system’s feedback must be given immediately. Otherwise the user will not accept the system for a long time. The dynamic assessment of valid indicators and their interpretation with respect to the student’s knowledge acquisition processes, goals, and problem-solving strategies still includes many open questions. Second, the systematic consideration of motivational and emotional aspects is an important question of research. Third the transitions between the content of the student model and the tutorial knowledge (system feedback, kind of help, strategic-didactical decisions, content and style of information presentation) needs more controlled empirical evidence. ITS are still an active topic of many research activities, not only because of the many problems to be solved but because of the ongoing growing requirement of computerized training and education.

3. Current Trends

The present research tends to be more and more interdisciplinary, slowly but surely making more effective the collaboration between course designers, cognitive scientists, computer scientists, and educators/psychologists/teachers. With regard to the uses of new technologies of information and communication in education, the main interest lies in exploiting Internet possibilities, especially the Web. Some important trends are outlined below: distance learning and teletutoring, collaborative learning, and Web-based learning. Finally, an active area surpasses them all: the evaluation of educational systems.

3.1 Distance Learning And Teletutoring

Distance learning and teletutoring are extremely active areas of research. Many courses are now being developed on the Web. Distance learning means that the learning material is offered as computer-based training (CBT) or as Web-based training (WBT) that can be used by students in their own learning environment at their place of work or at home. The access and interaction is completely self-organized. Teletutoring means that there is a teacher tutor who primarily takes the role of a moderator. This teacher supports ongoing learning activities in individuals but also in groups. Communication is the most important concept. One well-known variant of teletutoring is the online seminar in which a classical teaching form is virtually imitated. Some researchers are also investigating ways to use the Web meaningfully and effectively to develop new types of pedagogical paradigms (e.g., see the recent conferences on AI in education).

3.2 Collaborative Learning

In cognitive science, a whole area of research is striving to enhance computer supported collaborative cooperative learning (CSCL). The basic goal is to develop learning environments in which groups of students can communicate with each other, interact with each other, and learn and solve problems together. The cooperation might take place at the same time or at different times (synchronous asynchronous CSCL) and at the same place or at different places (local or distributed CSCL). Probably this field will play a much more prominent role in the future. It integrates the fast growing potential of new information technology (e.g., multimedia, interactivity), new learning theories (situated learning, learning as social activity, distributed learning) and the perspective of knowledge acquisition based on explanation, argumentation, and discourse. A good overview of ongoing research in this field can be found in the various contributions in Koschmann (1996) and Dillenbourg (1999).

3.3 Web-Based Learning

Since the Web is the major result of the present technological progress, it can be expected that it will have a steadily growing influence on computerized training and education (e.g., Eisenstadt and Vincent 1998, Forsyth 1998, Bruns and Gajewski 2000). However, since it is only a technological evolution, it will probably act more as a catalyst than as a conceptual or theoretical leading engine. At any rate, this domain will undergo two main positive changes. The first change is the expected gain in the pedagogical capabilities of educational systems. However, this gain will be effective only through a better intercommunication and a better collaboration between educational and psychological researchers, cognitive scientists, and computer scientists. But actually to benefit from this evolution, it will be necessary to overcome several traditional but strong barriers, essentially of a cultural or administrative nature.

The second major change stems from the fast growing communication and multimedia capabilities of information technology. Owing to such capabilities, students or trainees will undoubtedly become more and more autonomous, and will feel more and more responsible for their own learning or training. The responsibility of each subject for his or her own education and training will gain increasing importance. To adapt to this evolution, teachers of all types, including university professors, will have to modify their roles by becoming rather knowledgeable guides than knowledge providers, and by becoming more student-centered. Students will be more and more engaged in their own goals, searching the Web for the information they are interested in. The drawback of these changes is that all participants in this evolution— researchers and developers, but also educators and school administrators—will have to adapt to that evolution adequately, whether they like it or not. Information technology will evolve anyway, and we should actively use the new possibilities of this evolution.

3.4 Evaluation Of Educational Systems

Finally, with the variety and number of educational programs and prototypes developed, and with the variety of theories, methods, and techniques put to use in the field of computerized education, the need to better understand the pros and cons of that field gets increasingly stronger. Therefore the task of evaluating educational systems, and the methods and processes used to evaluate them, will receive growing interest.

3.5 Concluding Remarks

At the turn of the twenty-first century there are no up-to-date textbooks on intelligent tutoring. Sleeman and Brown (1982), Polson and Richardson (1988), Self (1988), and Wenger (1987) cover the basic ideas as well as the early systems. Lelouche (1999) gives a survey of the history of ITS and also a general outlook on the present trends of research. Self (1999) reviews the evolution of the ITS research, namely to make computationally precise and explicit the forms of educational, psychological, and social knowledge necessary for the design and implementation of computer-based learning systems. The latest work generally appears first in the proceedings of the conference on Intelligent Tutoring Systems or in the proceedings of the conference on AI in Education. The most relevant journals for this kind of work include the International Journal of AI in Education (http://cbl.leeds.ac.uk/ijaied/), the Journal of the Learning Sciences (Erlbaum), and Interactive Learning Environments (Ablex). There are also many Web-based educational systems available covering quite different subject matters. To get an impression of what is going on in the field, it might be helpful simply to use some of these systems.

Bibliography:

1. Anderson J R 1990 Analysis of student performance with the LISP tutor. In: Frederiksen N, Glaser R, Lesgold A M, Shafto M (eds.) Diagnostic Monitoring of Skill and Knowledge Acquisition. Erlbaum, Hillsdale, NJ, pp. 27–50
2. Anderson J R, Conrad F G, Corbett A T 1989 Skill acquisition and the LISP tutor. Cognitive Science 13: 467–505
3. Anderson J R, Lebiere C 1998 The Atomic Components of Thought. Erlbaum, Hillsdale, NJ
4. Brown J S, Burton R R 1978 Diagnostic models for procedural bugs in basic mathematical skills. Cognitive Science 2: 155–91
5. Bruns B, Gajewski P 2000 Multimediales Lernen im Netz. Springer, Berlin
6. Carbonell J R 1970 AI in CAI: An artificial intelligence approach to computer assisted instruction. IEEE Transactions on Man–Machine Systems 11(Dec): 190–202
7. Dillenbourg P 1999 Collaborative Learning: Cognitive and Computational Approaches. Pergamon, Amsterdam
8. Eisenstadt M, Vincent T 1998 The Knowledge Web: Learning and Collaborating on the Net. Kogan Page, London
9. Forsyth I 1998 Teaching and Learning Materials and the Internet. Kogan Page, London
10. Goldstein I P 1982 The genetic graph: A representation for the evolution of procedural knowledge. In: Sleeman D, Brown J S (eds.) Intelligent Tutoring Systems. Academic Press, New York, pp. 51–77
11. Koschmann T 1996 CSCL: Theory and Practise of an Emerging Paradigm. Erlbaum, Hillsdale, NJ
12. Lelouche R 1999 Intelligent tutoring systems from birth to now. KI Kunstliche Intelligenz 4/99: 5–11
13. Lesgold A, Lajoie S, Bunzo M, Eggan G 1992 Sherlock: A coached practise environment for an electronics troubleshooting job. In: Larkin J, Chabay R W (eds.) Computer Assisted Instruction and Intelligent Tutoring Systems. Erlbaum, Hillsdale, NJ, pp. 20–38
14. Mandl H, Lesgold A 1988 Learning Issues for Intelligent Tutoring Systems. Springer, New York
15. Murray T 1998 Authoring knowledge-based tutors: Tools for content, instructional strategy, student model, and interface design. Journal of the Learning Sciences 7: 5–64
16. Polson M C, Richardson J J 1988 Foundations of Intelligent Tutoring Systems. Erlbaum, Hillsdale, NJ
17. Self J 1988 Artificial Intelligence and Human Learning: Intelligent Computer-aided Instruction. Chapman and Hall, London
18. Self J 1999 The defining characteristics of intelligent tutoring systems research: ITSs care, precisely. International Journal of AI in Education 10: 350–64
19. Sleeman D 1984 An attempt to understand student’s understanding of basic algebra. Cognitive Science 8: 387–412
20. Sleeman D, Brown J S 1982 Intelligent Tutoring Systems. Academic Press, New York
21. VanLehn K 1990 Mind Bugs: The Origins of Procedural Misconceptions. MIT Press, Cambridge, MA
22. VanLehn K 1999 AI and education. In: Wilson R A, Keil F C (eds.) The MIT Encyclopedia of the Cognitive Sciences. MIT Press, Cambridge, MA, pp. 7–10
23. Wenger E 1987 Artificial Intelligence and Tutoring Systems. Morgan Kaufmann, San Mateo, CA
24. Young R M, O’Shea T 1981 Errors in children’s subtraction. Cognitive Science 5: 153–77