Chapter 2 Computer Based Learning Systems
2.3 CBL Systems – An Overview
Using human-made devices for learning is not a new idea. It has its roots in the 1920s when Pressey (1926) used his mechanical teaching machine for MCQ tests. This device was based on the theory of connectionism (Thorndike 1913), which represents the stimulus-response model of behavioural psychology. Three decades later, in the 1950s, Skinner used the idea of teaching machines, which modelled his operant conditioning theory (Skinner 1954), as a basis to create an instructional methodology known as
programmed instruction (Skinner 1958).
2.3.1 From CAI to ICAI
With the introduction of computers, the early computer-based learning systems that originated from programmed instruction were called linear programs. These early linear
programs were monotonous and boring until (Crowder 1959) introduced multiple paths in programmed instruction to alleviate their rigidity. The next leap in CBL evolution was the development of generative systems, which could generate appropriate problems tailored
to the ability of the learner (Uhr 1969). In general, the domain knowledge in these initial systems was not separable from the instruction code. Numerous systems were developed on this basis until Carbonell (1970) separated knowledge from control code. Inspired by contemporary AI research (Quillian 1968), he used the semantic net for knowledge
representation instead of pages of texts. Carbonell also included a Learner model in his system. The early systems are called CAI, and the new systems are have an added ‘I’, in the abbreviation - ICAI (Carbonell1970).
2.3.2 From ICAI to ITS
Triggered by the trend pioneered by Carbonell, several other researchers started to investigate the possibilities of incorporating some other AI techniques into CBL systems (e.g. planning, plan-recognition, natural language processing, etc.). Principally, all of these attempts had the aim of making the systems intelligent; that is, able to predict the status of the learner in order to select the most suitable pedagogical action, and to mutually interact with the learner in a user-friendly manner. The resultant systems are called ITS or AI-ED systems.
A generic ITS system consists of four key components; domain model, Learner model, teacher model and interface unit2 (Hartley
et al. 1973). Figure 2.2 describes the
functionality of a typical ITS system (Shute et. al. 1995). The domain model includes
subject knowledge. It is not just a sequence of text pages- it may contain declarative as well as procedural knowledge (as facts and rules), or it may include a rich semantic net. Additionally, domain models may include reasoning mechanism.
The Learner model is the most debated functional unit in ITS literature. Ideally, it should be able to assess the knowledge level of students during their learning process. However, the accuracy of this estimate is limited. Based on this estimation, the teacher model decides the level of pedagogical actions such as feedback (if any) and curriculum
2 An ideal ITS needs to have at least four intelligent components; subject expert, teaching expert,
accurate Learner model and an efficient natural language processing interface. If we have a good subject expert system the subject can be totally automated; and therefore, we do not need to learn this subject. This situation is referred to as ‘catch-22’ (McArthur et al. 1999)
Figure 2.2 Intelligent Tutoring Systems (after (Shute et al 1995)) Start Generate Problem Present Problem Compare Solutions Present Feedback Student Solution Update Learning Indicators Computer Solution Update Student Skills Model Tutor Curriculum Learner model Domain Expert Bug Library Student Advisor
Boxes: program decisions & actions Ellipses: program knowledge bases Bold Ellipses: Core ITS components
(Shute et. al. 1995). The interface unit may be a rich visual interactive environment.
Obviously, natural language based interfaces could closely resemble a human teacher. More discussions on Learner model and pedagogical action selection process are included in Sections 2.8 and 2.9 respectively. Section 2.10 discuss some issues related to using a particular formative assessment format (Multiple Question Test) in ITS.
2.3.3 From ITS to ILE
Many AI enthusiasts initially thought that creating intelligent tutors to replace human tutors would be an easy task. As researchers gradually realized this was not the case, they concentrated on the practical application of AI techniques to learning systems. Self noted this situation as
“A great deal of opprobrium was being heaped upon ITS research from all directions including from within itself, for many leading figures (such as Brown, Wenger, Clancey, Sleeman and Soloway) considered that AIED research was misguided, relying as it appeared to do on out-moded philosophies of knowledge and learning” (Self 1999, p. 354).
In the midst of many such critiques, another architecture called ILE emerged (McArthur
et al. 1999). Basically, the ILE is used to denote all the types of CBL systems that
incorporate active learning features and allow learners to explore their environment
freely. This type of system may not use AI totally; and therefore, it may not employ a Learner model and explicit knowledge representation.
Discovery World
Some researchers believed that computer programming might be used as a basis for learning all the plan-based studies (Papert 1972). However, the attempts to use computer programming in this manner failed (Palumbo 1990). Nevertheless, this thinking paved the way for a special kind of ILE called the micro-world or discovery world. This type of
system, in general, offers a limited simulation environment within which the learner may engage in constructing some artefacts or perform certain experiments by changing parameters, and this, in turn, enforces active learning (Papert 1972). Computers are now
capable of providing elegant, efficient, aesthetic and interactive environments. A carefully designed environment may be capable of engaging even an unenthusiastic student in active learning activities. Nevertheless, many critics argue that this type of learning system is suitable for only the kind of learners who like to learn independently (Shute et al. 1995).
2.3.4 From ILE to Bi-modus Learning Environment
Critics of ITSs complain that they are less learner-controlled, and the Learner model is not usually robust and efficient, whereas the critics of ILEs complain about lack of learner support (Shute et al 1995). A realistic approach that lies on the midway between the
above two positions comprises limited intelligent Learner modelling using AI techniques to guide learners while they explore or experiment with the environment. These systems provide facilities for an active learning environment with unobtrusive guidance to the learner. Many names have been used for slightly different systems that are based on this ideology: the list includes, Guided-Discovery Environment (Elsom-Cook 1990), Mixed- Initiative Systems (McArthur et al. 1999), Bi-modus Learning Environment (Otsuki
1993) and Intelligent Learning Environment (Shute et al. 1990a).
2.3.5 Collaborative Learning Systems
Another important type of CBL system that incorporates social learning features is called a Computer Supported Collaborative Learning system (Koschmann 1996). Human beings
are social animals. We learn many things from the environment. The philosophy behind Collaborative Learning systems is that learning is a social process (Lipponen et al. 2004).
The so-called solo CBL systems are designed for individual learners. Thinking aloud is
the basis for social interaction. Computers and networks are now powerful enough to provide rich collaborative learning environments. At the lower level, email may be used as a collaborative tool. The capacity of the Internet, web, LAN and other platforms for collaborative learning is being investigated constantly. Co-operative learning and collaborative learning are different. In co-operative learning, each member is responsible for a pre-determined portion, whereas in collaboration, all the members work together on a shared commitment in a coordinated manner. Lehtinen et al. (1999), in their influential
The term e-Learning system is being used to refer to a wide range of learning systems
including web-based systems. Laurillard defines it as “the use of any of the new technologies or applications in the service of learning or learner support”(Laurillard 2006, p.72). She is
enthusiastic about the potential of the learning environments (all e-Learning practices, in general) for active learning. She states “E-Learning could be a highly disruptive technology for education – if we allow it to be. We should do, because it serves the very paradigm shift that educators have been arguing for throughout the last century” (ibid, p.
73). She also worries that “despite the fabulous advance of the technology, there are surprisingly few real-time interactive simulation games in education” (ibid, p. 84).
Obviously, without some sort of centralised initiatives, it is impossible for individual institutions to create a highly sophisticated learning environment for a particular course.