Instruction Design on the Web

Introduction

This research focus is directed at student modeling and how it relates to Internet course delivery at Athabasca University. This paper is a summary of readings done to provide a background and describe a starting point for the research. It is necessary to first describe student modeling in how it is defined and structured. Understanding the student's knowledge base, as well as the structure of the instructional knowledge base is crucial to student modeling.

As Web-based learning systems are becoming common place, how the knowledge domain is structured, in the design of Web-based instructions has been extensively researched. This research has looked at how constructivist, or student-driven learning, and cognitive models of learning on the Internet are defined. These issues are described in this paper, as well as how student modeling can be used in conjunction with instructions systems on the Internet.

Student Modeling

Student modeling, as the model of a learner, represents the computer system's belief about the learner's knowledge. It is generally used in connection with applications computer-based instructional systems. The Alberta Research Council's Student Modeling study (1995), reported that in order to allow instruction to be individually designed, it is first necessary to capture the student's understanding of the subject. With this information, the difficulty of material, and any necessary remediation can be controlled within the instructional system.

Building a student model involves defining; the "who", or the degree of specialization in defining who is modelled, and what the learner history is; the "what", or the goals, plans, attitudes, capabilites, knowledge, and beliefs of the learner; the "how" the model is to be acquired and maintained; and the "why" , including to whether to elicit information from the learner, to give assistance to the learner, to provide feedback to the learner, or to interpret learner behavior.

In maintaining the student model, the factors that need to be considered include the fact that students do not perform consistantly, forget information randomly, and then display large leaps in understanding. For an intelligent instructional system to be tailored to the student, the student model is the essential component in individualized learning. It is the student model that builds and maintains the system's understanding of the student.

The knowledge base of the student model takes both domain and pedagogical knowledge into account. The method to elicit a student model is closely tied to the approach. In the overlay approach, first the expert domain is modelled as a set of correct production rules. The learner is modelled as a subset of these correct rules, plus a set of incorrect production rules. Each new learner requires an individualized student model. In developing the student model, the type of knowledge (ie, declarative, procedural) to be defined must be determined. It has to be decided whether or not to include student goals, and how to include these. The methods used include the users outlining their own goals, providing self-documentation, and submitting answers to a pre-test.

Instruction Design on the Web

The Web is a delivery medium, as well as a provider of content and subject matter. It is easy to use HTML to deliver all of these offerings in text, graphics, sound and video. The Internet links people world-wide, and contains a highly diverse, easy- to-update source of information. Web-based instruction is shown to be more useful for intellectual knowledge than affective knowledge, and is not appropriate if users do not have Internet connection, if the material requires large amounts of audio or video, or if the subject requires the teaching of physical skills.

The creation of instructions involves the organization of information to promote specific learning goals. It is critical to keep this a priority in the design of Web-based instructions. There are two schools of thought in the creation of learning instructions; objectivists and constructivists.

Objectivists find a series of steps to lead the learner to the final goal. This final goal is defined in terms of behavior, ie "The learner will be able to demonstrate behavior x". This method does not take into account the individual learner's differences regarding prior knowledge or present motivation. It does expect a minimum prior skill relevant to the knowledge domain. This approach may work well for procedural knowledge, which can be exhibited, but is not as effective with declarative knowledge, and higher levels of learning.

The constructivist approach differs from the objectivist in that the student takes control of the learning. One example of this approach is the Cognitive Flexibility Theory. This theory deals with the special requirements for attaining advanced learning goals. It views the learning methods as having a multi-dimensional perspective with a criss-crossing of the subject matter in a non-linear fashion. The Internet is an ideal tool for this approach.

The Hypermedia Design Model (HDM), in utilizing the above theory, defines the design methods as what knowledge the designer hopes will be attained by the learner, and how the learning environment is organized. The learner methods are what the learner hopes to learn, and how the knowledge is accessed by the learner. In the HDM, the learner objectives are of the utmost importance.

The HTM consists of the following steps;

1. Define the learning domain;

   - The boundaries are set. The larger these are, the less detailed the knowledge will be.

2. Identify the cases in the domain, which are the knowledge bits, plus the instructional elements, such as text, graphics, sound, and video.

   The model splits into two paths from here;

3. The guided path creates specific trails through the knowledge domain. The design goals are identified and highlighted, using maps and maps in a criss-crossed landscape. The learner is led through this, returning to the same elements from many directions.

4. The learner controlled path allows the learner to specify how to navigate without the aid of the design spcified trails. Learners create their own objectives, regarding the domain, which supplies them with the necessary tools to explore it on their own, (ie, search engine, concept maps).

Cognitive Models for Structuring Hypertext

Hypertext-based systems are described as static, and are often critisized for their lack of expert guidance in the instructional sequence. However, intelligent systems are often critisized for their structured tutoring and embedded expert's design as the basis for learning.

One of the most common problems in hypertext systems, especially with novices, is disorientation, or becoming lost in the links. A number of advanced navigational tools are used to prevent this. Punctual aids, such as transport buttons, and help buttons, assist the learner to move forward. Structural aids, such as maps, filters, and indexes, help the student to view the overall structure. Historical aids show the users where they have been.

Although these navigational tools assist the learner to move in a non-linear manner, studies have shown that many students will still use the material in the same manner they would read a written text. This is dependent on the student's locus of control, prior knowledge of the domain, and familiarity with hypermedia. Learners with an internal locus of control, strong domain knowledge, and high level of comfort with hypermedia, are more likely to use greater initiative with the navigational tools.

In defining the cognitive models for structuring hypermedia, both the domain and pedagogical knowledge are considered. Domain knowledge includes declarative, structural, and procedural knowledge. Pedagogical knowledge contains a knowledge of the domain structures, the state of the student understanding, knowledge of tutorial strategies, and a means to use these strategies to facilitate learning according to the domain structure and the student understanding.

Hypermedia Systems

Hypermedia Systems (HMS), which are based on hypertext, are non-sequential, non-linear methods of displaying text, graphics, sound, and video. They use interface design and advanced navigational tools, and assume that the student's interpretation is more meaningful that the expert's. Hypermedia Systems use a constructivist approach to learning, where learning is regarded as the formation of "constructs" of understanding by the learner. The learner builds the knowlegde based on previous understanding by interacting dynamically with the domain structure. HMS provide a suitable means for this approach because they allow the learner to take control.

This system allows the learner to make an informed decision regarding where to proceed in the material. The hypertext is a node-link structure, which allows the user to move through the information using advanced navigational tools. The structure and the sequence is expert-defined, but the learner defines his own path to follow. This provides the students with the greatest opportunity to learn on their own.

Adaptive Hypermedia

Adaptive hypermedia involves the research of understanding student traits using computer interface. It attempts to trace student knowledge and provide individual advice, by creating an adaptive system with an interface management approach. A student model is used to adapt the display characteristics of the interface to the needs of the learner.

One study involving adaptive hypermedia, at the Swedish Institute of Computer Science, uses stereotypical knowledge of the learner. This is derived from observing several users. It uses an adaptive search and filtering mechanism that involves the use of clustering and direct user input, and avoids the need for advanced organization of hypermedia knowledge.

Another example is the Adaptive User Modeling System (Kobsa, Muller, and Neill, 1994). This system addresses the two main problems users face with hypertext; navigation and comprehension. It allows objects to alter according to the user's knowledge state. There is an assumption here that the user will bypass familiar knowledge, but choose the knowledge areas they wish to learn.

Semantic Net Architecture

Hypertext structures reflect a model of learning based on schemas. This knowledge exists in semantic memory, which is a network of interrelated concepts. In Semantic Net Architecture, such as Hypercard or Authorware, domain knowledge is organized as concept nodes. These nodes are connected via links and menus, which describe the relationship between the nodes.

OCRT Model

The OCTR model of learning styles, describes "Orientation" as relating to the learner's prior knowledge, "Coaching" as apprenticeship learning, "Tuning" as finetuning the knowledge, and "Routining" as increasing student autonomy. This approach emphasizes the constructivist approach, and the creation of links between new knowledge and old.

Summary

Eklund, in his research, "Cognitive Models for Structuring Hypermedia"[1995], draws the following conclusions and implications;

1. Browsers, such as Netscape, take a constructivist approach as they allow the student to take control of learning.
2. In the node-link cognitive models, learning is the accumulation and organization of knowledge structures, which can be represented as nodes and links. The nodes are the declarative elements, and the links are the procedural and structural elements.

Eklund suggests these possibilities to learning on the Web;

1. To use an expert's construction of the domain to form the basis for the creation of links and nodes.
2. To incorporate advanced navigational devices such as concept maps.
3. To provide online help, which may be intelligent help if the student is modelled.
4. To use adaptive interface based on stereotypical user classes to modify the environment to suit the needs of the individual learner.
5. To provide adaptive advise, and model the student's aquisition of knowledge through their use of the environment, (including navigational use, answers to questions, help requested), to intelligently suggest a preferred path through the knowledge base.

The above possibilities show how the use of student modeling can be related to learning on the Web. Online help may be specific to domain or generalized to more global assistance, (depending on whether the the problem occurs with the knowledge domain or with the tools).

Adaptive interface applications and adaptive advice would both require use of intelligent systems, with student modeling being a requirement for development. Further research could explore the above situations, focussing on other research and applications that address these areas. This research would be directed to how this information could be used in Internet course delivery

References

Alberta Research Council
ACE Project - Student Modeling

Eklund, John. Faculty of Education, University of Sydney.
Cognitive Models for Structuring Hypermedia and Implications for Learning from the World Wide Web