• No results found

"Instructional Design" or "Design Science" is that part of the general field of education which is concerned with prescribing optimal methods of instruction to bring about desired changes in knowledge and skills for a given set of students. It is clearly related to learning theory but has a different focus . Instructional design focuses on the entire environment in which the pupil is placed and how this will effect his learning. This is not to deny the importance of the learner but merely to focus on creating the best learning environment and then to look at the learner-environment interaction from this point of view. Briefly, instructional design deals with how best to teach a given course to given students and tries to provide a blueprint of what methods to use for particular desired outcomes and students.

Reigeluth suggests the following framework for use when evaluating theories of instructional design;

• Methods • Conditions

• Outcomes

Relevant variables that can be controlled.

Relevant variables that are fixed.

Results of the application of particular methods in particular conditions

A particular theory of ID must consider these three general aspects and provide a coherent set of principles for choosing appropriate methods for given conditions and desired outcomes.

Methods can be broken down into three categories;

• Organisational strategies Elemental methods of organising subject matter

su::h as the use of examples or diagrams and sequences of the selected content.

• Delivery strategies Such as the choice of teachers and media and their

Chapter 6 Instructional Design and Training Artificial Neural Networks 6.4

• Management strategies Which are methods of choosing or sequencing the organisational and delivery strategies.

The organisational strategies can be further partitioned into either micro strategies, which deal with a single idea or macro strategies, which deal with the presentation of more than one idea.

An instructional model consists of a complete set of strategy components and their relation to conditions and outcomes. A prescriptive model or theory will prescribe the best methods for a given set of conditions and desired outcomes.

A prescriptive theory would thus contain a number of prescriptive principles of instruction such as;

To increase long term retention, begin instruction with an overview that epitomizes the content rather than summarises it. Then gradually elaborate on each aspect of the overview, one level at a time, constantly relating each elaboration to the overview.

A very useful set of criteria by which instructional theories may be judged is

-7 suggested by Reigeluth [Reigeluth

198)1

ibid. p25] and reproduced here with

minor modifications; • Theory • Consistent • Limitations • Valid • Simplicity • Usefulness

Is it a theory or just a list, model or set of beliefs?

Is it internally consistent?

Are its boundaries and limitations clearly defined?

Is it contradicted by any known data?

Is it unnecessarily complex? (Like modem physics)

Does it provide useful predictions or assistance in design?

• Comprehensive How much of the total variance is covered?

• Optimal Is there anything better?

These criteria provide an excellent yardstick for evaluating a theory of ID or any

other scientific theory. Theory construction is, however, an iterative process and

many of the present theories of ID do not satisfy the criteria suggested above. A

common fault appears to be the simple classification of conditions or methods described as part of a particular model are actually found to be unsuitable when

Chapter 6 Instructional Design and Training Artificial Neural Networks 6.5

principles relating conditions to methods are determined. The simple and tidy classification of conditions is then modified or extended in an ad hoc manner. Following the next section, which considers the difference between instructional design for people and instructional design for artificial neural

networks, a number of ID theories are examined in detail. For this analysis, in

addition to the criteria above, each theory is considered in terms of the three aims of this chapter.

• Does it provide a context for the heuristics proposed in chapter five?

• Does it provide any additional heuristics for training artificial neural

networks?

• Does it assist in laying foundations for a theory of instructional design for

ANNs?

It is important not to be too literal in looking for analogies. Each concept or principle used in an ID theory needs to be considered in relation to training ANNs. Is there an obvious analogy? If not, is there some facet of ANNs which is loosely analogous? It is well to be wary of deciding too soon that there is n o analogy for a particular aspect just because ANNs are s o much simpler than people. For example, the concept of motivation or the use of cues might b e discarded as irrelevant to ANNs but with further thought it may b e possible to relate these ideas to the gain term in a learning algorithm or to pre-trained sub networks of a special type. Once the analogy has been found then simplified versions of the instructional design principle or model aspect under consideration may be found quite useful in the training of ANNs.

Chapter 6 Instructional Design and Training Artificial Neural Networks 6.6

6 . 2 D IFFERENCES BETWEEN ID FOR PEOPLE AND ID