Email: sedjkc@thor.cf.ac.uk
S. Hurley, S.B. Jones and N.M. Stephens, Department of Computing Mathematics, University of Wales College of Cardiff, PO Box 916, Cardiff. CF2 4YN, U.K.
Email: steve@cm.cf.ac.uk
Currently many universities in the United Kingdom are teaching units on high performance computing in their undergraduate degree programmes. Many more recognize the growing importance of this topic now that the technology has come to prominence and these institutions would welcome assistance in the form of readily available computer based teaching material to get their courses off the ground. In some cases they wish to provide a more comprehensive coverage of the subject area.
The teaching material (which goes under the name of ParTeach) produced by the programme will consist of ten modules of courseware, each corresponding roughly to four hours of teaching material. Delivery of the courseware will be in hypertext format and will link in to examples and code which can be run on networks of PC's, Macintosh computers and from X terminals and on the local parallel processing facilities (an nCUBE2 at Cardiff, transputer networks at Kent and Southampton and an AMT DAP at QMW). The modules will cover material on (i) occam and transputers (Kent), (ii) scientific applications and system support for scientific computation (Southampton), (iii) computer architecture and Fortran 90 (QMW) and (iv) algorithms and paradigms (Cardiff).
Several platforms for presenting hypertext tutorial material were investigated, including hypercard (for the Apple Macintosh) and UnixHelp and DosHelp. The study revealed that although these platforms had certain advantages over HTML documents residing on WWW servers, for example, finer control over display formatting and robust authoring tools. However the ability to make courseware publicly available in an accessible manner was an overriding advantage. Also, the mechanism of web servers provides an added advantage of being able to launch arbitrary processes upon request from a browser. The pilot study conducted at Cardiff discovered that it was possible to allow users of the parallel programming tutorial to launch real applications on the departmental parallel computer, an nCUBE2, and to have the results of the run displayed in the window of their browser. This remote running of real example programs is particularly useful in the demonstration of parallel programming principles, especially when the distribution of such resources is so scarce, and access to the hardware would otherwise be impossible.
Until the advent of WWW, courseware was comprised of stand alone packages. These generally implemented hypertext type systems themselves. A recent example of this is Microcosm [Hall94] developed at Southampton University for a PC running Microsoft Windows. This presents a framework for providing a given lecture course and allows the inclusion of text, images and video.
The Ceilidh (pronounced kay-lee) courseware [Zin91] package is an interesting development. It provides an online tutoring system for programming languages such as Pascal, C, C++ and Modula2. Ceilidh provides course notes and an integrated program development system that includes the automatic marking of programs. It has been adopted by many UK universities.
Over the past year a variety of courseware has appeared on the WWW. Many courses provide notes and lecture material in the form of HyperText textbooks --- where sections of text are broken up into hypertext links that naturally reflect the layout of a book. Some have extended this work further to include links to other documents worldwide. Some level of interaction between the student and the course has also been facilitated.
One notable case is the Global Network Academy's attempt to set up a Usenet University Project [Butts94]. Their long term goal is to set up a fully accredited on-line university. For the time being they are concentrating on developing and giving courses across the Internet. They employ WWW and other networking approaches e.g. electronic mail and file transfer). They have developed a correspondence course in C++ which provides hypertext notes on the WWW and tutorials via email.
The Phoenix project [Lavenant94] at the University of Chicago aims to develop an integrated academic information system providing full Internet connectivity and wide-area distributed hypermedia services for all teaching and research. They have developed a WWW system for teaching biological subjects that provide course notes, syllabus and announcements in a integrated environment that allows course instructors to convey this information in an easy manner.
One novel use of the WWW has been produced at the University of Geneva where they have produced a system that "animates" Pascal algorithms used in a data structures course [Ibrahim94]. This allows students to run programs (or program segments) and interact with the program execution to step through programs and examine contents of variables and data structures. Graphical representation of complex data structures is also provided.
Developing courseware is not merely a matter of preparing a series of lectures, linking them together and packaging them as a course. A comprehensive design strategy must also consider how to implement and manage a course, how to evaluate the materials that are used, and how to assess the learners. With that in mind, there was a conscious effort to take a systematic approach to the development of course materials. According to Rowntree [Rowntree82] the systems approach should incorporate four basic strands:
Designing such materials confronts many well-established and accepted instructional design principles. The first step in many instructional design models, and the one that the course development team agreed to use, is to analyse the learners who will use the materials. Analysis of even a subset of potential users, however, would have proved expensive and time-consuming. Consequently the development team has tried to compensate for that by putting effort into ensuring that the material would appeal to a broad audience.
To achieve this, the team designed the model that included links to all relevant material ( i.e. in the same unit) in the unit's index document. (This document also contains links to all diagrams within the unit.) Each document in the unit has a link to the index, the link being grouped with the other unit traversal links, e.g. next page, home page etc. This design was adopted because courseware has a more logical progression than do many hypermedia software packages. A benefit of this design is that a novice user can proceed safely through the work, since the actual notes are generally free from links which break the flow of information presentation. On the other hand, the experienced user can easily access any related topics via the index page. This feature is available at all levels of the unit, thus minimising unnecessary link traversal - a problem which tends to disorient users and hinder the effective use of the courseware [Marchionini88].
The use of Gagné's events of instruction as an evaluation tool led to further changes in the development of a suitable model and particular changes in courseware content. For example instructional event three, stimulating recall of prerequisite learning, led to the insertion of additional references to other units.
Utilisation of the ASSURE model and Gagné's Instructional Events is not intended to provide a prescriptive design model. Rather, it provides a framework based on sound instructional strategies within which it is possible for development team members to develop a dialogue about design strategies. The tools provide a common ground for collaboration. To illustrate the function of these tools in the development of educationally sound courseware, the implementation of each of the steps of the ASSURE Model will be discussed.
After evaluating the uses of the courseware it may be necessary to develop separate modules adapted to meet the various teaching environments, but which cover the same topics. At present, the team is developing the courseware to be robust enough to stand as an independent teaching tool, with the aim that it will satisfy most of the demands made upon it. Lecturers may then be selective about the courseware material they use in their courses, adapting their courses as necessary.
Therefore, the material is designed for users with no previous knowledge or experience of the courseware content, and who lack adequate understanding of related principles. Hence, the lecture notes that were used as a foundation for the courseware content needed to be expanded thoroughly to provide a more satisfactory coverage of the topics presented, since they were originally designed to be used to teach experienced computer science students within the university.
Jargon has to be eliminated or explained thoroughly where unavoidable, with relevant links to a glossary provided in each (main) document. Most concepts must be explained in simple terminology, minimising the need for a familiarity with the topics covered or any points made therein. It is a goal of the development team to restrict the flow of information as little as possible. For example, demonstrations and examples are included via linked documents, rather than being contained within the main unit notes (This also conforms with HTML style guide policy by decreasing the length of documents.)
Learners are informed specifically of what they should learn from the courseware. This approach increases their awareness of the key points in the notes, and hence increases the potential of the courseware to deliver the education required. Although each unit is prefaced by an introduction, the stating of objectives also serves as an advanced study organiser.
There is another reason for stating specific objectives explicitly. If the courseware is to be used in any way in conjunction with lecturing courses, they inform the lecturer of precisely what they aim to present, and thus the lecturer can accurately adapt the courseware into his/her own course.
According to the model specification, a unit contains a list of objectives which relate to the unit as a whole, yet which can still be defined specifically. Each sub-topic within a unit also has a list of objectives which cover that particular sub-topic. Thus the stating of objectives reflects the hierarchical design of the model. Some sample objectives are:
Whilst HTML provides the possibility of linking in various hypermedia features eg. mpeg movies, audio clips, in-lined images etc., it requires a suitable browser to handle them. Mosaic is such a browser, allowing the potential of hypermedia to be utilised fully. However, another factor limiting the use of hypermedia is the hardware used to run the browser. A computer with no external speaker cannot play audio clips, no matter which browser is used. Therefore, the "effectiveness" of a piece of hypermedia software, however well designed, is ultimately determined by the browser used and the machine upon which that browser is run. It has been assumed that the implementation of the courseware will most likely occur in university settings where hardware and software capable of presenting all media employed by the courseware is readily available.
Whilst these algorithms are not overly complicated nor lengthy, the visualisation of their execution for non-simple processor networks is quite challenging. This problem could be eliminated by the use of hypermedia, in particular by using mpeg movies.
For most of these algorithms, their execution has been modelled onto a three-dimensional hypercube (a common processor network, resembling a simple cube with a processor at each corner.) This hypercube has been illustrated graphically (by use of a raytracing package) together with the messages which are passed between the processors. By producing seqences of these images (in some cases 168 individually raytraced frames.) small mpeg movies which graphically illustrate the operation of these algorithms have been compiled. When presented with sufficient explanatory imformation they are an effective means of presenting the courseware. Feedback from users indicate the movies have clarified their understanding of the algorithms. A snapshot from one of the movies is shown in Figure 1.
Simulated annealing is a non-trivial process whereby the loads on a processor network may be minimised (usually not optimally). Observation of students has shown that involvement in the implementation of this algorithm improves their comprehension of it. The courseware implements a simulation which allows the user to execute the algorithm on a simple linear processor network. The user is able to see the results of the algorithm by means of a graph which plots the load on the network against the "work" done by the algorithm. To enhance understanding the user can adjust the various parameters which affect the algorithm's performance. The algorithm can be rerun with different parameters, and the new results are plotted on the same graph with the previous results. This allows the user to compare and contrast performances, and understand how the parameters affect the algorithm, and consequently more fully understand how the algorithm works.
The demonstration of simulated annealing is possible because HTML allows links to executable programs and scripts. As long as a suitable HTML document is produced by the program(s) called, the user is unaware of all the "behind-the-scenes" operations that are taking place.
The main feature of HTML which allows user interaction is the HTML form, which enables user input to be passed to the programs which produce the HTML documents.
Some snapshots of the simulated annealing Interactive Example are shown in Figure 2 and Figure 3.
The simulated annealing interactive example, described in the previous section, illustrates one method of achieving learner participation. However, whilst demonstrations such as this serve to increase the user's understanding of the relevant topics, they still do not guarantee the learner's correct processing of the material presented. Each unit therefore includes exercises developed specifically to examine a range of learning objectives, from the testing of simple verbal knowledge, to more complex problem solving.
Presenting a variety of question styles is simple. However, for the interaction of the user to be complete, some form of assessment and feedback needs to be implemented. This can be achieved either by user's self assessment or by external assessment. User self assessment involves providing a model answer for each question presented. The user is left to assess the correctness of submitted answers.External assessment involves the submission of the user's solutions to a tutor or instructor, who provides appropriate feedback to the user.
To evaluate the effectiveness of the teaching materials, learners will be asked to rank their confidence for understanding a particular learning objective. Each learning objective for the unit will be stated, and confidence ranked accordingly. The feedback received from this data will allow designers to evaluate the effectiveness of the material at the level of the individual unit.
More detailed evaluation will depend on how the material is implemented and how learners are assessed. Consider the following scenarios:
The second option, providing the materials as an additional resource, gives students the opportunity to study the relevant material at their own pace. This can be a boon for students who have missed lectures or need to reinforce the notes they have taken with additional material. The benefits of such an approach are primarily to the student. Use of the materials in this matter would have little financial impact on course delivery. The cost of designing such a resource, however, must then be absorbed wholly by the developers. Given the increasing number of students in Higher Education, however, such benefits should not be ignored. From an educational perspective, some research has gone so far as to suggest that hypermedia should be limited to use as a resource [Smeaton91, Laurillard93].
The final implementation scenario is a mix between the former two. In this scenario, implementation would involve the integration of the hypertext material with lectures. This would allow some savings in terms of lecture time and management and administrative costs. For example half of the course might be delivered by lectures and half by hypermedia. This combination would avoid the need for course organisers to write customised learning materials as any subject-specific input could be handled within lectures.
The construction principles we have used during the development stage are easy to employ. Finally, our students find the packages easy and intuitive to use, and allow them to work at their own pace.