The growing interest in web-based education a nd tra i n i ng is spurring new resea rch and development into i mproving systems for the authoring and delivery of on line learning. lt has also prompted some old ideas to be revisited .
This section reviews some significant trends in this research that have a bearing on i mproving the functionality, usabi lity o r accessibi lity of e-lea rn ing software for extramural study .
3 . 5 . 1 Anywh ere, anytime study
One-way Broadband
Cook (1 998) writes that the biggest problem in providing Internet access in the developing cou ntries is "that of 'the last mile' (or first mile) linking between the towns a nd the villages" (p. 22). Advances in com munication tech nology - based u pon va riations of wireless m icrowave or satellite technologies - have opened u p new possibilities for widening the bandwidth beyond what the local telephone infrastructure can offer for delivery of learning materia l into rural areas. However, this is by no means a solved issue , with many of these new tech nologies also failing at "the last m i le", even in the developed countries.
O ne line of attack uses wireless networki ng tech nology to provide a 2-way i nternet service , which is typically several times faster than that available over standard copper wire lines. A wireless modem has the advantage that it can be used anywhere within a wire less network's coverage , a lthough its bandwidth is below that recommended for streaming aud io/ video , very large downloads or viewing graph ic-intensive Web sites (NZWireless, 2004 ).
Unfortunately, node-to-node links in a wire less network require a line-of-sight connection over a relatively short distance . This is difficult and expensive to provide in remote and hilly terrains.
Hand-held wireless systems using mobile cel l phone technology have also been explored (Jo et a l . , 2001 , Boada et al, 2003). Mobile phone networks face similar performance issues as wireless networks. Moreover, they exacerbate the usability issues relating to e-learning because of poor performance and inadequate user interface (Qu i n n , 2002 ) .
Satellite tra nsmission technology is opening up promising new possibilities for high bandwidth delivery to individual distance students as telecommunication and med ia
Chapter 3 : Web-based cou rseware 50
com panie s d rive research into linking the I nternet with digital television services (Bi yth , 200 1 ) . Two-way broad band satellite transmission is an important technology that is used by learning centres a nd schools. Satellite communication s have the added advantage of being a g lobal technology that does not req uire a well-developed local com m u n ications infrastructure . Thus satellite broadcasts have been used for some years to deliver d istance education on an international basis to learning centres in Africa , Latin America and elsewhe re in the developing world (Cook, 1 998, p. 29).
For individual use rs, satellite reception is as straightforward as satellite television reception. Satellite transmission is more problematic and expensive 1. To be cost effective, home u sers would have to be linked through some other technology such as a landline to a shared transmitter. S uch transm itters are not generally available in ru ral areas. Moreover, links to them face the same performance issues as for ru ral telephone o r wireless commun ications.
A more practica ble approach for the ind ividual extra mural distance student would be to employ a hybrid approach which u ses satellite reception for rapid down loading of learn ing material a nd a terrestrial link to the transmitter for receiving commun ications from the student machine, includ ing system information necessary to maintain the connection a nd ensure the integrity of the data transferred . Early efforts along these lines manag ed download speeds fou r times faster tha n a dial-up (Arora et a l , 1 996) . Pioneer NZ inte rnet service provider, I H U G , has since developed a commercial version of this tech nology providing down load speeds of up to 40 times that of dial-up connection. In this way , rural customers could be offered a one-way broadband service at an afforda ble price that uses a standard dial-up internet connection to link to a central satellite transmitter, and special satellite reception hardware and software to receive downloads from the Web via the transmitter (IHug, 2004).
While video-conferencing over the I nternet is technically straightforward , the results for the rural home u ser, employing sta ndard copper wire connections, mirror those described in a computer journal as "teleconfe rencing where you have a Webcam on your PC and look at a postage sta m p-sized fuzzy image changing twice per second" Honeyball (2002a, p. 1 72 ) . H owever, i ntegrating satellite broadcasts into e-learn ing not only provides a more effective medi u m for streaming video to a ru ral home computer, but opens up possi bilities for using para llel television broadcasts to complement computer-based learning materials.
Chapter 3 : Web-based courseware 5 1
"A truckload of tapes "
To ensure that a student's cou rse material is accessible anywhere, anytime, Gehne et a l . (200 1 ) advocate d istributing pa rt or a l l of the system's fun ctional ity to the student's machine, so that the system will work off-li ne.
The TILE system dupl icates web server functions on the student's computer, a nd requires that the student machine and the central server be periodically synch ronised on line (Gehne et al. , 200 1 ). If necessary, teach ing materials can then be distributed by mail using portable media like C D-ROMs. Gehne et al. cite computer pioneer E.E.
Dykstra ("never underestimate the bandwidth of a truck-load of tapes", p.2) in support of this approach.
A similar tack is taken in Dietinger et a l . ( 1 998) and Bark (2001 a).
3.5.2 I n d ividual isation
Efforts to individualise ed ucational software systems have proceeded along two axes - system-in itiated adaptivity and u ser-in itiated adapta bility (Oppermann et a l . , 1 997). The major d ifferences between adapta ble and adaptive approache s are listed in Table 3. 1 .
Adaptation may be implemented across a range of dimensions as illustrated in Table 3 . 2 . A learning system may e mploy adaptation across a l l , some o r none of these d imensio n s . Some features, e . g . the help system, may be adaptive , while others , e . g . the interface look and fee l , m a y be adaptable.
In arguing for ada ptive interfaces, Benyon and D . Murray a sserted that the drawback with the adaptable approach is that "the user must lea rn functions wh ich a re tangentia l t o their m a i n task. Although some routi nes will o n l y have to be set once, others involve learning specialised com mands . Tailoring facilities are typically very general, do n ot take fine-g rained , ind ividual d ifferences into account and do not cater for a user's task needs, either perceived o r implicit" ( Benyon et a l . , 1 993, p. 1 98).
Efforts to imbue cou rseware with adaptive capabilities have d rawn u pon three decades of experience in the development of intelligent ITSs. One method of developing adaptive web-based systems has been to customise existing standalone ITSs so that they can be accessed from anywhere via the Internet. SOL-Tutor (M itrovic and Hausler, 2000) and PAT (Ritter, 1 997) are examples of standalone tutors that have been successfu lly parted to the Web. However, the na rrow subject focus of such tutors
1 In an October 2004 personal interview, the principal of a rural primary school reported that 2-way satellite
C h a pter 3: Web-based courseware 52
m a kes them an u nsatisfactory choice for teach ing whole courses. Efforts h ave therefore expanded along the lines of:
• integrating tutors and other adaptive learning tools into broader cou rsewa re ; a nd
• the development of adaptive systems that can teach whole cou rses.
Property Adaptability Adaptivity
Primary research discipline HCI AI
Manner of adaptation Static Dynamic
Locus of control User System
Flexibility of user view More Less
Com plexity of user view (cognitive load) More Less
Hidden complexity (im plementation) Less More
Ta ble 3 . 1 : Contrasting properties of adaptable and ada ptive computer systems
Dimension
I nterface "look and feel" - how user interacts with the system Sequencing - what part of the course is presented next Learning style - how the material is presented to the student
Learning level - the degree of difficulty of the material
presented
Learning help - what help is presented to the user a nd when
Table 3.2: Forms of ada ptation in lea rn i n g systems.
Chapter 3 : Web-based courseware 53
Another approach , which is discussed u nder the subheading of "Collaboration and Help", is to incorporate software agents that collaborate with and assist the user in the learn ing task.
Interactive tools to foster multidimensional learning
The United Kingdom-based Byzantium project sought to ach ieve more practical results for artificial intelligence in education by limiting the goal to produci ng intelligent tutori ng tools that extend the scope of a lecturer rather than replacing him or her (Pate I et a l . , 1 997). Patel et a l . ( 1 997) propose that Web-based versions of such tools be integrated into broader courseware .
Audiograph (Jesshope et al . , 1 998) a nd MAN IC (Stern et al. , 1 998) support learning by-lectu re th roug h providing synchron ised audio and slide presentations. MAN IC supports individualisation by adapting presentation sequencing to suit each student. A n umber of web-based tools have been developed, at least at the p rototype level, to offer better support to active lea rn ing dimensions such as learn ing by doing, by exploring, by tuto rial and by But even when they have been successfu lly prototyped, lack of sta ndard isation makes it difficult to integ rate them into commercial cou rseware systems. The "Model for Distri buted Curriculum" (Murray , 1 998a) proposed sta ndards by which an e-learning system could dynamically locate relevant tutorials on the World-Wide Web. More recent standardisation efforts have focussed upon reusable
learning objects, which will be discussed l ater in this chapter.
lnterBook (Brusilovsky et a l . , 1 998) and Metalink s (Murray et a l . , 2000) are electronic textbooks wh ich g u ide students while allowing them to access materia l in the order of their choice. In this way they provide an environ ment for students to learn by explori ng . This is especially so of Metalinks, which enables the i ndividual student to adapt navigation paths and the presentation of material to their own learn ing level and preferences.
Adaptive courseware
Adding adaptivity to educational software may come at the expense of other lea rning goals. For instance , with ITSs the locus of control is typically with the system rather than the user.
Overall ITSs have h ad a narrower teaching focus than web-based courseware. They h ave taught a particular problem-solving skill in a well-defined and structu red topic. WITS however, is a course-oriented intel ligent tutoring environment which g ives
Chapter 3 : Web-based cou rseware 54
students the more coarse-grained treatment a imed at the average student that they would receive from a class teacher (Callear, 1 999).
Other researchers h ave explored ways to bu ild adaptation directly into the Web techno logy itself so as to produce adaptive courses. Adaptive courseware systems employ a user model a nd an adaptation strategy to generate an ind ividualised user view. Two early efforts were the Dynamic Course Generator (DCG) (Vassi leva, 1 997) a nd E LM-ART (Weber et a l . , 2001 ). DCG generates individual cou rses accord ing to the learners' goals and previous knowledge and dynam ically adapts the course according to the learn ing prog ress of the student (Vassileva , 1 997). ELM-ART is an i ntelligent, interactive , educational system to teach L I S P which incorporates tools that e nable students to com m u n i cate with each other and with their tutor (Weber et al., 2001 ).
Both DCG and E LM-ART were developed to teach computing-related subjects. They have not made the transition into broader ed ucational use.
l nterbook, an offsh oot of ELM-ART, is a tool for authoring adaptive electron ic textbooks. lnterboo k uses a depth-first strategy, i.e. encouraging the learner to demonstrate thorough knowledge of one topic before moving on to another topic.
ITSs a re typically n ot adaptable by the user. An a lternative adapta ble , breadth-first approach represented by Metalinks will be discussed in the following section .
The customisable (adaptable) interface
Benyon et al. ( 1 993, p. 1 97) wrote that the practical results from the adaptive approach had been "rather disappointing a nd problems h ave proved far harder to dea l with than was first expecte d . " Even today, more than ten years after that article was written , add i ng adaptive featu res to courseware remains at the level of research prototypes .
T. Mu rray ( 1 999) and Bork (200 1 a) argue that individualisation goals can be achieved more simply by paying g reater attention to design of the student interface . Mu rray writes that since basic g raph ics authoring is considered a solved problem, most ITS researchers have n ot prioritised student interface desig n, but have restricted their systems to predefin ed screens a nd styles. This "severely constrains the types of tasks a nd i nteractions that a n ITS can have with the student" (Mu rray, 1 999, para. 28).
M u rray and his col leagues (Murray et a l . , 2000) present evidence from the development of the M etalinks project that adapting the content to the individual student may be better and more si mply achieved through deeper design of the interface, keeping the locu s of control with the student. They argue that underestimating the importance of inte rface design (adaptability) in learning systems can lead to
Chapter 3 : Web-based courseware 5 5
overestimating the importance o f artificial intelligence ca pability (adaptivity) .
Metalinks uses a breadth-first approach to allow the student to traverse the course at the level they choose . lt was designed to accommodate intelligent software featu res but was implemented in a less sophisticated version. This version was then eva luated for evidence that students would benefit from add ing more advanced featu res. The resea rchers were somewhat surprised by the resu lts, which ind icated that well designed interface features a nd non-intelligent tools satisfied many of the user needs that they expected would have to be addressed by addi ng intelligent software capabilities. "Th is work supports the notion that good interface design and passive but powerful user features ca n sometimes provide the benefits that are ascribed to more sophisticated o r intelligent featu res (which ca n be more presumptive , controlling, o r intrusive than passive features)" (Murray e t a l . , 2000, para. 42).
The authors felt this experience also underscored a more general lesson that inadequacies in the design of the interface or usability of educational softwa re will overshadow any benefits from more sophisticated features it may possess (Murray et a l ., 2000) .
3.5.3 Collaboration a n d Help
Collaborative agents
Benyon and D . Murray ( 1 993) point to the development of interface agents as a more promising (and less ambitious) approach to imbuing systems with adaptive capabilities. With th is approach , interfaces employ agents that can interact with u sers and with other agents to assist users to accomplish specific tasks individually, or co-operatively across a network. "Essentia lly agents are adaptive systems, but systems which are specialised and know about only a very small part of the world" (Benyon et a l . , 1 993, p.
200).
Collaborative agents are a n adaptation of intelligent tutoring tech nology to s upport collaborative forms of learn i ng beyond the simple email, chat rooms and threaded d iscussion lists of commercial courseware .
Collaborative learning has not generally been the focus of adaptive tutors, which are geared to the individual student. Vassileva argues that wh ile an adaptive system "truly 'cares' for the learner", by tailoring the course to the m , it can reinforce the problem of the "lonely learner". She holds that adapting to other's ideas and experiences is part of the learning process, so, "maybe sometimes the learner should adapt, n ot the
Chapter 3 : Web-based courseware 56
environment?" (Vassileva , 2001 , Slide 1 2).
1-Help (Greer et a l . , 2001 ), uses i ntelligent agents to attempt to overcome the learner's isolation by o rganising and mon itoring com puter-based collaboration within a com m u ni ty of learners . 1-Help attem pts to find the most suitable member of such a com m u ni ty to respond to another m e mber's query. A variant to this approach (lshikawa et a l . , 2002) channels q uestions posted to a cou rseware bulleti n board to other students o r teachers currently online to answer.
V I N C E N T ( Pai va and Machado, 1 998), which is built around an animated softwa re agent that acts as a n intelli gent hel per to assist trainees to complete their courses , intervenes when students get off-track. Lopez e t a l . (2002) and Rasseneur et a l . (2002) use virtual student software agents to foster collaboration and di scussion among networked students.
Natura/ language querying
Another approach to the lonely lea rner problem is to provi de a means by which students can ask the system questions about a subject and get a meaningfu l response , even i f a h uman tutor is not available online. The Flexible Structured Cod ing Language a pproach enables querying of a database, using a natu ral language-like syntax, which can retu rn known facts to the student (Heinrich et a l . , 200 1 ). Hei n rich and Kemp propose exte nding th is approach by u sing artificial intelligence techniques to enable the system to i nfer add itional facts from thi s database to pass on to students (Heinrich et a l . , 2000).
3 .5.4 I nteraction
Specialised languages
Bork criticises the "very l i m ited forms of interaction" possible wi th coursewa re using HTML-based browsers (Bo rk, 200 1 a , p. 59). He counterposes a g raphical scripting language with which experienced teachers can design hig h ly interactive, natural languag e-based tutorials (Bork, 200 1 a , p. 64; Bork et al. , 1 992).
Specialised languages for simplifyi n g the programming of learni ng material are not a new idea. They were a n integ ral part of the PLATO and TICCIT projects. This approach has now been taken up more widely in the form of developing speci a li sed languages