A glass Box Approach to Adaptive Hypermedia
Download 0.88 Mb.
|
- Navigate this page:
- POP Is an Adaptive Hypermedia System
- Knowledge Acquisition
Classifying POPEven if we have not discussed the POP system in detail yet, it might be beneficial for the reader to see how POP can classified in terms of the framework outlined in this chapter. POP’s User ModelPOP infers users’ information-seeking task from their interaction with the system. We have a similar view on tasks as Vassileva has in the system HYNECOSUM, (1994, 1995). We shall motivate how adapting to the task will subsume both adapting to users’ knowledge and occupational category. When discussing different kinds of user models, we also discussed models of users’ cognitive abilities and personality traits. As our system is a hypermedia system, we suspected that cognitive ability would be related to their ability to search for and find information. We did find that there was a correlation between spatial ability and ability to navigate in hypermedia, but we choose not to adapt to users’ spatial ability for reasons we shall discuss later on. Our results did verify that cognitive characteristics might be very useful in predicting individual user’s ability to efficiently make use of the POP system. Our user model is explicit, short term and individual. The reason why the model is individual is that we only model one aspect of the user, their information-seeking task. The different tasks will activate different sets of stereotypical answers that the system believes to be fitted with the inferred task, but, as discussed above, we regard this to be an example of a stereotypical answer, not a stereotypical user model.
Figure C. Classification of POP marked in bold.The way we infer users’ tasks, is by continuously monitoring them and making assumptions on which task users are currently attempting. To do this we use a form of local plan recognition (Wærn, 1996) which will be explained later on. As said in the introduction we are concerned with finding a combined user- and system controlled adaptivity. We therefore provide different means for users to override the assumptions made by the system. In terms of Kühme et al.’s control classification, POP can be described as a combination of the two figures in Figure C, i.e. on the one hand users can always change any adaptations made by the system (B), but the adaptive system will also continuously try to infer the information searching task (A). When the user wants to change the assumed task, the system presents the alternative tasks in a menu, which is why the system proposes the possible adaptations. There is no proper, separable, domain model in POP. We structured the domain concepts into an object-oriented knowledge representation, and the user model is inherited into each object and directly referring to the various pieces of information deemed relevant to users’ tasks. The domain model is therefore part of the overall structure of the database. In the POP system, we have a very simplistic model of the dialogue with the user. All we keep track of is some simple aspects of the dialogue history. We have therefore not implemented any proper discourse model. 
(A) (B)
Figure D. Classification of POP according to aspects of control. Finally, concerning the timing of the adaptations, POP will in the current implementation adapt after each action of the user. In the version which was evaluated in the final comparative study, the system would only adapt when the user was moving from one page to another. We shall come back to the reasons for changing this behaviour later on. POP Is an Adaptive Hypermedia SystemAccording to the adaptive hypermedia classification above, POP can be characterised as affecting the content of a page through a combination of additional explanations and explanation variants. We achieve this through a powerful stretchtext-technique which will hide or open quite large chunks of hypertext. In this respect our approach is different from the kind of stretchtext employed in both MetaDoc, (Boyle and Encarnacion, 1993), and in KN-AHS, (Kobsa et al., 1994). The adaptivity in POP only affects the navigation through the hyperspace in a minor sense. POP can order the links available in the hotlists according to their relevance to the information seeking task. Knowledge AcquisitionThe target domain of the POP system is a software development method, named SDP. In order to decide whether and how to create an adaptive help system that was fitted to SDP users’ needs, we had to study them and their potential problems. Our problem was similar to that of knowledge acquisition for knowledge-based systems, which is why we were inspired by a method for system development named Cognitive Task Analysis (CTA) (Roth and Woods, 1989). The CTA method is divided into two parts: first assess the total problem situation, and then perform a focused task analysis directed at the most prominent problems identified in the first phase. So, we first tried to understand the problem area, focusing on the types of tasks that users typically were involved in when addressing the manual. This was done to avoid ending up with a solution that would only solve an example scenario, but not capture the users’ real needs. The first part was realised by a set of informal interviews. In the second phase, the actual task analysis and collection of a corpus of questions was done. The task analysis did first and foremost result in a task hierarchy which described users’ tasks with respect to their information needs. It also gave us the basis for deciding which queries and which answers would be best fitted to meet users’ needs. But, the task analysis also revealed some interesting differences between users that we wanted to study further. One difference was in how well the users perceived that they could navigate by the graphs in the available on-line manual. Some users wanted to avoid the graphs and others liked them. The other interesting difference between users was their understanding of fundamental building blocks in the SDP method. These two aspects of users, their ability to navigate and their background knowledge, caused us to do two focused studies on these particular aspects. There exists few other studies of object-oriented methods (Detienne and Rist, 1995), none focused on users’ needs for documentation support. Instead, they focus on how well, for example, novices and experts are able to apply a particular method, or transfer effects when moving from procedural design approaches to object-oriented approaches. When designing help systems, as pointed out in (Breuker, 1990) it is quite common to discover that many problems could be overcome through redesigning the target system, or in this case, the SDP method. Some of these problems we discovered with SDP originated from the SDP process itself, others from a mismatch between the process and the organisation in which it was applied. These findings have proved useful in the subsequent development of SDP-TA, but they lie outside the scope of the help system – the help system is supposed to give help on the method as it is, not compensate for it. It should also be noted that whatever method or tool that is put under such close scrutiny as done in the studies described in here, would reveal bugs and mistakes made in the design. The finding in our studies should not be taken as evidence that SDP is a bad method. SDP tackles a truly hard problem: how to design large software applications. There are no simple short cuts to good quality of such software. Still, the SDP method is complex and difficult to get a quick and intuitive grip of and this, of course, affected how well engineers at Ellemtel were able to make use of it. We start by providing the necessary background of the target domain, SDP, and the PUSH project in section . We discuss knowledge acquisition methods and our reasons for choosing CTA in section . Then, go through four of the studies done in PUSH during the knowledge acquisition phase: the initial interviews, section , the task analysis, section , the study on the relation between cognitive abilities and navigation in hypermedia, section , and the study on the differences between novice and expert users of SDP in section . Finally, we summarise the design demands imposed by these four studies in section . Directory: ~kia -> papers ~kia -> Usability and fun An overview of relevant research in the hci community papers -> Comparing Two Approaches to Context: Realism and Constructivism ~kia -> B. Cahour (1), P. Salembier (1), Ch. Brassac (2), J. L. Bouraoui (1), B. Pachoud (3), P. Vermersch Download 0.88 Mb. Share with your friends:
|
