Our interface contains several different parts that are each designed to meet certain needs: the graphs should support navigation in the hyperspace, the hotlists should make it possible for users with little knowledge of the domain to ask follow-up questions on unknown concepts, and the adaptivity should help prevent users from being overloaded with information within the page, etc. From our studies, we can see that we have met some of these goals, in particular, the adaptive system was preferred by the subjects, it required fewer actions within the page, and the choices made by our adaptive system influenced subjects’ choice of information entities to include in the solution.
What is potentially lacking from our design is additional help on how to navigate between pages. As we have included quite a lot of information in each page, we have made the information space smaller than it would have been if each page had been divided into several smaller pages. So our adaptivity is, in fact, helping to make the information space smaller and thereby easier to navigate in. Still, a global map or a dialogue history, or potentially even an adaptive solution that affects the navigation between pages might have improved the situation.
What is needed next, not only for our system, but for intelligent interfaces in general, is to prove that they are also useful in a longer time perspective. This can only be shown through studying users in real working conditions under longer periods of time.
To make a system to adaptive to its users’ abilities or preferences is an intuitively appealing idea. It has the potential to help users to more efficiently search for and filter out information from large information spaces. Still, very few such systems have managed to become widespread and considered usable. Researchers in the adaptive systems’ field have to devote some more attention to usability issues before this will happen.
In this thesis we have been concerned with how to design one particular adaptive hypermedia system, POP, taking usability considerations into account. Ensuring the usability of any system involves a number of considerations. There should be a metaphor upon which the interaction with the system can be built. There should be a method for system development that at various stages during the design work considers usability issues. Once the system is implemented, there must be ways of evaluating the design, preferably in realistic conditions with the targeted user group.
We have argued that the black box in a glass box metaphor is a useful vehicle for adaptive systems’ design. Through the glass box the user is provided with some insight into what the adaptive system is doing, while the exact mechanisms in all its details are hidden in the black box. The ultimate goal is to provide an interface that is transparent, predictable and thereby controllable.
Since the glass box metaphor is fairly general it needs to be interpreted for any specific system design. In the POP system, the glass box was realised through giving domain-dependant names to the set of information seeking tasks that the system adapts the information content to. From the task name users can get a (more or less) clear picture of what information content that task corresponds to. Furthermore, through making the systems’ choice of task and corresponding adaptation of the information visually obvious, users stand a chance to learn the relation between task and adaptation. Since users are allowed to alter the task themselves, they can control the adaptive behaviour if necessary. We make the systems choice of task and corresponding adaptation visually obvious through:
-
the names of the information entities,
-
the linguistic style in the information entities,
-
the colouring (in red) of headers connected to the task,
-
an explicit sentence at the top of the answer page that explains which task the system currently assumes that the user is performing.
The second usability consideration concerned the method for developing the system. The method we used to develop POP was focused on the targeted user groups problems and needs. We investigated various aspects of how the users differed in terms of possessed knowledge, roles, cognitive abilities and information seeking tasks, and how those differences influence their ability to learn and make use of the targeted domain information. When designing POP we attempted
to meet all these differences, but through various means. POP has:
-
hotlists to enable novices to learn about concepts and principles unknown to them,
-
maps of the information structure to enhance navigation through the hyperspace,
-
a guide frame that structures the content of the node,
-
query-menus so that users can ‘jump’ to specific nodes in the hyperspace.
The POP system can be characterised as an adaptive hypermedia system that adapts the contents within the node. It does so through a variant of stretchtext where whole chunks of texts, information entities, are stretched or collapsed. The user can alter the adaptations made through clicking on the headers of the information entities whereby the text under the header is stretched (or collapsed).
POP uses a combination computer-aided adaptive and self-adaptive strategies. On the one hand the user can set the information task they think that they are performing, choosing from a menu of tasks or through clicking on the hotlist in the sentence at the top of the answer page. The latter renders a list of tasks ordered by their probability as inferred by the plan recognition component. On the other hand, the plan recognition component will continuously monitor users’ actions and possibly change the task to better fit with the users’ actions.
The choice to adapt explanations to users’ information seeking tasks is somewhat unusual. Most systems that adapt explanations based on some user model do so based on users’ knowledge. Adapting to users’ information seeking tasks through combining information entities as done in POP has many advantages: it is easier to address users’ real needs, the authoring process is simplified, and the adaptations are conceptually simple both to users and to authors of the information.
Finally, concerning the third usability issue, that it is important to evaluate the system under realistic conditions, we have done a set of combined bootstrapping and evaluation studies. The last, comparative study showed that the adaptive system was preferred by the subjects. In it we used tasks collected from users in their daily work situation. Still, putting the system to test within the company for real usage would be the ultimate test of its usability.
Share with your friends: