Usability and the World Wide Web next up previous
Next: Terminology Up: Improving Web Usability with Previous: Introduction

Usability and the World Wide Web

The Web raises many usability issues [9]. In this paper, however, we focus on two specific ones which are fundamental to understanding how to enhance the link. The first of these is content: how users determine that the document referenced by a link meets their requirements. The second is the temporal behaviour of the link: how long the selected document will take to download. These two factors together: quality of the document and quality of service are key determinants of users' selection and use of links. These factors interact, however, so their combined influence cannot be determined a priori. For example, a user may be prepared to tolerate slow download where the document quality is known to be high. Conversely, if time is pressing, then a lower quality -- but more speedily downloaded -- document may be more acceptable.

Neither document quality, nor its corollary, relevance, can be properly assessed except in the context of use. The only certain way that a user has of determining these properties is to download the document. However, following this strategy blindly may require the user to invest a lot of time and effort. This is simply impractical when the user has many documents from which to choose, such as when picking from a list of documents generated by a search engine. One solution is to make use of whatever contextual information is available. For example, the choice of anchor may provide a meaningful summary of the content of the document which it references. The wider context in which the anchor appears may also provide further information.

The problems of relying on contextual information partly explain the current wide interest in developing a standardised document markup language such as W3C's Extensible Markup Language (XML) [29] so that content providers can add meta-data -- information about information -- to Web documents. One goal of adding meta-data is to facilitate the automatic processing of documents by Web applications such as search engines. We argue, however, that meta-data may also be an important resource for the design of an enhanced link. The first practical objective of our work is therefore to investigate how meta-data might be presented through an enhanced link.

As any Web user soon learns, the simple underlying abstraction of a seamless, distributed information space is rarely sustained in practice. Instead of an instant and deterministic response to link activation, the user typically must cope with unpredictable download times arising from variable latencies and network bandwidths. System response times in general are recognised as a major factor in determining the usability of an interactive system [4,5,6,14,21] and investigating how Web delays affect user behaviour is an area of growing interest [10,17,15,25].

Studies have looked at how users' strategies change to cope with extreme delays or system failures [7,17]. The unpredictable nature of Web download times means that users are unable to apply consistent coping strategies which leads to frustration and error. Attempts have been made to formalise this behaviour with an aim to helping developers design for download latencies [13,15]. Other studies suggest that users' perceptions of document quality on the one hand, and their tolerance of download times on the other, are not independent. In particular, they indicate that download times may influence users' perception of document quality and relevance. For example, users' ratings of how interesting a document is have been shown to decrease for longer download times [23,25]. Conversely, users' tolerance of download delays is influenced by their expectations of document quality [15]

Despite the well-known detrimental effects of unpredictable delays on users, browser user interfaces all but ignore the problem. Whilst abstractions are useful tools for protecting users from unnecessary details of system behaviour, rigid adherence to a single abstraction may deny them the information they need to act effectively [8,16]. The link user interface, we argue, is a good example of this problem. The account it offers of the behaviour of the underlying system is too abstract and so fails to provide users with the resources for informed decisions and effective improvisation when potential problems are detected. Our second practical objective, therefore, is to investigate how to provide all users with a more informative account of download behaviour [20] which is helpful both for decision-making and for learning.

next up previous
Next: Terminology Up: Improving Web Usability with Previous: Introduction
Rob Procter