Introduction

The present generation of World Wide Web browsers provide users with too little information to help them decide whether to choose a link. Whilst there may be a number of factors that may influence this decision, the only information that the user can rely on being provided by a browser user interface is the document's URL, and an indication of its activation history. Sometimes, the user may be able to determine from contextual information whether to activate it or not (see Figure 1). For example, by choosing an appropriate anchor, the author of the document in which the link is embedded can help the user to determine quality and/or relevance [19]; similarly, a description of the document's size and type may help the user to assess how long it will take to download, and whether he or she will be able to view it locally. Since this kind of information is entirely discretionary, however, authors often do not go to the trouble of including it. Where such explicit information is lacking, users may still be able to infer useful information about the document and its download overheads. For example, the URL may contain implicit information such as the document's type and the physical location of its server. Such resourcefulness suggests that users would benefit if this kind of information were made more explicit, perhaps through the agency of an enhanced representation of the link. The selection and activation of a link is not the end of the user's dilemma, however. Download times are often long and unpredictable and may leave users uncertain about when -- or if -- it will be completed. Yet, browsers (see Figure 1) generally provide only limited feedback of download progress. If anything can be read into the scale and placement of this feedback, it is that it is unimportant. Yet numerous studies have shown that users of interactive systems are very sensitive to factors such as delays.

  

Figure 1: A typical browser's `link user interface'.

We have been conducting an investigation of how users make link selections and of their behaviour subsequent to link activation. To date, this has included re-enactment protocols of videoed Web sessions and interviews with Web users where they were asked to interpret a series of URLs. The results show how users fall back on heuristics and improvising strategies drawn from past experience. One important issue that emerges is that users' heuristics are often flawed simply because the picture of Web behaviour that they get from the link user interface is too abstract. Based upon these results, we discuss specific ways in which links might be enhanced to improve users' understanding of Web behaviour and provide more reliable information about its contents. The concept of enhanced or `rich' links has been the subject of some discussion in the hypermedia literature (see e.g., [2]). Our objective here is to investigate what these might entail from a user perspective.

We begin by discussing some critical determinants of World Wide Web usability. We then present a model of link selection decision-making behaviour and relate this to our empirical observations of users. We then review relevant usability principles and finally, we propose ways in which users' decisions could be better supported through the `link lens', a prototype of an enhanced link based upon Magic Lens concepts of movable visual filters [1,27].