CHI 96 Workshop: HCI and the Web, Position Papers

David Gilmore

Computer Science Department, Bowling Green State University
davidg@cs.bgsu.edu

The world-wide web is an interesting bundle of contradictions when it comes to questions of usability, and maybe we should regard this as a clue to the fact that what usability there is is more accidental than deliberate. Thus, the workshop announcement attributes the success of the WWW to HCI factors and yet it is regular occurence for most web-surfers to meet at least one "File not found" message every day. The difficulty of constructing and maintaining a web-site is substantial and user interface issues of site construction remain a thorny problem. And as HTML grows and people wish to achieve more and more through web pages, so the complexity of using web pages will increase.

HCI of web browsing

As HTML grows, so more and more levels of interaction can be engaged in upon the www. But arguably some of the original success of the www was the very fact that it was limited and, therefore, easy to design consistently. Furthermore the fact that it had to be machine-independent (and arguably browser independent) meant that simplicity was the order of the day. But, for example, Netscape have introduced frames into web browsing - for which we can ask the usual HCI-style questions:

Is the functionality required?
Is the implentation usable?
Are frames easily comprehended?

Netscape's own arguments concerning the first question are based on a valid criticism of existing web browsing - namely the loss of the context of what one is viewing. By providing a broader overview of a site's structure through frames, the reader may well be more likely to find information more quickly and more comfortably.

By contrast, Netscape's own apologies about their implementation of frames gives the answer to the second question - No. As one example, there is the problem of going back through frames, where (for no apparent reason to the reader) going back takes one back to some long gone document, and not the previous page. A side-effect of frames seems to be that one's browser no longer knows what page one is looking at - one cannot bookmark with a frame, nor can one go back to the screen before last (except through the previous one). These changes not only muddy the waters of navigation, they also confuse the user's model of what the browser is doing.

In this respect the third question also gets a negative response. Although frames ought to be more easily used, it seems that Netscape have failed to appreciate the importance of simplicity in web browsing. In fact, only a few days ago Ii read an article in which someone commented how web-surfing had been remarkably more successful than CD-ROM exploration and they concluded that it was the sheer simplicity and consistency of it which had won people over.

Even more than many computer-oriented tasks, web-surfing is one where the interface must not be yet another distraction from the task of finding some information.

HCI of Web Page construction

Most web pages seem to sit on two ends of a continuum. they are either very complex and almost certainly controlled automatically (for example, The (London) Times) or they are really very simple, being primarily a collection of links and pictures.

Support for the construction of in-between pages, with quite a lot of text and the need for links between them, but only occasionally updated is minimal. Adobe PageMill is one application which claims to fill this niche - it even claims that you can write your web pages directly in it, rather than importing them. And yet, Pagemill offers no search and replace functionality, or spelling checking - surely two critical word-processing tasks.

Indeed the parallels between PageMill and early PageMaker are quite striking. Both performed admirable tasks, but were unable to function as stand-alone packages. early Pagemaker could at least read files from standard word-processing packages, whereas PageMill only reads rtf.

Furthermore in many situations web-pages are documents that have a life elsewhere as well and must therefore be maintained, printed, etc. Most current authoring packages do not support these other functions. A highly desirable product would seem to me to be something which can take a document from (say) Microsoft Word and convert it automatically, but customisably, into html. For example, it could identify headings, keywords, lists etc and show them to the user and ask them to confirm how they should be translated. If it is a large document, then the package could advise on how to divide it up.

HCI of Web site management

Managing a network of documents with internal links is very different form managing one page with external links. But some problems, like validating the existence of all links in a page have been remarkably slow to be supported. Furthermore, managing passwords in a network of documents is also not well-supported.

Yet both of these kinds of task will, I believe, become more common. The existence of browsers on everyone's machine can make html an extremely effective dissemination device. For example, the ability to break lecture materials and teaching resources down into small chunks can enable us to provide students with links to material which they will not feel inclined to print out (each unit is too small), but which is easily revisited. However, such materials are probably not intended for access from outside the instititution. Or, consider the use of html for disseminating the minutes of meetings across an organisation (or organisations). In these situations there is a clear need to link access to passwords, without necessarily worrying about where the reader is.

Frames provide the reader with the context for the page they are reading, but there is (as yet) no equivalent for the site manager, showing the network of documents and enabling links to be checked, maintained and constructed.

Summary

The world-wide web is still primarily at the enthusiast level, which means that human factors issues have not been of primary concern. However, browsing is now moving beyond enthusiasts only and before too long one can expect there to be pressure for page and site management to move beyond the enthusiasts. New tools are needed and needed fast and they must be designed with a focus on the tasks that the many different users will actually have.

In many respects these problems are a result of the technology being so new that no-one had or maybe even has a good idea of why and how it will be used. This makes user-centred design an extremely challenging enterprise. Without it, however, there could prove to be substantial disappointment concerning the real, practical power of the web.

CHI 96 Workshop: HCI and the Web, Position Papers