Jakob Nielsen

The need for HCI input to web design has been changing rapidly in tune with the change in dominating styles for websites over the last few years:

• 1993: Just having a server on the Web was enough to show that you were a pioneer! I remember faithfully going to "What's New With NCSA Mosaic" every day to see who were putting stuff up on the Web: the pages may have sucked like a vacuum but they came to us from the other side of the world which caused a major novelty effect.
• 1994: The main thing in 1994 was to show users how much information you had. This was the year of home pages that were no more than glorified hotlists with long bulleted lists of links. At this relatively early stage of the Web, people were still easily impressed by anybody who had real, *useful* content.
• 1995: Focused value-added information became key as users suffered under ever-increasing information overload. The preferred style for home page design in 1995 was to provide a clear sense of priority for the user and to showcase a small set of high-quality information, recognizing that users normally cover no more than about five pages when they visit a site. The design of Sun's home page was done according to this principle. Less is more definitely became a key design strategy.
• 1996: I think that web-surfing is dead. Sure, users may check out a few new sites every now and then, just as they may buy a new magazine from the newsstand when they are stranded in O'Hare. But to continue the magazine analogy, most users will probably spend the majority of their time with a small number of websites that meet their requirements with respect to quality and content. Hotlists can only grow so big (especially with the lousy user interfaces for bookmark management in current webbrowsers), so only a few websites will be graced with substantial numbers of repeat visitors.

The relationship between a website and a user can be enhanced, for example, by allowing the user to indicate an interest in a specific information object and then inform that user when the object changes or has been significantly updated (you don't want to bother users every time you fix a typo). We have done this at Sun for some time with our bug fixes. Subscribing customers who are particularly bothered by a certain bug can mark the SunSolve page for that specific problem and get notified when a patch is available.

Much as I dislike praising HotWired's user interface, they do deserve credit for their customized What's New page. Not only does HotWired keep track of when each user last visited so that new stories can be highlighted, they also allow users to customize the What's New page to list only stories within those sections of the site that interest them. An excellent way to cut through information overload and focus (and thus capture) a user's attention.

An even more interesting way of supporting user relationships is the personalized view of Ziff-Davis' ZD Net. Users can store an interest profile on the server so that subsequent visits will produce a list of new stories that match their interests. Of course, simplistic information filtering based on keyword matching will never do a perfect job, but they do seem to direct my attention to stories that mostly interest me, so the feature has succeeded in enhancing the relationship between me and Ziff. Future personalized view servers can be expected to use advanced information filtering methods like relevance feedback and various synonym matching ideas and may also allow users to help each other find relevant information.

In the early years, hacking HTML was the main requirement for doing a Web site, and user interface design was an afterthought. Recently, a real HCI contribution has become essential for Web design, with the need to draw upon our existing knowledge of icon design, knowledge elicitation (discovering appropriate information space structures), and usability testing. Users very easily get lost in hyperspace, and much existing hypertext research could benefit web designers as well as the designers of web software. The end of 1995 was probably the high tide for the value of traditional HCI research for the Web. Due to the change to relationship web design in 1996, we may now be more in need of task analysis techniques and ways of finding out what users would like to do even if they don't do it now.

At the time of the workshop, I expect the Web to span about 200,000 servers, and in two or three years we will have a million-site Web. One of the major implications of this cornucopia of webpages is a dramatically increased need for web designs to stand out in the crowd. Consider, for example, the home page for the Web magazine Word. The page is 766 by 1122 pixels large and thus cannot be seen in its entirety by users who do not have large workstation monitors. I would normally be hesitant to recommend such a large design for most user interfaces where one traditionally tries to design rather small dialogue elements that do not overwhelm the user or dominate the screen too much. But for its domain, the Word design may be appropriate. It is probably more important for them to look "cool" than to be highly usable: after all, people don't browse this kind of magazine for productivity reasons.

The implication of the need to be "cool" is that future HCI research must focus much more on what makes interfaces seductive and pleasant and less on what makes them easy to learn and fast to use. At the same time, the need for traditional HCI methods is not going away. On the contrary, Web sites are growing ever larger, with the resultant need for better navigation support and clearer information structure. I think, though, that most of the necessary HCI knowledge is already well-established in the more traditional areas and is simply waiting to be utilized. For example, there is a clear need for usability methods that will allow a large number of "gatekeepers" (business unit content providers) to enhance the quality of Web pages contributed by their local groups to a central corporate Web site. Heuristic evaluation would seem to be a perfect match for this need.

Participant Bio

Jakob Nielsen's recent books include "Multimedia and Hypertext: The Internet and Beyond", "Usability Engineering", (with Bob Mack) "Usability Inspection Methods", and (with Elisa del Galdo) "International User Interfaces". He also writes the monthly Alert Box column on Internet user interface issues.

Dr. Nielsen's previous affiliations include Bell Communications Research (Bellcore), the Technical University of Denmark, and the IBM User Interface Institute at the T. J. Watson Research Center.