Guidelines for Creating and Evaluating Reference-Based InterfacesBy:Gary J. Dickelman As the work we perform becomes more and more computer-mediated, a number of phenomena occur. First, the extent to which user interfaces (UIs) have evolved from transaction-based data collectors to on-line reference bases has been dramatic. Second, the user's reaction to the software tools they use has becomes more and more important. In this paper I will deal with both issues - the notion of the interface (or graphical user interface - GUI) for on-line reference tools, and how the construction of the GUI affects our tolerance to use it. The factors that affect our reactions are:
Examples of interfaces that violate most items in the list are easy to find, particularly on the Internet. These are the web pages that provide too many links, create associations between unrelated items, place you in endless loops, contain "link rot" (links that have expired), require you to remember things as you move forward, and the like. A recent link analysis of the web site of a respected company revealed the following:  Figure 1: First-Level Links from a Home Page The center of the picture contains a representation of the Home Page. The "spokes" represent independent links to other items on the page or to other pages. This page contains over 100 distinct links. Even more frightening is a picture of the first two levels of links:  Figure 2: Two Levels of Links Associated With a Home Page The small circle in the middle of Figure 2 is Figure 1. Shading is formed by links from one item or "node" to another. There are approximately 5,000 jumps in a region that scrolls 1,000+ pages containing 700,000+ words. Observations:
The site was not developed by a hacker but was created by software experts. How can this be? It seems that the elements of the interface that focus on reference material - the hypertext - was not expertly constructed, even though other interface elements were. In fact, there are many misnomers regarding hypertext, particularly in the age of the Internet (where many novices possessing the HTML hammer feel that every word is a nail). Hypertext Engineering What is hypertext? In Robert J. Glushko’s SALT 1990 presentation Practical Introduction to Hypertext and Hypermedia he suggested several options: a) A right-brain artistic panacea that is a revolutionary new concept b) ‘Hype’ + something else c) An evolutionary concept for increasing the accessibility and usefulness of on-line text If your answer to the question is c then follow along. Constructing a hypertext (or hypermedia, performance support system, CBT, or enterprise application) requires engineering discipline. A favorite definition of engineering is "a science by which the properties of matter and the sources of energy in nature are made useful to human beings…" (Webster’s New Collegiate Dictionary). Even if an interface is intended for recreation, the developer typically had a purpose and a goal in mind. Engineering, therefore, consists of meeting the goal by creating a usable interface with respect to content, structure, semantic encoding, and wayfinding. In a 1996 usability study entitled Navigational Issues In WayFinding On The World Wide Web (D. Bachiochi, E. Choinard, N. Conlan, D. Way of Aetna Life & Casualty, and M. Danchak of The Hartford Graduate Center) Wayfinding is defined as providing human beings with the answers to four simple questions:
The study concludes that the tools inherent to web browsers are not sufficient for effective wayfinding; I would add that neither are conventional tools and techniques for graphical user interfaces. They must be augmented with:
Dillon, McKnight, and Richardson ([7], pp. 172 - 175) explain that there are important distinctions between physical representations and semantic intentions of hypertext. It is no surprise that the Wayfinding study concludes that usable web sites must include both structural and semantic constructs.
It is frustrating when you jump to a web site or start a GUI application and find yourself waiting for the page to load - not because of graphics, but because of the length of the document. Usually people experience a very long scroll bar (indicating lots of text below) and tend to run out of patience before the entire scroll is available for viewing. Developers of such pages are often called Scrollers. They develop or convert large quantities of text as single scrolling entities while missing important usability principles. Scrollers hang on to the artifact of scrolling - a traditional means of accessing text on a computer - but confuse it with hypertext, where restructuring, "chunking," semantic encoding, and wayfinding maps must be present if value is to be added. According to Nielsen ([9], p. 263): Traditional text is extremely heterogeneous as can be seen by comparing a mystery novel with a corporate annual report. You do not need actually to read the text to distinguish between the two. But the same two texts would have looked exactly the same if they had been presented on-line .... During a recent intranet design session one of the participants said, "Let’s not waste time analyzing this manual to death. It has an index and a table of contents. Just use them." But direct conversion may render a document largely unusable the moment it’s on the screen. If care is not taken the document may lose structure, meaning, and even content as so much is suddenly hidden. Consider restructuring each of the following as hypertext: A Prayer for Owen Meany by John Irving [4] The Dilbert Principle by Scott Adams [1] The Oxford English Dictionary (OED) Value would decrease with an Owen Meany hypertext. Nielsen ([9], p. 287) cites that readers of general interest materials have better recall when the materials are linear than when they are structured as hypertext. Books are intended to be read linearly, people do not prefer to read on-line for very long, and the physical characteristics provide important clues that effect motivation - like how many pages remain. The Dilbert Principle might better serve us in hypertext format as corporate reference material. Just imagine jumping directly to Mandatory Self-Deprecation, linking to Strategies For Survival then Idiots. Or go directly to any cartoon that includes Catbert, the Evil HR Director. Consider the following Web site organization for The Dilbert Principle:  Figure 4: A Proposed Web Site for The Dilbert Principle The Site Map & Title Area is static and has no scroll bar. It contains an explicit representation of the site structure as conceived by the designer. The title and picture of the book jacket do not scroll away as you navigate to any major element (chapter), thus context is not lost. This Site Map is the book’s table of contents (TOC), which is how the author conceived its structure. When you click on a chapter it appears in the Content Area. The Entry Point Area is reserved for links other than those related to the explicit structure. It links to content based on how the site is used by human beings from a mostly semantic perspective. Entry points are determined by observing and analyzing how people use the document and make associations between its words and the larger context. The clues of usage for a printed work are usually pretty obvious: dog-eared pages, sticky notes or other papers hanging out of the book, dirty binding (because of frequent use), folded pages, highlights, scribbles, and the like. Entry Points represent implicit structures and semantic encoding of the document from the reader’s point of view and typically serve 80% of users as they perform typical tasks. Entry points can be independent of or dependent on the site Map. The choice is based on how much intra-document navigation (versus inTER-document navigation) is required and how the structural elements of the site Map must be augmented by semantic clues for the site to be usable. In the Dilbert example the entry points are independent of the site Map. We might have analyzed each chapter for implicit structure and presented a separate set of entry points for each one. The user would be presented with a different set of entry points each time a different chapter was selected from the TOC, thus forming a hierarchical browser - another useful but often abused wayfinder. Hierarchies must be applied with care. They should be as flat as possible and their use must not result in loss of context. In the previous example the user never loses site of the levels of hierarchy because the Site Map Area and the Entry Point Area never disappear. On the other hand, the conventional way of navigating via hierarchies goes something like this:
When asked how people find their way around a printed work the answers are surprising: "I flip through the pages back-to-front until I see the weather map or the picture of such-and-such. I know it’s just past the middle, on the back side of a page." They are referring to artifacts of the printed work - which may have value as on-line wayfinders. They define an implicit structure for the user. Artifacts help us to fine-tune our knowledge of the global structure to find our way in an instance. When we analyze entry points we find that they suggest the importance of being able to "see" how much stuff there is, what it’s about, and where specific content lies in relation to the whole. Wayfinding, therefore, is incomplete without both site maps and entry points. Readers must have artifacts to make inferences analogous to physical size, shape, and dimensions of the printed work. Donald Norman ([13], pp. 104, 105) notes that our information-based technological world has an inherent problem: It is invisible. "It is up to the conveyor of the information and knowledge to provide shape, substance, and organization...[via artifacts]." Without appropriate design of artifacts in our virtual domains, Norman says, "...our tools will continue to frustrate, to confuse more than clarify, and to get in the way rather than merge with the task. The power of information artifacts is that they provide an unrivaled opportunity to enhance our lives." Entry points may seem inconsistent compared to a site map. For example, Mandatory Self-Deprecation, Cartoons, and Stupid Corporate Slogans have nothing to do with one another. It doesn’t matter because entry points are what people find useful for navigating Dilbert as a reference book (determined through analysis, such as contextual interviews). The topics are functional. Think about the dashboard of your car: Audio Controls (radio), Environmental Controls (heat and air conditioning), and State Indicators (speed, oil pressure, charge, and temperature) are quite distinct but nonetheless comprise functional groupings. The Oxford English Dictionary (OED) was structured as hypertext according to the same principles applied to Dilbert, including structural and semantic constructs in the wayfinding schema. The user interface, however, is quite different. The OED can be found on the web at http://www.oed.com/oed.html. A description is included in Nielsen [9], pp. 330-332). Highlights:
According to Nielsen, "We should also note that the OED is one of the few examples where hypertext is actually more readable than paper." Notice the flexibility in the design of the OED interface with regard to the dual metaphors for long and short entries. Some designers are adamant on the issue of scrolling and insist that only a single "screen" or node of text is appropriate. A colleague recently asserted that she didn’t want any node of information to be larger than a single 640x480 screen. Neither those who profess using a card metaphor nor the scrollers should win the battle. Applying usability principles, observing human beings work, and analyzing the nature of a document will suggest whether or not nodes need to be restricted in size. What we’ve seen in the examples suggest that GUIs that include a reference or hypertext component must:
Conclusion: The ideal GUI and user interface therefore, must reflect the task at hand, the functionality required by the application, and the usability and readability principles discussed earlier. Primarily among them is the ability to maintain context, providing a common framework for information that reflects well though out mental model, a look & feel that makes optimal use of white space and a navigation paradigm that provided value with every click. If you are successful in aligning these various, and hard to achieve goals, you will have a system that minimizes unwanted surprise, and maximizes the value and learning experience of the user. Isn’t that what it’s all about anyway? About the Author Gary Dickleman is Vice President of GURU, Inc. He is a leading authority on the application of computer learning technology, information technology, and business process reengineering applied to the concept of performance support systems and performance centered design. He has published several articles and written several books on the subjects of high performance user interfaces and technology based learning and systems that enhance human performance. Bibliography
|
