Chapter 10: User Interfaces and Visualization - by Marti Hearst
Contents
|
Modern Information Retrieval Chapter 10: User Interfaces and Visualization - by Marti Hearst |
Contents |
What makes an effective human-computer interface? Ben Shneiderman, an expert in the field, writes (p.10) [shneiderman97]:
Well designed, effective computer systems generate positive feelings of success, competence, mastery, and clarity in the user community. When an interactive system is well-designed, the interface almost disappears, enabling users to concentrate on their work, exploration, or pleasure.
As steps towards achieving these goals, Shneiderman lists principles for design of user interfaces. Those which are particularly important for information access include (slightly restated): provide informative feedback, permit easy reversal of actions, support an internal locus of control, reduce working memory load, and provide alternative interfaces for novice and expert users. Each of these principles should be instantiated differently depending on the particular interface application. Below we discuss those principles that are of special interest to information access systems.
Offer informative feedback. This principle is especially important for information access interfaces. In this chapter we will see current ideas about how to provide users with feedback about the relationship between their query specification and documents retrieved, about relationships among retrieved documents, and about relationships between retrieved documents and metadata describing collections. If the user has control of how and when feedback is provided, then the system provides an internal locus of control.
Reduce working memory load. Information access is an iterative process, the goals of which shift and change as information is encountered. One key way information access interfaces can help with memory load is to provide mechanisms for keeping track of choices made during the search process, allowing users to return to temporarily abandoned strategies, jump from one strategy to the next, and retain information and context across search sessions. Another memory-aiding device is to provide browsable information that is relevant to the current stage of the information access process. This includes suggestions of related terms or metadata, and search starting points including lists of sources and topic lists.
Provide alternative interfaces for novice and expert users. An important tradeoff in all user interface design is that of simplicity versus power. Simple interfaces are easier to learn, at the expense of less flexibility and sometimes less efficient use. Powerful interfaces allow a knowledgeable user to do more and have more control over the operation of the interface, but can be time-consuming to learn and impose a memory burden on people who use the system only intermittently. A common solution is to use a `scaffolding' technique [rosson90]. The novice user is presented with a simple interface that can be learned quickly and that provides the basic functionality of the application, but is restricted in power and flexibility. Alternative interfaces are offered for more experienced users, giving them more control, more options, and more features, or potentially even entirely different interaction models. Good user interface design provides intuitive bridges between the simple and the advanced interfaces.
Information access interfaces must contend with specialkinds of simplicity/power tradeoffs. One such tradeoff is the amount of information shown about the workings of the search system itself. Users who are new to a system or to a particular collection may not know enough about the system or the domain associated with the collection to make choices among complex features. They may not know how best to weight terms, or in the case of relevance feedback, not know what the effects of reweighting terms would be. On the other hand, users that have worked with a system and gotten a feeling for a topic are likely to be able to choose among suggested terms to add to their query in an informed manner. Determining how much information to show the user of the system is a major design choice in information access interfaces.
The tools of computer interface design are familiar to most computer users today: windows, menus, icons, dialog boxes, and so on. These make use of bit-mapped display and computer graphics to provide a more accessible interface than command-line-based displays. A less familiar but growing area is that of information visualization, which attempts to provide visual depictions of very large information spaces.
Humans are highly attuned to images and visual information [tufte83][kosslyn89][larkin87]. Pictures and graphics can be captivating and appealing, especially if well designed. A visual representation can communicate some kinds of information much more rapidly and effectively than any other method. Consider the difference between a written description of a person's face and a photograph of it, or the difference between a table of numbers containing a correlation and a scatter plot showing the same information.
The growing prevalence of fast graphics processors and high resolution color monitors is increasing interest in information visualization. Scientific visualization, a rapidly advancing branch of this field, maps physical phenomena onto two- or three-dimensional representations [keller93]. An example of scientific visualization is a colorful image of the pattern of peaks and valleys on the ocean floor; this provides a view of physical phenomena for which a photograph cannot (currently) be taken. Instead, the image is constructed from data that represent the underlying phenomena.
Visualization of inherently abstract information is more difficult, and visualization of textually represented information is especially challenging. Language is our main means of communicating abstract ideas for which there is no obvious physical manifestation. What does a picture look like that describes negotiations over a trade agreement in which one party demands concessions on environmental policies while the other requires help in strengthening its currency?
Despite the difficulties, researchers are attempting to represent aspects of the information access process using information visualization techniques. Some of these will be described later in this chapter. Aside from using icons and color highlighting , the main information visualization techniques include brushing and linking [eick95][tweedie94], panning and zooming [bederson96], focus-plus-context [leung94], magic lenses [bier94], and the use of animation to retain context and help make occluded information visible [robertson93][card91]. These techniques support dynamic, interactive use. Interactivity seems to be an especially important property for visualizing abstract information, although it has not played as large a role within scientific visualization.
Brushing and linking refers to the connecting of two or more views of the same data, such that a change to the representation in one view affects the representation in the other views as well. For example, say a display consists of two parts: a histogram and a list of titles. The histogram shows, for a set of documents, how many documents were published each year. The title list shows the titles for the corresponding documents. Brushing and linking would allow the user to assign a color, say red, to one bar of the histogram, thus causing the titles in the list display that were published during the corresponding year to also be highlighted in red.
Panning and zooming refers to the actions of a movie camera that can
scan sideways across a scene (panning) or move in for a closeup or
back away to get a wider view (zooming). For example, text
clustering can be used to show a top-level view of the main themes in a
document collection (see Figures ![[*]](cross_ref_motif.gif)
and
). Zooming can be used to move `closer,' showing
individual documents as icons, and then zoom in closer still to see the
text associated with an individual document.
When zooming is used, the more detail that is visible about a particular item, the less can be seen about the surrounding items. Focus-plus-context is used to partly alleviate this effect. The idea is to make one portion of the view -- the focus of attention -- larger, while simultaneously shrinking the surrounding objects. The farther an object is from the focus of attention, the smaller it is made to appear, like the effect seen in a fisheye camera lens (also in some door peepholes).
Magic lenses are directly manipulable transparent windows that, when
overlapped on some other data type, cause a transformation to be
applied to the underlying data, thus changing its appearance (see Figure
). The most straightforward application of magic
lenses is for drawing tasks, and it is especially useful if used as a
two-handed interface. For example, the left hand can be used to position
a color lens over a drawing of an object. The right hand is used to
mouse-click on the lens, thus causing the appearance of the underlying
object to be transformed to the color specified by the lens.
Additionally, there are a large number of graphical methods for depicting trees and hierarchies, some of which make use of animation to show nodes that would otherwise be occluded (hidden from view by other nodes) [furnas94][hendley95][johnson91][lamping95][robertson93].
It is often useful to combine these techniques into an interface
layout consisting of an overview plus details
[greene97][plaisant95]. An overview, such as a table-of-contents
of a large manual, is shown in one window. A mouse-click on the title
of the chapter causes the text of the chapter itself to appear in
another window, in a linking action (see Figure ).
Panning and zooming or focus-plus-context can be used to change the view
of the contents within the overview window.
From the viewpoint of user interface design, people have widely differing abilities, preferences, and predilections. Important differences for information access interfaces include relative spatial ability and memory, reasoning abilities, verbal aptitude, and (potentially) personality differences [egan88][shneiderman97]. Age and cultural differences can contribute to acceptance or rejection of interface techniques [meyer97]. An interface innovation can be useful and pleasing for some users, and foreign and cumbersome for others. Thus software design should allow for flexibility in interaction style, and new features should not be expected to be equally helpful for all users.
An important aspect of human-computer interaction is the methodology for evaluation of user interface techniques. Precision and recall measures have been widely used for comparing the ranking results of non-interactive systems, but are less appropriate for assessing interactive systems [lagergren98]. The standard evaluations emphasize high recall levels; in the TREC tasks systems are compared to see how well they return the top 1000 documents (see chapter 3). However, in many interactive settings, users require only a few relevant documents and do not care about high recall to evaluate highly interactive information access systems, useful metrics beyond precision and recall include: time required to learn the system, time required to achieve goals on benchmark tasks, error rates, and retention of the use of the interface over time. Throughout this chapter, empirical results of user studies are presented whenever they are available.
Empirical data involving human users is time consuming to gather and difficult to draw conclusions from. This is due in part to variation in users' characteristics and motivations, and in part to the broad scope of information access activities. Formal psychological studies usually only uncover narrow conclusions within restricted contexts. For example, quantities such as the length of time it takes for a user to select an item from a fixed menu under various conditions have been characterized empirically [card83], but variations in interaction behavior for complex tasks like information access are difficult to account for accurately. Nielsen [nielsen93] advocates a more informal evaluation approach (called heuristic evaluation) in which user interface affordances are assessed in terms of more general properties and without concern about statistically significant results.