>> Ressourcen > Download Area für Papers > Jonassen, D.; G[..]

DynamIcons as dynamic graphic Interfaces: interpreting the meaning of a visual representation

David H. Jonassen

Pennsylvania State University, University Park, PA, USA

Ricki Goldman-Segal

University of British Columbia, Canada

Hermann Maurer

Institute for Information Processing, University of Graz, Austria

Abstract

Iconic interfaces in operating systems and hypermedia knowledge bases have become the default standard. The content, purpose, or function implied by icons may be enhanced by making them more dynamic. Dynamlcons can use representational, abstract, or symbolic graphics to convey the categorical or functional information or inferences and implications about its referent. Conveying more information in the design of icons can make iconic interfaces even more efficient.

1. Assumptions and purposes

It is without question that iconic interfaces enhance our ability to use information and operating systems. They may be especially useful for hypermedia and multimedia systems that contain large quantities and diverse types of information that are supported by a variety of computing functions. That is, icons may denote the nature of the links implied by the buttons, the nature of information in the nodes, or some specific functionality in the hypermedia system. Most icons are static icons: traditionally, they are displayed at the bottom of the screens. For example, the following static iconic buttons are consistently found in the HyperCard version of the Proceedings from the Hypertext '87 Conference at the University of North Carolina (Smith, 1987).

These icons indicate the type of information that will be presented when the buttons are clicked. From left to right, they indicate: go back one card in the stack; go to the Home stack; go to a previously viewed card; go to an overview map of the hypertext; go to the next stack; find designated text in the stack; go to the index of topics in the stack, and finally go to the next card in the sequence. The meanings of only one or two of these buttons is obvious to the novice, though most users can learn the conventional meanings of such a small number of icons in a short time.

Static icons never move; they never change shape, style, form, or colour; and they are always in the same spatial relationship to each other. The designers of hypermedia- knowledge bases will argue that consistent placement and functionality of the buttons enhances usability. When there is limited functionality in a hypermedia document, that assumption may be true: however, we will argue that for many hypertext and hypermedia knowledge bases, valuable information may be hidden or 'lost' by consistently using static icons.
The assumption of this paper is that icons used as buttons in hypermedia interfaces often tend to be 1) too static and 2) not graphic enough in terms of what each icon conveys and in terms of the spatial relationships between the icons. In order to convey more detailed information about the structure, functions, and meanings contained in hypermedia documents, we believe that icons can and should become more dynamic by changing form, shape, colour, position, or by becoming animated (Baecker, Small & Mander, l991). As the importance of information differs with each user's needs, perspective, and familiarity with subject matter, so should the nodes and links change. Icons should designate the changes made by users accordingly.

This paper will first describe the attributes of icons as graphic images along two dimensions and then describe how each of three types of icons (static, dynamic, and animated) can help users to better understand the intention underlying each type of iconic button.

2. Icons as graphic images

In order to discuss the uses and limitations of icons as they exist in current hypermedia systems, we will show how icons vary along at least two dimensions or axes (see Figure 1).

First, they differ in terms of how they function. We break this dimension into three categories: (a) icons that describe or exemplify a type of activity or function; (b) icons that define the purpose of an activity; and (c) icons that show the inferences or implications of an activity or chunk of information. Second, icons differ in terms of the way they communicate visually. Graphics can convey (a) representational, (b) abstract, or (c) symbolic information.
In the following section, we will describe these two dimensions in more detail. Then, we will describe static, dynamic, and animated icons in relation to these two dimensions.

2.1 How icons function

Icons may be used to depict at least three different kinds of information. Some icons represent the contents to be found at a location in the knowledge base. In other words, you double click on a folder and files open up.

The folder denotes the type of information in that folder. Other icons define the purpose of an activity. For example, an icon in the shape of an eye infers that you will see something if you click on it. The third kind of icon, more subtle than the previous two, indicates the relationships existing among various types of data by inference or implication. For example, an icon showing a puzzle may mean you have a problem to solve that relates to the chunk of data you are exploring. 2.1.1 To describe and exemplify type of information Many icons are designed to describe or to illustrate either the type information that is accessed by that button or the type of activity that will be initiated. For instance, in a hypertext built by Jakob Nielsen to describe some conference trips, he uses the following illustrations as a series of buttons.

Each book spine describes the type of information contained in it. Clicking on specific icons shaped like books on a shelf brings the user to either a dictionary, clip art, poetry, or whatever else is illustrated. This iconic representation not only conveys topics, but also uses the book as a metaphor for describing the contents. This type of graphical representation is definitional or descriptive. These icons describe classes of objects or ideas that share the same characteristics. They are simple concept lessons. Concepts are best learned by being shown three or more prototypic examples of the concepts. So, describe/exemplify icons could show examples of the concept in the icon itself. It defines or describes the contents of the information, but not the function of the link.

2.1.2 To describe purposes

Many icons describe purposes implied by the links. That is, the icons describe functions or activities that may be accessed through the link. For instance, by clicking on the icons below,

you might be able to (a) see something, (b) see it closer or magnify it, (c) paint something, (d) lock or unlock (the exact purpose is not clear from the icon, so it will rely on context), and finally (e) remember something. This type of functional icon is among the most common and probably the easiest to interpret.

2.1.3 To describe inferences and implications

Understanding and accurately interpreting icons whose function is to infer or to imply is a more difficult task. This is not a mode of information processing that users are adequately prepared for in their traditional schooling. Therefore, illustrating such relationships among units of data can be quite diffficult. It is more likely that such icons will have to be explained to users.

For instance, the icons above may infer that the user (a) will get a surprise, (b) should solve a problem, (c) had better look out, or (d) may have to juggle many things if taken on to another task.

2.2. How icons communicate graphically

Graphical representations convey information in three ways: representationally, abstractly, and symbolically. We see these three ways of communicating as points on a continuum rather than as separate or exclusive categories. In other words, an icon that conveys a straightforward depiction of what it shows is more representational. As an icon grows more abstract, it falls in the more symbolic range.

2.2.1 Representational graphics

The most basic level of visual communication is the representational level. At this level, the icon tries to represent the idea as faithfully as possible. In the figure below, the image shows two hands pointing toward each other.

While this representational graphic does not have photographic accuracy, it is simple to describe what the images are. It is very doubtful that anyone would misconstrue what the images are. However, it requires some degree of inferencing to interpret these hands as coming together in a handshake. Again, given the positioning of the image, it is likely that most people in a society that engages in handshaking would interpret this graphic as a handshake, regardless of whether it will occur or has just occurred.

Representational graphics try to represent the object in a straightforward fashion. They are more diffficult to produce and to reproduce. However, when clarity of the object is important, representational graphics are worth the extra effort.

2.2.2 Abstract Graphics

Abstract images are intentionally placed on the continuum between representational and symbolic (symbolic images, as we shall see, are very abstract and completely arbitrary images that represent an idea or object). Abstract images are certainly more abstract and more arbitrary than representational images; however, they bear some resemblance to the idea or object that they portray. The abstract images below represent (a) sound (as replayed through a speaker), (b) graphs and charts that represent quantities, and (c) speech, talk or an explanation as depicted by a speech balloon.

Abstract images are playing an increased role in our lives. For instance, highway signs, especially throughout Europe, now rely on a common set of abstract representations of crossings, routes, obstructions, or dangers ahead. These abstract images are easier to create and reproduce, and are extremely helpful for conveying a message to those drivers who might not share a common language. However, they require additional cognitive effort to learn their exact meaning. They also require cultural interpretation. For example, consider the designated icons for women's and men's washrooms in international locations such as airports. In Moslem cultures, it is not uncommon for men to be in long robes; in other cultures, it is not unusual to see women in pants more often than they are in the knee length skirt depicted in the icon. Nevertheless, having associated the sign with the meaning over time,"readers" of these icons are less likely to misinterpret the meaning. The message depicted by abstract icon is possibly more readily differentiated and understood once it is learned. In short, abstract icons tend to be efficient.

2. 2.3 Symbolic graphics

Symbolic images are designed to represent something, but they do not necessarily bear any resemblance to what they represent. Alphabets are symbol systems. The symbols below represent (a) a specific sound as designated by the note, (b) a question, and (c) a command or piece of information using alphabetical symbols.

Because symbols are completely arbitrary representations that bear no resemblance to the object that they are representing, they require the greatest effort in order to learn. That is, there is more cognitive load when using arbitrary symbols as icons in hypermedia interfaces.

3. Attributes of Graphics and Icons
3.1 Focus on Clarity and Expressing Meaning

One of the most important criteria of any graphic is how well it communicates its intended message. (Similarly, when one tries to understand the people of another culture, one closely examines their artifacts or representations in order to uncover the intended message.) Clarity is the main attribute that contributes to the ability of an icon to transmit its intended message to an audience. The problem is how to make sure that each user interprets more or less the same message. We understand that everyone will interpret even the clearest graphic within some range. For example, the graphic for the handshake may mean reaching an agreement to some people; to others, it may only signify a meeting or a parting. What we hope for in any visual communication is that the response to the graphic falls into a specific and reliable range. To accomplish this, we need to (a) accept some level of misinterpretation and the accompanying frustration, and (b) require the users to learn some set of iconic conventions to be able to interpret new icons. If the user is required to memorise specific icons, then clarity does not matter. If we expect that the user will be able to interpret new icons met along the path, then we need to establish conventions that will lead the user to make the best possible interpretation. However, as ethnographer Clifford Geertz (1973) points out, the final interpretation by multiple readers of any text may at best only fall into the same range and not elicit the exact meaning for everyone. According to Geertz, the message is conveyed not only by the medium to the reader, but also is a reflection of the mind of the designer or author of the medium.

One could argue that it is reasonable to expect users in a complex hypermedia system to invest enough effort to learn the meanings for a set of conventional images. Mannes (1985) argues that iconic interfaces can be confusing, wasteful, and ineffective. This may be true in cases where the quantity of operations is large. The question is whether or not the increased functionality is worth the increased cognitive overhead to learn and rehearse the meanings of sets of icons. In an experiment by Baecker, Small,and Mander (l991),it was found that turning static icons into animated icons was indeed useful and helpful for beginning users. In fact, adding animation to a static icon (a static pen icon becomes a moving pen icon that can mark the space in the icon box) reduced the learning curve for the inexperienced user. However, they found that expert users paid little or no attention to the animated icon and completed their tasks before the icon could help them.

The ability to best express the meaning rests with the clarity of the icon - whether static, dynamic, or animated. Overly expressive icons can often be inversely related to clarity. This is not to say that expressive icons do not have their place in depicting complex relationships. For example, in the interface Learning Constellations, a star icon represents a chunk of video, text, or sound data (Goldman-Segall, 1991). Each star chunk can be related to other chunks in different constellations, or groupings of data star chunks. The user decides the kinds of relationships that will exist among the stars. The constellation icon acts metaphorically: the underlying message is that users interpret within their own point of view, from their own universe. In short, if expressive graphics are used the relationship between the message and the icon may yield a spectrum of intricate relationships that may, in fact,be worth the extra cognitive load.

3.2 Attributes In DynamIcons

In order to describe DynamIcons, it is necessary to describe the ways in which icons could become dynamic. We focus on the following four possible attributes: significance and realism; hue and saturation; motion, and position.

3.2.1 Focus on significance and realism

The level of realism of any graphic object typically remains the same. However, with dynamic objects, there is no reason why the level of realism should not change. In other words, as an object designated by an icon is visited more frequently or as nodes and chunks connected to it are constructed, the icon may become more representational or more realistic to indicate that the idea is becoming more clear or more real.

For example, in Learning Constellations 2. O, a significance DynamIcon has been designed to measure the relative importance of various topics entered into a descriptive field (Goldman-Segall, R., Marcovici, M., Halff, L, & Flinn, C., 1993). As users chunk and describe various chunks of data, they can rate the significance of the topics that describe the chunk. This Significance Measure DynamIcon is a vertical coloured scale, resembling a thick thermometer [see (a) below . The measure is more saturated in colour at the top; higher saturation means higher significance. When users want to rate the significance of various related topics to a chunk, they first select or create a topic in the Topic List (since the window for Topics is green, the Significance Measure becomes green.

If the user is rating Participants, which lives in the blue field, then the same Significance Measure turns blue.). The user applies a sliding bar to set the rating, moving it from bottom (unsaturated colour and numerically low&emdash;0) to the top (saturated colour and numerically high&emdash; 10). The result of this rating is displayed visually, by listing the item in relation to its significance, and numerically as well (see above).

Ben Schneiderman describes a similar graphic used in an educational program designed by Christopher Ahlberg for conducting what he calls dynamic queries. The user changes the values using a slider.

A slider serves as a metaphor for the operation of entering a value for a field in the query. Changing the value is done by a physical action - sliding the drag box with a mouse - instead of entering a value by keyboard. By being able to slide the drag box back and forth and getting immediate query results, it is possible to do dozens of queries in just a few seconds....The results of the query are displayed in a graphical format near the sliders (Schneiderman, 1992, p. 669).

Clearly, the impetus to use dynamic graphical representations has begun.

3.2.2 Hue and Saturation

As illustrated above, icons may change colour in order to illustrate changes in the content or functioning of a hypermedia knowledge base (Goldman-Segal, 1993). When using the Significance Measure in Learning Constellations 2. O, there is a change in the colour saturation, connoting more or less significance. As grayish pixels dilute the colour, the viewer understands that the significance is lowered. Colours are also used in the lists of various chunks that act as links to other chunks. Annotations, authors, participants or actors, and topics are all separately colour-coded. To find the most significant topics (blue chunks) in a cluster of data or constellation, the user simply looks for the most saturated blue and its position in the list.

Colour can also indicate functionality. For instance, many pull-down menus in standard operating systems show operative functions in black on the list where inoperative functions are in grey. Another application could be to fade a trail to show temporal paths through the space. For example, if a user were tracking another user's trace or audit trail, that user could follow iconic footsteps going from node to node. The longer that it has been since the user accessed that node, the more the footstep might fade. Perspective could be used to add a spatial distancing from the main paths.

3.2.3 Focus on Motion

There is no reason why icons should not move. Movement of icons may be within the icons as in the case of Motion Icons called MICONS™. The term MlCONs was first coined by Russell Sasnet (1986) and then built by Hans Peter Broadma in 1989. An example of a micon is the postage stamp movie used in Elastic Charles, a hypermedia journal designed by Glorianna Davenport and her team about Boston's Charles River. Motion within the icon signals the changing nature of the information contained in the node designated by the icon. The micon itself displays moving images. These micons play continuously in loops and they

...indicate a link to full motion video that has a distinct visual signature. When linking from video to video a micon appears on the video screen as the link it represents, becomes relevant and disappears once the link is no longer relevant (Broadma and Davenport, 1 990).

As the user navigates down the digital map of the Charles River, micons appear as related links become possible.

As Broadma and Davenport have illustrated, icons should signal to the user the kind of link it is. That is, a slightly larger than icon-sized screen element may display a video image that describes the kind of video information contained in the node. Moreover, motion icons, micons, should appear as they become relevant and disappear or 'face' into the background as they become less relevant. This is another kind of motion.

3.2.4 Focus on Position

Icons as node designators may also change location on the screen as the relationship among nodes changes. As more information is included, or as the similarity or linkages between nodes changes, so too should the spatial orientation of the nodes. Suppose that you were using a graphical browser, such as the one in Figure 2 from a hypertext on hypertext and hypermedia Jonassen, 1989). As the hypertext grows, new nodes are added, causing a realignment of the former nodes (Fig. 2b).

A number of mathematical models and semantic networking tools are available for determining the current alignment of ideas in a knowledge domain Jonassen, 1991).

Another example of positioning uses the idea that images can be piled one on top of the other into stacks, somewhat like a flip chart with the edges showing (Elliot, 1992). Links with video streams tend to be problematic to display on a single screen. The problem is: if there are many video links, how does a user know where to look? Elliot has built an interface called the Video Streamer which lines up the frames so that only the edges show. His approach is to use the single desktop screen.

The video streamer positions frames of digital video sequentially in front of each other with a slight offset from frame to frame; visually this appears as a three dimensional extrusion of the video stream in time which emphasises differences along the side and top edges of the adjacent frames. In this way the video streamer helps us to see characteristics between frames and across shots such as transition types and cutting rhythms (1992, p. 1).

Elliot is also developing a way to move through the layers as if they are transparent, and being able to select features that are common. This use of moving icons (micons) that are the video stream rather than being pointers to a video stream may change the way we conduct dynamic queries in the future.

4. Types of Icons

In the following section, we will describe three types of graphic icons-static, dynamic, and animated-according to the categories of attributes described in Figure 1.

4.1 Static icons

Static icons are those which are designed, usually using abstract graphics and placed somewhere on the screen. They are technically not Dynamlcons. Static icons typically remain in their appointed location, and they do not change form, shape, colour or any other attribute. They are used in a consistent manner, because they have a consistent and reliable function and appearance. If reliability is most important in a hypermedia system, and it frequently is, then static icons may be appropriate. Despite their lack of dynamicity, we will describe a range of static graphics.

a) Describe/Representational

This type of icon provides a commonly interpretable representation of the information accessed by that button. For instance, if you were constructing a hypermedia document on an election and were organising the information around issues, the user could click on a picture about the candi-issues

b) Describe/Abstract.

This type of icon would use abstract images to convey the content that would be available. For instance, the first icon (a) might indicate information about food or nutrition; the second, something written as opposed to graphic/sound information or something spoken (see sound icon and speech balloon icon in 2.2.2), and finally, the third icon, information about plane schedules or transportation.

c) Describe/Symbolic.

Commonly used or easily understood symbols may also be used to convey the type of information available in a system. For instance, the first (a) uses the international symbol for information, which could be made available to a user. The second (b) could indicate the court's perspective, a ruling, or information about justice. Finally, the third (c) would indicate a royal perspective, or information about royalty.

d) Function/Representational

Probably the most common use of icons, especially representational ones, is to designate system functions, like (a) cut, (b) paint, and (c) push the button. This type of icon describes the function or purpose provided by the program, knowledge base, or system. These are very descriptive, easy to use, and part of the repertoire of most computer users.

e) Function/Abstract

Again, very common symbols are used to designate activities that may be supported by a system. These are typically interpreted as going forward or back (a), and examining a topic very closely (b). These meanings may not be immediately obvious in a hypermedia system and may have to be taught. Probably the most common abstract messages defining functions are those in (c).

f) Function/Symbolic

Common symbols can be used to indicate available functions, purpose, or perspectives. The first (a) would indicate the female perspective on an issue. The second might be used to ask for a prescription from a program. For instance, in an adaptive learning system, the learner might want to know how many practice items should be completed, asking the program to prescribe an appropriate number. Finally, if you needed a higher level of help, you might access a prayer.

g) Inference /Representational

The result of an experimental action taken in a system might be succinctly communicated by this image:

h) Inference/Abstract

Abstract images can also be used to indicate inferences or implications. In the first example (a), time may be running out. The second (b) would indicate that an idea was good, and the third could provide an indication of what might happen if a certain course or strategy were selected.

i) Inference/Symbolic

These are very difficult to conceptualise and very unlikely to be needed.

4.2 DynamIcons that changeform, colour, or position

Dynamic icons (or DynamIcons) are those that, for a variety of reasons, change shape, form, colour, position, or any other attribute while the hypermedia document is being used. Generally, though, we believe that they are of two types: first, those changing in form, shape, colour, or content, and second, those changing position on the screen. We will attempt to provide an example of each type for each of the types of graphic illustrated in Figure 1.

Symbolic reasoning requires that users abstract and synthesise meaning from a group of ideas or objects. Most often, this meaning is socially negotiated by the group using the symbols, such as through the learning of reading and writing. Clearly, symbolic reasoning is more difficult. Therefore, using symbolic icons adds to the cognitive load of the program, since learners must invest some cognitive effort in learning the meaning of the symbols. This would argue for the replacement of some static icons with more dynamic or animated icons. Symbolic reasoning is a dynamic intellectual process, so we suggest using dynamic icons to replace static images in icons in order to facilitate that processing.

We now present some verbal and visual descriptions of examples of the different types of icons in imaginary hypermedia systems. For the types of icons, you are referred to Figure 1.

a) Describe/Representational. Assuming that each icon were a slide screen of some size, pictures of the contents of a particular node could be projected onto that screen. The pictures are representational examples of the concept being defined. For instance, in a geographic hypermedia knowledge base, an icon that cyclically shows pictures of elephants, gazelles, giraffes, monkeys, and tigers would indicate that the node contains information about the range of animals in the geographic region being studied. Images of different types of dogs inserted in an icon would convey a narrower concept (dogs) in a knowledge base on, for instance, hobbies.

b) Describe/Abstract. An icon that grows as the amount of information in a node on Harry Truman's house grows or as the number of links to that node expands would be an example of an abstract, descriptive type of icon.

c) Describe/Symbolic. Symbolic, descriptive icons are more difficult to conceive as well as to understand. Therefore, some pre-learning of these icons will probably be necessary. Suppose that a node icon cycled through various types of formulas, such as math's formulae, chemistry formulae, or statistics formulae. The hypermedia viewer would be led to believe that this node contains technical information. If each type of formula were displayed on separate icons, then each icon would lead presumably to different interpretations of the content, e.g. a mathematical interpretation, a chemistry interpretation, etc.

d) Function/Representational. The most common example is a picture of the destination for an object. For instance, if you were browsing a node and wanted it printed out, you should be able to drag an icon for the node over to a representational icon of a printer in order to have it printed out.

e) Function/Abstract. In a multi-user access system with limited numbers of users who could access a single node, the node could fill up with people as more people were looking at it.

f) Function/Symbolic. In a networked hypermedia system that is supported by a movie server that can serve only a limited number of patrons simultaneously, the icon could be green if there was no competition for the resource, changing to yellow as it slows and to red when it is full. Another example is the Significance Measure described in 3.2. l. The higher the rating, the more saturated the colour.

g) Inference /Representational . Showing a cycled representation of a cause-effect relationship in an icon would infer the kind of information available in the node. Cycling back and forth between these images would infer that a surprise is in store if the user presses this button.

h) Inference/Abstract. A clock image that appears on the screen implies that a limited amount of time is left or that an operation or function will likely take some time. This is a very common icon for implication. i) Inference/Symbolic. In a multi-user hypermedia system in which users can encode different types of comments or arguments about a node, those arguments can be signaled by the icon depicting that node.

4.3 Animated icons.

DynamIcons may change screen position or form, but that may not be enough to convey truly dynamic relations. Icons can become animated and move on the screen in order to convey additional meaning. As Baecker, Small, and Mander (1991) say, icons need to be brought to life.

a) Describe/Representational. Icons that describe things representationally can be as simple as the pencil icon. Instead of having the static view of the pencil, one can see the pencil leaving its trace as it marks the space in which it is found. This kind of animated icon was built and used by Baecker, Small, and Mander (1991).

Another example is a video icon of a person talking who acts as a guide through the system. The earliest version of this was used in Palanque, an interactive program designed by Kathy Wilson (1988) and her team at Bank Street College, for exploring the ancient Mayan site of Palanque. The user clicks on the icon of a young boy, and he becomes a navigational guide, describing various aspects of the site.

b) Describe/Abstract. Animating icons that describe things or events in an abstract sense could be as simple as the battery icon on the laptop currently being used to write this article. As the battery runs out of power, the eight little boxes turn from filled to empty. Eight full ones means the computer will last me the length of this trip. Another example of an animated icon is the little person running on the bottom of our monitors when a very large program is loading up from a server. Seeing the icon also lets the user know that everything is being done as fast as possible. Granted these icons could be more informative, but that is only a matter of time and imagination.

c) Describe/Symbolic. This kind of animated icon is more diffficult to conceptualise. How do we animate a symbol that stands for a thing or event?We could animate existing symbols, such as a Cross, a Crescent, or a Star of David to signal to the user that something of a religious nature is about to happen. Or, easier to imagine, we could animate a spider's web to connote the intricacy in our connections. It may be that either users will need to learn new metaphorical meanings, or they will have the means to construct those that are most attune to their own conceptions of symbolic representations.

d) Function/Representational. An example of an animated icon which shows its purpose or function to the user is an animated fill icon in MacPaint, one that fills the box when it is selected. In other words, the icon teaches the user what it does by clearly showing or demonstrating its function (Baecker, Small, and Mander, 1991).

e) Function/Abstract. To illustrate this particular category, we will choose icons from a child's paint program, KidPix™. The erasure takes many forms. The regular brush icon erases with the usual hand gesture, deleting what is on the page. The more interesting erasure icons are those which cause a more dramatic vanishing of what is on the screen. One icon, a bomb shaped object, blows up the page. Another sets a theatrical stage and then shows the day and date. In fact, the child chooses from one of many ways to make the page disappear. Clearly, the user is not using a clear one-to one correspondence. Choosing the effect, for the child user, is more important than choosing the functionality. Children seem to find this animated choice delightful. We think that adults, although more concerned with task-oriented operations, will also find animated functionality more lifelike and therefore more engaging. Certainly, there are many of us who enjoy the process and result of recording our own error messages instead of hearing the standard 'alert' sounds: quack, beep or clickety clack. In other words, animated icons add choice and interest along with the potential for efficiency.

f) Function/Symbolic. Mitch Resnick has designed what we think are animated symbolic icons (Resnick, 1991 & 1992). Resnick has built upon Papert's graphic interface using turtles to help children and adults reflect upon their thinking processes (Papert, 1980) by identifying with the movements of the turtle as they program it to create mental models of their own thought processes. In Resnick's system called *LOGO (star LOGO), turtles "live" and "grow" in turtle communities. Resnick uses his animated turtles as a means of visualising how large communities of intelligent agents would interact. This approach of using graphic interfaces is highly symbolic.

g) Inference/Representational. Animated icons that infer connections are what has been planned in smart electronic manuscripts. For the electronic newspaper of the future, Nicholas Negroponte, MIT Media Lab Director, often speaks about an interface where the computer would be able to infer a range of articles you might be interested in reading by examining the choices you have previously made.

h) Inference/Abstract. Imagine a 3D desktop where icons change as data are entered. The user would need to infer the relationship, but the chart or spreadsheet would develop and grow according to the data. The user could specify what representations would best convey the message.

i) Inference/Symbolic. These icons are perhaps the most difficult to conceptualise. However, let us imagine that as we approach New grange, a pre-recorded historical tomb north of Dublin, an iconic flashing light in our car computer displays a list of other sites in the area that might be of similar interest. In other words, the icons would infer a potentially symbolic relationship.

5. Conclusion

In this paper we have argued that iconic interfaces convey more information than textual, and that iconic interfaces can be made more effective by adding dynamic qualities to the icons, such as changes in character, motion, positioning, and so on. In this paper, we have focused generally on using dynamic icons in hypermedia interfaces. While we have concentrated on icons in interfaces, many of these ideas could generalise to other types of graphic in hypermedia and multimedia as well.

5.1 Prospects and Problems of Dynamicons

User interfaces are typically confusing, especially to novices. Dynamic icons that change location or appearance may, at the least, be distracting for the user. Adding the complexity of dynamic icons only increases the information processing requirements for the learner, perhaps to the point of information overload. A researchable issue is the level of complexity that can be added to any hypermedia interface before users performance is impaired or before the utility of the knowledge base is limited by the complexity.

We believe that symbolic icons, in general, would require that new conventions be developed and learned by users of any interface with DynamIcons. Systematic use of symbolic icons, especially dynamic ones, requires a new way of designing and reading 'interfaces. This, in turn, may entail more effort and a learning curve for users. Future research examining the use of these more complex graphical interfaces will be needed before we are able to depend upon their usefulness. We hope this paper will provide a structural model for future investigations where categorisation of icons is needed.

6. References

Baecker, R., I. Small, & R. Mander. (1991)

Bringing Icons to Life. Proceedings ACM CHI'91.

Broadma, H. P, & G. Davenport. (1980).

Creating and viewing the elastic Charles -A hypermedia journal. In R. McAlesse & C. Green (Eds)Hypertext: State of the Art .

Elliot, E. (1992).

Multiple views of digital video. Unpublished paper. Cam- bridge, MA: MIT Media Lab's Interactive Cinema Group.

Geertz, C. (1973).

The Interpretation of Cultures. Basic Books: NewYork.

Goldman-Segall, R. (1991).

A multimedia research tool for ethnographic investigation. In I. Harel & S. Papert (Eds.) Constructionism.. Norwood, New Jersey: Ablex Publishers.

Goldman-Segall, R., Marcovici, M., Halff, L, & Flinn, C. (1993).

Learning Constellations 2.0 [Software].Vancouver, BC: MERLin Lab, Faculty of Education, University of British Columbia.

Goldman-Segal, R. (1993).

Semantic layering tools for video analysis. Proceedings EdMedia'93.

Jonassen, D.H. (1989).

Hypertext/hypermedia. Englewood Cliffs, NJ: Educational Technology Publications:

Jonassen, D.H. (1991).

Representing the expert's knowledge in hypertext. ImpactAssessment Bulletin.

Mannes, S. June, 1985).

Pushing the Picture-Perfect: Smash that Icon! PC Magazine, p.64.

Papert, S. (1980).

Mindstorms: Children, Computers and Powerful Ideas. New York: Basic Books.

Resnick, M.(1991).

Beyond the Centralized Mindset. Proceedings of The In- ternational Conference on the Learning Sciences. Evanston, IL.

Resnick, M. (1992).

Beyond the Centralized Mindset: Explorations in MassivelyParallel Miaoworlds. PhD Dissertation, Dept. of EECS, MIT, Cambridge, MA.

Sassnet, R. (1986).

Reconfgurable Video. Masters inVisual Science Thesis, Cambridge, MA: MIT.

Schneiderman,B. (1992).

Dynamic queries: Database searching by direct manipulation. Proc.ACM CHI'92.

Smith,J. (1987).

Proceedings of Hypertext '87. Chapel Hill, NC: University of North Carolina.

Wilson, C. (1988).

Palanque [Videodiscs and Software]. New York: Banks Street College.