Feelings Stuck in a GUI web:
or
The most important tool a physicist has is the trash can. --Albert Einstein
Why do you people in the sciences always ask for so much funding for your
labs? The math department only asks for paper, pencil and trash cans. And the
philosophy department doesn't even ask for trash cans.
--anonymous
university dean
A notoriously quirky feature of working with a Macintosh computer is that for a user to eject a diskette from the computer, an icon of the disk must be dropped onto the Trash Can icon. It is well known that the guiding paradigm behind the Apple Macintosh OS is something called the DESKTOP metaphor. In the DESKTOP metaphor, the computer screen is a virtual 'desktop' with electronic 'folders,' 'documents,' 'disk icons' and a 'trash can' which are patterned after the physical objects in the physical office. The user interacts with this virtual world by using a mouse, an input device which allows one to move an arrow across the desktop and pick up an electronic object by clicking and holding a mouse button. To get rid of a physical document from my desktop, I toss it in the trash can. To delete an electronic document in the virtual world of the Macintosh OS, I simply grab the document and move it on top of the electronic trash can, into which the document disappears. Deleting a document from the computer is patterned after the action of throwing out a paper document--in both cases, place the document in the trash can. (Like the real world trash can, the electronic documents tossed into the Macintosh trash can remain there until the trash is emptied.) But while deleting a file using the trash can makes perfect sense to most users, ejecting a diskette by placing it in onto the trash can meets with reluctance and dismay. Many experienced users even confess to feeling a twinge of anxiety every time they eject a disk using the trash can, though they know from experience they are not deleting the information on their disks. I have known some users who, when faced with a need to eject a disk, prefer to shut down the system and have the disk eject automatically rather than throwing their disk into the trash can. This feature of the Macintosh desktop is so quirky and counterintuitive that a computer lab I have known had a sign on every Macintosh which explained that "in order to eject a disk, the user should drag the disk icon onto the trash icon and drop it."
Why should this action be so counterintuitive? Collins[1] explains that using the trash can metaphor to eject disks is "an example of magic that goes too far."[2] He suggests that metaphors are most intuitive to users when they are fairly literal, as in deleting a document by tossing it in the trash can. But metaphors can also be "magical;" that is, metaphors can be extended in important ways which do not precisely mimic their analogues. For example, an electronic spreadsheet can magically recalculate the sums of columns and rows on-the-fly when values in the columns are changed, a capacity spreadsheet users find useful in order to project the answers to "what-if" scenarios. Yet this capacity is a significant deviation from the way written account books work; the magic of the metaphor lies in its ability to be extended in flexible and creative ways which do not precisely mirror the source of the metaphor. Collins notes that the extension to the spreadsheet metaphor to automatic recalculation is an example of sympathetic magic[3] which surprises and delights users. The extension of the trash can metaphor to ejecting diskettes, however, is baffling rather than sympathetic magic: it is counterintuitive. What is it about the magical trash can that goes too far?
I began teaching users the Apple Macintosh OS in 1987, at about the same time I adopted metaphor as my major research interest. Both as a teacher of new users and as a metaphor researcher I have been hard pressed to come up with an explanation of magical trash cans which both gybes with the DESKTOP metaphor and overcomes the deep-seated feelings of anxiety that this usage of the trash can causes. My initial response was to assume that the users' discomfort was irrational, fleeting and would be assuaged by a little conditioning practice from which they can ascertain that ejecting a diskette does not delete information.[4] Despite practice however, many users still self-report discomfort at putting a disk icon into the trash, even to the extent of choosing a different procedure--shutting the machine down--to eject a diskette. The first step to a successful explanation is to realize that users' anxiety is real and well-founded here: Users do occasionally make mistakes that delete their documents, and therefore any action which links an object associated with deleting information (the trash can) to an action which does not involve deleting information (ejecting a diskette) naturally causes trepidation. I mean naturally in the strongest sense of the word--as biological organisms, humans experience false negatives much more saliently than false positives. Suppose we are examining unfamiliar plants in an effort to find a new food source. However, some of the plants may be poisonous. The consequences of a false negative are much more extreme than a false positive--after all, if we falsely identify a plant as poisonous, don't eat it, and it turns out not to be poisonous we will not be harmed and will survive (assuming other food sources are available). But if we falsely identify a plant as not poisonous, eat it, and it turns out to be poisonous--a false negative--we will be harmed and may well not survive. Given a number of similar problems over many generations, it just makes good evolutionary sense for human rationality to have developed a strategy of feeling more cautious about making false negative judgements than false positive judgements--a point succintly summed up as "It is better to err on the side of caution." Of course users' naturally feel trepidation at ejecting a diskette with the trash can--because it requires overcoming their deep-seated tendency to assume that the consequences will be negative (and harmful) is false. Thus it is unsurprising and quite reasonable that users shun even the possibility of accidentally deleting all their files. Further, it is quite rational that even experienced users often feel a twinge of angst when they drop a disk in the trash, for their response is deeply seated in their bodies.
A somewhat better explanation (but one initially far less palatable to me) required that I begin by accepting the users' gut feelings as fact. Although I was aesthetically drawn to the simplicity and elegance of desktop, I had to consider the possibility that something was wrong not with the users' rationality but within the metaphor itself. Under the DESKTOP metaphor, it made no sense to pattern ejecting a diskette after the action of throwing a diskette in the trash, for we normally re-use a diskette once it is out of the system. Further, why even have the user eject the diskette electronically--wouldn't a simple manual eject button suffice? Such a solution seems much more consonant with what we do when we put the disk into the drive--which after all, is a mechanical action taking place in the physical world. Why not just keep ejecting diskettes physically? Perhaps the Macintosh designers had made a fundamental design flaw. This explanation required that I admit the magical trash can was fundamentally inconsistent within the DESKTOP metaphor. I was unhappy and reluctant to put it so directly to new users. Part of the magic of the Apple desktop lay in the fact that users could grasp the basic metaphor system and then infer what to do in a related situation and act accordingly, thus shortening the learning curve. Admitting the DESKTOP metaphor had some fundamental flaws discouraged its application; users felt betrayed by the interface and lost the trust necessary to learn by doing.
Struggling to keep this trust and still provide a reasonable explanation of magical trash cans, I took the pedagogical approach that the DESKTOP metaphor was not really the foundational metaphor system, despite Apple's pronouncements. Instead I suggested that the DESKTOP metaphor was only part of a larger story. Beneath the DESKTOP metaphor there was a more abstract PHYSICAL WORLD metaphor in which the mouse was used to grab objects and move them. Ejecting a diskette, I suggested, was removing an object from a system and just as the trash can served as a way to get unwanted documents out of their office, the trash can served as a way to get a disk out of the system. To make this point hit home, I asked users if they really had trash cans on top of their desks, as the Apple desktop did. I pointed out that the magical trash can was consistent with this higher-level PHYSICAL WORLD metaphor even though it was inconsistent with the DESKTOP metaphor, and so it inherited the functionality of removing objects from view. However, my explanation did not prove as helpful to users as I had intended. Rather than refocusing attention onto a consistent underpinning of a more abstract computer as PHYSICAL WORLD metaphor, pointing out any inconsistencies between it and the DESKTOP metaphor had the opposite effect. Users abandoned trust in the rule of metaphor completely; they would not readily generalize on its basis to related cases. Again, the explanation required something vital to spirit of the Macintosh design be tossed in the trash can.
Frustrated, I turned to technical explanations. It made good sense, I suggested, for the designer to require that the user drop a diskette onto the trash can icon to remove it from the desktop because the trash must be emptied[5] (and electronic documents in it deleted) whenever a disk is ejected. If any of the files to be deleted was on the diskette and the trash was not emptied, ejecting the diskette would require that files either be automatically deleted or left intact since there was no way for the Mac OS to know what was happening to the disk once it left the system. Asking my students to consider the designer's point of view, whatever was going on with the trash can had to be finished before a disk could leave the system. In short, the system needed to perform a little "housekeeping" every time a disk left the system, and the designers simply wanted to remind us of the consequences of our actions by making us use the trash can to eject diskettes. The introduction of the designer's point of view had a similar dampening effect on users' trust in metaphor, however; instead of naively viewing this computer as an extension of their desktop, they began to attribute an other's consciousness to the computer--in short, they began to conceive of themselves as having a conversation with their computer, rather than conceiving of the computer as another tool to do their work. Once again something remarkable about the uniqueness of the Macintosh DESKTOP metaphor vanished; this time into the seething hell of conversational interfaces.
I say that conversational interfaces are a seething hell not because I profoundly dislike them, but because from a pedagogical and aesthetic perspective they slow down users' ability to learn and work with computers. Rather than naively working with the interface and feeling out its potential, users puzzle at every response the computer screen gives them, as if they were learning to understand and speak a foreign language. Although attributing consciousness to the computer is probably an inevitable conceptual crutch for users, I believe we work best when we can forget about the computer's (or the designer's) point of view and concentrate on expressing ourselves and getting our work done. Though every interface is at least somewhat conversational in nature--the Macintosh has its 'dialog' boxes--I want to suggest that another important tension in interface design is between feeling-based and conversation-based interfaces. Feeling-based interfaces are aesthetically designed in that they strive to provide common ground for the user to have correct intuitions. Conversation-based interfaced are well designed to communicate information about the internal states of the computer, but do poorly at providing common ground for users to have intuitions.
The answer to the question "What is it about the magical trash can that goes too far?" requires that we realize feelings are not secondary, epiphenomenal responses to user-interfaces. Instead, feelings are constitutive of good interface design. The magic of the Macintosh trash can "goes too far" because it violates users gut feelings. No further explanation about how the inconsistencies can be resolved will assuage this fact for users because the felt response to a user interface is not epiphenomenal. Instead of educating people out of their feelings, the challenge for feelings-based interface design is to rethink the magical parts of the interface. Assuming it's necessary to have an electronic method in order to keep the Macintosh OS informed of the contents of the disk drive, Collins suggests that the magical trash can be replaced by a magical electronic analogue of a manual eject button that could be dropped onto disk icons. However, I don't find his solution aesthetically elegant. The magic of a trash can has to do with its being a portal to the beyond in the PHYSICAL WORLD metaphor--the beyond of the landfill, the beyond of the electronic bit bucket, and the beyond of the world outside the computer. A better solution to the problem might be to drop the diskette onto a disk storage box icon rather than into a trash can. Or why not simply have the trash can metamorphisize into a disk storage box when the diskette icon hovers above it? The trash can already bulges when it is full of files to be deleted. If my memory serves me, it was somewhere between system three and system five that the trash can began to change color when an object--a disk icon or file icon--was moved near enough to it to be dropped into it. To me, Collins' suggestion makes some aesthetic sense within the PHYSICAL WORLD metaphor, but metamorphizing trash cans makes some sense in both the PHYSICAL WORLD metaphor and more sense in the DESKTOP metaphor--you simply drop the disk into a storage box. But the acid test of any solution to the magical trash can problem is to build it and then ask users how they feel about the solution.[6]
My purpose in this long introduction is not to further dent the much maligned Macintosh trash can, but to introduce via anecdotes the tremendous challenge metaphors pose to interface design. I see not only a tension between the literal and magical qualities of metaphor, but a tension between the users' feeling that the computer is an extension of their bodies and believing that it is an "other"--a sentient being with a consciousness of its own (and usually a malevolent consciousness at that). The magical qualities of the DESKTOP metaphor are inextricably bound up with the users aesthetic feel of the system. Somehow their sense of the aesthetics of the metaphor system is disoriented by the magical trash can; they feel this deep in their bodies. I have deliberately ignored the designer's viewpoint in favor of the user's because in the rest of this paper I will focus on the bodily feelings metaphors evoke. I believe that it is not enough to say merely that metaphors evoke feelings, but that it's necessary to try to tell a story about what these patterns of bodily feelings are. To tell this story I will turn to recent work in philosophy and psychology on the embodiment of metaphor. In short, I argue that it is time to bring the users' bodies back into interface design.
The first theoretical obstacle in the interface design literature is what I call the tendency to disembody users. This tendency is deeply rooted in our common-sense picture of the way the mind works, a metaphysical picture which we have inherited from the philosopher Rene Descartes and which underlies much psychology and cognitive science. Descartes' picture works as follows: human beings have minds and bodies, but they are radically different kinds of things. Whereas bodies are physically extended in space, minds are not. Though we speak about ideas as mental objects and talk about minds as if they could contain these objects, that is only a matter of how we happen to express ourselves. These PHYSICAL WORLD metaphors are just one way we have to describe metaphysical things such as minds, ideas, concepts and souls. Ideas have an existence (albeit a non-physical one) outside our own minds--in that they are public and shared features of the world we can just "get" or "grasp." It is because of the non-physical existence of ideas and minds that you and I can truthfully say, after we reach a moment of genuine understanding, that we have "the same" idea in mind. Though we have to dress our ideas up in language and send them across the gulf between our bodies, the language we use is optional and incidental to the ideas they express. Since we are both rational creatures, our minds can strip away the extraneous language and we just "see" the same ideas by the light of pure reason. Anything bodily just interferes with the clarity and distinctness with which we perceive ideas; thus subjective emotions and feelings are eschewed in favor of dispassionate objectivity. Metaphor, on this view, is just a conduit for the transportation of ideas between minds.
The upshot of this for interface design is that Descartes' picture of mind is the one we ordinarily use to communicate on an everyday basis. In the course of our ordinary conversation we say to each other: "Do you get it?", "I'm not sure we are thinking of the same thing," "Try to put the ideas a little more clearly next time," "Your anger kept me from seeing what you meant," and "Can we get your view on the Jones proposal?" In thinking about interface design, we often fall into the Cartesian trap of trying to figure out how to transfer the designer's conceptual model for the user interface from inside the designer's mind into the user's mind. On this view metaphor is just another way, possibly a very helpful way, to transfer into the users mind the model of the program the designer thinks the user should have. Ultimately the user will learn to think of the user interface in almost the same way as the designer does; they will converge on the same model. The metaphor just falls away like scaffolding once it has done its work; the designer and the user have the same model and the same ideas in mind when they use the user interface.
I believe this picture to be fundamentally wrong, both as a theory of mind and as a theory of interface design. Minds are not just attached to bodies; minds are embodied. The problem of user interface design is not just the problem of finding the right metaphor to carry the designer's model into the users mind; but about ways in which the metaphor shapes the design process itself. Metaphor is not about simply transporting a system of ideas from the designer's mind into the users; it is constitutive of both our theories of minds and our software. Without metaphor there could be no software and no theories of mind. Human beings are fundamentally metaphoric animals, and all our creative intellectual endeavors (including both software and philosophy) are constituted by the patterns of bodily feelings which motivate metaphors. The metaphors we use to understand ideas, minds and user interfaces are not separable from the "things themselves." There are no minds which are metaphysically distinct from bodies, and there are no ideas or user interfaces which are metaphysically distinct from bodily metaphors. Successful user interface metaphors tap into a reservoir of bodily feeling on the part of the user and successfully exploit our embodied knowledge. The problem of disembodied users is that we ordinarily think of user interface design as if the users were disembodied minds when they are not.
One of the more surprising consequences of feeling-based interface design is that users may sometimes understand the model better than the designers do. Consider the case of the Macintosh clipboard. One of the inconsistencies between the electronic clipboard and a physical clipboard is that the Macintosh clipboard can only hold one item at a time, while a physical clipboard can hold a number of items piled on top of one another. The inconsistency comes up only when the user wants to retrieve something other than the last thing cut to the clipboard. On a physical clipboard we can simply rummage down through the pile of papers till we find it, but on the original Macintosh clipboard we were simply out of luck. The solution to this problem was a shareware add-in[7] called the Scrapbook. When something vanished from the Macintosh clipboard with Scrapbook installed, it vanished into the Scrapbook. Depending on the amount of memory devoted to it at installation time the Scrapbook could hold a number of cut items, and users could rummage through their Scrapbook to find what formerly would have been "lost" items. The point of this example is surprising: sometimes users feel an inconsistency in a metaphor and design a solution without any input from the system designers. But if the users can feel how the user interface should work even better than the designers, the traditional picture of metaphor as the mere conduit of the system model from designer's mind to the user's mind is inadequate.
This is not a clarion call to reform our language and ways of thinking, but it is a clarion call to think about how people have bodies and how their bodies, not just their minds, respond to and learn from the computer system. In the case of the electronic clipboard, paying close attention to how people's bodies interact with physical clipboards produces a useful extension to the electronic clipboard. Avoiding the temptation to disembody users is difficult, but focusing on feeling and metaphor as constitutive of user interface rather than incidental to it is the first and most important step.
My aim in this section is to begin to give a logic of feeling. Such a logic differs from a logic of objects because it is inherently subjective, not objective. Generally this project has been dismissed as hopeless relativism because feelings are individual and personal. More recently exploring a logic of feeling has been politically difficult, as many past attempts to find a logic of universal feeling have led to charges of sexism, racism, ethnocentrism. Sometimes these charges are justifiable, as in the cases of scholars who claimed universal patterns of feeling were exhibited only by classical European music, for instance, but sometimes they are not justifiable. My aim here is somehat less ambitious than finding universal patterns of feeling; I will settle for common patterns of feeling.
Hold your head perfectly still a moment and move this paper closer to your eyes. As it approaches, it grows incrementally larger and obscures more of your field of vision. Now consider the way document windows operate on a Macintosh. Though the Macintosh desktop is limited to the two-dimensions of the computer screen, the Macintosh manages to mimic the feeling of a paper moving closer to your eyes. When users double click on an icon to expand it to its document size, the windows zoom out at the users--an effect created by rapidly drawing and then erasing a series of ever larger rectangular outline boxes until the window reaches full size. Motion in three dimensions is mimiced in two dimensions. Similarly, when users put a Macintosh document away, the document zooms away in a series of concentric rectangles toward its location.
Zooming is more than just a nice touch however; it is one of the best examples of how user interface design can draw on common patterns of feelings. Zooming is a pattern of feelings that takes place in and through time; the realization that all feeling takes place in and through time is the most important step in thinking about users' bodies. Part of being embodied is being a creature in time, and being in time is part of what the Cartesian theory of mind and ideas as objective entities hides from our attention. Zooming windows are an extension of the PHYSICAL WORLD metaphor which draws on the common pattern of feeling we experience when an object approaches us. Though computer events usually happen fairly instantaneously by our standards, zooming windows are a deliberate attempt to make the PHYSICAL WORLD metaphor of the user interface to include both three-dimensional space and time.
Why make windows zoom? Why make an event that could happen instantaneously on the computer screen take longer than necessary simply for the sake of metaphorical consistency? Making users feel at home in their user interface is important to the development of users abilities to imagine and intuit how the interface will work. When Wittgenstein, in the most famous passage of his Philosophical Investigations, asks us to think of how games form a category based not on shared objective properties but on family resemblances, he specifically asks us to imagine "proceedings of games." It is only by imagining games in and through time that we begin to see that no single criteria defines the category, but instead that the category is made up of members who share a complex web of similarities and differences. By evoking a pattern of feelings consistent with our bodily experience, the user's imagination is stimulated to investigate how the world--or the Macintosh desktop--is a complex web of similarities and differences.
Zooming is an excellent example of one of the patterns of feeling the philosopher Mark Johnson calls image schemas. In The Body in the Mind, Mark Johnson defines an image schema as a recurrent motor or visual pattern common to the activities of bodily experience. Another example of an image schema is the IN-OUT (container) image schema. We regularly experience a number of events as containers: In the morning we wake up out of a deep sleep, drag ourselves out of bed and into the bathroom. Later that morning we might open up the newspaper and become lost in an article. The container schema is everywhere in the Macintosh desktop: With drag and drop file editing, we pick up files with the mouse and drag them into another folder or put them into the trash. We might cut the appendix out of one document and drop it into another. These imaginative metaphorical extensions of bodily patterns of feeling are what allow our minds to organize the world. In The Body and The Mind Johnson catalogs a number of different image schemas, including balance (homeostasis, symmetry) and links. Designing user interfaces well requires that designers consider how the image schemas and the patterns of feeling underneath metaphorical conceptualization can be instantiated in the user interface.
Patterns of feeling are at the heart of how the imagination works. When in music we hear the pattern of an ascending scale: do-re-mi-fa-so-__ and it is interrupted, we have a sense of anticipation for the last note of the sequence, and we will often supply it ourselves. This pattern of feeling build-up and release is basic to human bodily feeling and the basis for our intuitions and our inferences. The aesthetics of user interface design are inextricably bound up with the aesthetics of the human imagination, whether musical, visual, or kinesthetic. The infant researcher Daniel Stern argues that the activation, build-up and release of emotional tension is among the earliest and most foundational of our prelinguistic experiences. "For instance, in trying to soothe the infant the parent could say, 'There, there, there ...' giving more stress and amplitude on the first part of the word and trailing off towards the end of the word. Alternatively, the parent could silently stroke the baby's back or head with a stroke analogous to the 'There, there' sequence, applying more pressure at the onset of the stroke and lightening or trailing it off toward the end ... the infant would experience similar activation contours no matter which soothing technique was performed."[8] As infants we experience these patterns of feeling before we develop a conversational and intersubjective self, and these patterns of feeling are not unique to any one perceptual modality but have a structure which is shared across them. The well designed user interface metaphor taps into this same shared structure in as many places as possible.
In summary, I've argued that the aesthetics of user interface design requires thinking about subjective, preverbal bodily patterns of feeling. Thinking about bodily patterns of feeling instead of abstract, symbolic, and verbal communication is at the core of my distinction between feeling-based and conversation-based user interface design. The development of good user interfaces depends on careful phenemonological and psychological research on subjective bodily experiences. To write good user interfaces we must be willing to embody our magic.
(c) Tim Rohrer, 1995
Comments are welcome to Tim Rohrer (rohrer@darkwing.uoregon.edu)
More papers available at the Online Center for the Cognitive Science of Metaphor