Rules of etiquette at the global cocktail party: issues of metaphor in CSCW design
This document considers the effects of using metaphors in systems design. It uses the work on metaphor in the literature of HCI, cognitive science and computer-assisted learning to assess the effects of metaphor use in the development and use of CSCW systems. In particular it raises the issue of etiquette in social activity and how this evolves in a computer-mediated social setting.
||Fri, Jan 15, 1993|
1. Advantages of metaphor: A rosy picture 1
2.2. Evolution of metaphors: The growth of dinosaurs 4
2.3. Metaphor: A Pandora's Box of requirements 4
5. Using the cocktail party metaphor in CSCW 8
6. Metaphor and etiquette 9
Computer data 10
There are numerous claims (eg. Carroll & Thomas 1980, Rumelhart & Norman 1981) that when designing a computer system to perform a task, a real world metaphor offers a number of interlinked advantages:
• A metaphor as an aid to learning/understanding can offer further benefits by providing 'gratuitous' information. That is, if the analogy between the real-world system and the computer system is close so that there are many additional sub-correspondences, the user may not need to actively learn about certain features because they can infer their use from the metaphor. Likewise they can use the metaphor to effect the mapping from their goals to the available operators to achieve those goals. The metaphor can also serve as a means of error correction. By using the metaphor one of the mappings from the analog to the action structure of the system becomes more meaningful, indicating the correct action.
• If the metaphor used is a near-direct mapping of existing work practice, it can offer advantages during continuing use as well as during the initial learning period. By matching the existing work environment, existing work practices can continue unhampered while benefiting from additional features offered by computerisation. An example would be the desktop metaphor of the Macintosh interface.
• Besides offering advantages, it may be that a metaphor is the only way to initially design a new system. It is very difficult to create a completely new application. Instead one uses existing computer systems or real-world systems as a source of ideas. Also, it is difficult for a user to understand the usefulness of a completely new system (all the other points refer to usability). Thus the best way can be to take an existing system and computerise parts of it. By using a metaphor of an existing successful system or practice one can build on the usability features that have evolved with the original.
2. Disadvantages of metaphors: The spectre at the feast
2.1 Cognitive Limitations: Every rose has its thorn
A metaphor can also perpetuate restrictions on activity that had a real basis in the original version but are no longer relevant for a computational version. This can include ignoring the new features of the system that do not map over from the analogical case. An example is the use of the Find command in word processors that is not available on a typewriter.
It can restrict design creativity, because one always thinks only in terms of slavishly adapting existing non-computer systems. As Goldstein (1981) notes of the development of word processors: "If we program these machines to imitate paper ... we will never know if qualitative improvements in the handling of words can be obtained".
For a complex system, a single metaphor may be insufficient to account for all the activity. This greatly increases the complexity of understanding for the user as she must now decide on which metaphor to employ in each context. There may still remain cases where all the metaphors break down (Halasz & Moran 1982). For example a computer's file management facilities may be explained by analogy to a filing cabinet. However, this analogy fails for password protection. If one expands the metaphor by describing a filing cabinet with a combination lock, one still has to account for protection of folders and individual files. As more and more special cases are added to the real-world analogue in order to explain the computational system (nested folders, links etc), it becomes increasingly baroque and less like any familiar real-world entity. An alternative is to propose a new analogue for file protection only, such as a guard who retrieves documents only when given the correct password. Now the user is in danger of reasoning about guards when she should be reasoning about filing cabinets opening up endless possibilities for errors and confusion.
Different use of terminology between the system and the analogue can lead to confusion for users. For example in the learning of probability the terms 'permutation' and 'combination' have very precise meanings. Unfortunately certain real-world analogues that illustrate the concepts use the terms with quite opposite meanings: a combination lock is based on permutations and 'perm' in football pools are based on combinations! No teacher would choose to provide students with such dreadful analogies but the danger arises that the students spot the analogies themselves leading to inevitable confusion.
As features and options are added to the computer system, the originating metaphor can get increasingly stretched as more features are provided that are meaningless or contradictory in terms of the metaphor. The question arises of whether it remains productive to talk in terms of the metaphor, or do the contradictions outweigh the advantages of employing it for beginners and users. Do old metaphors collapse or simply fade away?
At this stage probably not, but we can begin the work by setting up case studies of how metaphors, have been used and attempt to find rules for classifying successful and unsuccessful metaphors.
By analogy(!) with HCI research, we can feasibly hope not to create criteria for spotting world-beating ideas, but perhaps rules for avoiding real lemons. That is, most HCI advice is negative rather than positively prescriptive (eg DON'T have too many colours, DON'T require the user to memorise a lot etc) Similarly we can aim to produce some negative rules for metaphor construction. For example:
The spreadsheet is an example of where metaphor is used to inform the development of the system, but does not have a direct effect on ease of use. The fact that novices do seem to find it easy to learn the general concepts of spreadsheet use illustrates that accessible metaphors are not essential for the successful learning of a system. However, by being based on a successful analogue, it inherits (and substantially adds to) the tested ease of use of the latter.
The essential features of a real world cocktail party relevant to this use of the metaphor are that it consists of a significant number (more than 10) of participants co-located. Individuals have their own personal goals which require conversation with a number of other individuals. They achieve this by forming sub-groups of two or more for interaction. These groups grow and shrink as people choose to move to other groups in pursuit of their goals. In addition to any immediate goals, the individuals may obtain useful serendipitous information that can lead to the setting of further goals.
As if all this wasn't complicated enough, there is a time dimension. Not only may the party have a definite (but usually flexible) start and finish, but individuals may leave and arrive at different times. Furthermore there may be a host/hostess who considers one of her goals to be responsible for maintaining the dynamics of the party, introducing individuals who are expected to appreciate this and perhaps encouraging the breaking and reforming of groups.
Given the goal to talk to various people and perhaps search for unspecified information, a participant must not only monitor the activity of her group, but also that of other groups in case an opportunity to talk to a particular individual arises. This monitoring can be visual, which has the advantage of not interfering too much with an existing verbal interaction, but it can also be auditory. One of the remarkable abilities of humans is the ability to selectively attend to one conversation out of many going on simultaneously. Auditory monitoring is very valuable in providing information about the content of another group's interaction, at the cost of losing information about one's present group.
This brings us to an important issues for all forms of social interaction and hence necessarily for CSCW, that of rules of etiquette. In the case of the (real) cocktail party this is complex, revealing numerous issues. At the party, conversations are not as private as normal, people are allowed to join groups, and perhaps to overhear or tune into groups as a prelude to choosing whether to join in. However, visible monitoring of other groups while participating in a group is considered bad manners as it implies minimal commitment to the group. Nevertheless monitoring does occur but in a subtle manner.
For electronic mail, the metaphor is naturally the paper mail (but also has aspects of telephony: see later). This is asynchronous and democratic: anyone may write to anyone else. Mail can generally be read faster than it is created and it is possible to filter it (with a human secretary for paper mail, or monitor programs for email).
Until recently there have been few problems with telephones either. Telephonic communication is exclusive. However, new services offer the ability to indicate to a phone user that another person wishes to speak to him. This has been advertised by British Telecom with scenarios of inessential conversations blocking the line and the advantages of being able to interrupt with an important message. The extent to which the consequences of this feature have not been thoroughly explored is illustrated by considering the potential irritation of circumstances where the situation is reversed: an important conversation gets cut short for the benefit of a (more) trivial one.
Consider the problems in cscw-cocktail-party mode. There may be restrictions on entry (admission by invitation only). But within the party, anyone can approach you and engage you in interactions. At the same time the people you particularly wish to see are also in danger of being mobbed. You can't prioritise or make appointments in this metaphor, and so whether you achieve your interaction aims can be semi-random. There need to be ways of coping with the problems of large numbers of people and excessive popularity. In the real world analogue, this is done by techniques such as exclusivity (only a select group of invitees), hiding, disguise, minders or simply avoidance. It is possible to envisage computational analogues for all of these.
Electronic mail also can be regarded as the product of a mixed metaphor. It has some features analogous to paper documents (normal mail) and others more analogous to telephonic communication. This ambiguity has contributed to the debate in the USA on the legal privacy status of email (Elmer-Dewitt 1993). A US government judge has barred the Bush administration from erasing tapes of email on the grounds that they were documents like paper. Nevertheless in general private use, the assumption remains that private message such as gossip should carry the same legal protection from third party monitoring as telephone conversations. However the Electronic Communications Privacy Act of 1986 only prohibits "outside" interception of email by a third party without proper authorisation. Within the organisation which owns and uses the system it is quite legal for supervisors to monitor employees' conversations.
The problem is that we seem to have a different etiquette for what is acceptable in a physical system compared to a computer system, and are generally more strict about computer systems. So in the hotel keyrack case, even if the computerised system had a facility to query if someone was in their room, it might be considered unacceptable to offer such a facility because it could be abused by for example tracking the position of single female residents. So a computer feature that offers identical functionality to a physical feature may be scrutinised more carefully and considered unacceptable precisely because it is computerised.
In some countries (such as Sweden) this manifest anomaly has led to the introduction of laws of privacy that apply regardless of the recording medium. What is significant here is that it is the act of computerisation that changes the etiquette of acceptability that is then applied back to existing systems, rendering them now unacceptable.
Even though a metaphor may seem to be useful for a CSCW application, the application that embodies it may be unacceptable or unusable because it violates certain rules of etiquette. These rules for the computer system may not be the same as those that apply to the real world system. Just as these rules of etiquette can arise, so they can be revised with familiarity. However the system would need to offer substantial benefits for the user for her to overcome her reluctance to break existing etiquette. The continuing incremental development of a CSCW system participates in a co-evolution of user requirements and etiquette.
Cool, C., Fish, R. S., Kraut, R. E., & Lowery, C. M. (1992). Iterative design of video communication systems. In J. Turner & R. Kraut (Ed.), CSCW '92, (pp. 25-32). Toronto: ACM Press.
Elmer-Dewitt, P. (1993). Who's reading your screen? Time International, 141(3) 30.
Fish, R. S., Kraut, R. E., & Chalfonte, B. L. (1990). The VideoWindow system in informal communications. In Proceedings of the Conference on Computer Supported Cooperative Work (CSCW '90), (pp. 1-11). Los Angeles, California: ACM Press.
Gentner, D. R., & Grudin, J. . Why good engineers (sometimes) create bad interfaces. Proceedings CHI 90, Seattle. 277-282
Goldstein, I. (1981). Writing with a computer. In Proceedings of the Third Annual Conference of the Cognitive Science Society, Berkeley:
Halasz, F., & Moran, T. P. (1982). Analogy considered harmful. In Human Factors in Computer Systems Conference, 82, (pp. 383-6). National Bureau of Standards, Gaithersburg, Maryland.
Ishii, H., & Arita, K. (1991). ClearFace: Translucent multiuser interface for TeamWorkstation (Research report No. NTT Human Interface Laboratories.
Lauwers, J. C., & Lantz, K. A. (1990). Collaboration awareness in support of collaboration transparency: requirements for the next generation of shared window systems. In CHI Proceedings, (pp. 304-311). Seattle, Washington. April: ACM Press.
Lawson, B. (1980). How designers think. Chatham: W & J Mackay Ltd.
Orlikowski, W. J. (1992). Learning from Notes: organizational issues in groupware implementation. In J. Turner & R. Kraut (Ed.), CSCW '92, (pp. 362-369). Toronto: ACM Press.
Robinson, M. (1993). Introduction to the "common artifact" (Working Paper No. COMIC-SF-4-1). Sageforce Ltd.
Root, W. R. (1988). Design of a multi-media vehicle for social browsing. In Proceedings of the Conference on Computer-Supported Cooperative Work (CSCW '88), (pp. 25-38). Portland, Oregon: ACM Press.
Rumelhart, D. E., & Norman, D. A. (1981). Analogical processes in learning. In J. R. Anderson (Eds.), Cognitive Skills and their Acquisition (pp. 335-360). Hillsdale, NJ.: Erlbaum.
Tang, J.C., & Minneman, S.L. (1991). VideoWhiteBoard: video shadow to support remote collaboration. In CHI '91, New Orleans.
Whittaker, S., Brennan, S. E., & Clark, H. H. (1991). Coordinating activity: An analysis of interaction in computer supported cooperative work. In ACM SIGCHI Conference on Human Factors in Computing Systems, (pp. 361-368). New Orleans: ACM Press.
Wolf, C. G., Rhyne, J. R., & Briggs, L. K. (1992). Communication and information retrieval with a pen-based meeting support tool. In J. Turner & R. Kraut (Ed.), CSCW '92, (pp. 322-329). Toronto: ACM Press.