INTERNATIONAL CONFERENCE ON ENGINEERING DESIGN ICED 01 GLASGOW, AUGUST 21-23, 2001
DIMENSIONS OF COMMUNICATION IN DESIGN Claudia Eckert and Martin Stacey Keywords: Design teams, collaborative design tools, computer supported cooperative work, design information management, taxonomies, ethnography.
1 Introduction In industry designers interact and collaborate with each other in many different ways, often on the same day. Many of these different types of interaction can be supported effectively with computer tools – but a tool must meet the needs of the situation it is used in. However research on collaborative designing and on design groupware often focuses exclusively on one interaction paradigm. This paper explores the variety of interaction in design by developing a set of dimensions for classifying design scenarios. It discusses a number of important interaction scenarios we have observed in our studies of design practice, and their implications for the development of computer support tools.
2 The Problem The term design covers a large number of different activities. Designing as a mental process is the generation of a description of a new product, through a repeated cycle of reformulating the problem, synthesising a potential solution, evaluating it, and reformulating the problem again. However, design in its broader cultural sense includes all activities involved in the generation of a complex product. Many of these activities do not contribute directly to generating the product description, but instead to establishing the requirements it must meet or testing its properties. At present there is no complete taxonomy of different design tasks. Frost [1] provides a useful classification of design products, which can be used to assess characteristics of their design processes. While there are many attempts made to describe engineering design in general in taxonomic form [for instance 2], detailed taxonomies address only particular issues. For instance Ullman [3] classifies decision problems in design; and Kaplan et al. [4] are concerned with the information requirements of tasks requiring interaction between designers. In contrast to Kaplan et al. [4], we are looking at communication activities within large design processes, where the mode of collaboration is not necessarily predetermined by the task rather than by the organisational set-up. Medland [5] classifies communication activities into four types: delegation, reporting, awareness, and problem handling (for resolving conflicts between design elements or constraints).
3 Research on Design Communication How designers communicate, and how designers could communicate, has been studied from a variety of different angles and intellectual perspectives. But discussions of collaborative
designing usually consider only a handful of activities; and support systems for cooperative design are developed for specific scenarios, when consideration of a wider range of uses could reveal a broader range of requirements and potential pitfalls.
3.1 Groupware for Engineering Design A variety of computer technologies enable designers to collaborate effectively across long distances or when involved in large projects. Information management systems that support communication through record-keeping are one important strand of development, especially important in applying methodologies such as SSADM to software development. Other developments include systems for enabling participants in face-to-face meetings to share software and computer representations of information; and support for transmitting a variety of material including video messages through time and space to colleagues working separately, to replicate for message passing the resources available in meetings [6]. There has been extensive research on computer support for collaborative designing by remote designers using shared workspaces simultaneously. Systems are in commercial use that support video-conferencing, and working collaboratively on shared CAD models across the Atlantic. Successful research systems have provided a face-view video channel as well as voice, and allowed remote users to share sketches [for instance 7], and also desktop views and near-screen hand gestures. Research on shared workspace systems has both driven and been driven by experimental and observational studies of design meetings [notably 8, 9, 10].
3.2 Studies of Communication in Collaborative Designing Research on design collaboration has largely focused on team meetings. Many studies have given a group a design brief and analysed the resulting design activities [see 11 for twenty detailed analyses of the same episode of collaborative design by various different researchers]. Design conversations almost always employ sketches, drawings, prototypes or other visual referents (but these can sometimes be imagined [12]). Communication in joint designing is multi-modal: speech, drawing and gestures are used in combination, with each channel used to explicate and disambiguate what is expressed in the others [8, 9, 10]. This multi-modal communication involves the use of argumentation strategies and rhetoric [10, 13, 14, 15], and subtle modulation of the degree of commitment with which a proposal is put forward [10, 15]. Minneman [10] points out that describing the design itself is just one aspect of design discourse. He classifies the content of design communication according to a 3-by-3 matrix: Communication can be about an artefact, or a process, or a relation (between individuals or groups, or between people and tools, rules, representations etc). It can describe the state (that something is now in), or how and why something got to be the way it is (making sense), or how something might or should develop (framing the future). There has been extensive research on how using computer technology influences how people interact in meetings. One important finding is that people will exploit ways to communicate that don’t exist in conventional face-to-face interactions. For instance by drawing or gesturing in the same place at the same time in a virtual workspace [7]. Another is that using group support systems influences what happens in meetings, but how they change what happens depends on both the technology and the purpose of the meeting; for instance decision-making is different from idea generation [see 16]. Minneman [10], Bucciarelli [13] and Henderson [14], among others, have studied large-scale engineering processes as participant observers. They report that complex designs are
developed largely through social processes of argumentation and negotiation. They view designs as arising through a process of negotiation between participants, where information is activity communicated and actively made sense of, rather than seeing it as passively flowing through an organisation. However this view downplays the role of designers working on their own and needing to communicate actively to pass on the results of their work. Henderson [14] shows that graphic representations play a critical role in structuring the design process and conveying information between people with different knowledge and responsibilities. Designers’ learned representation-specific interpretation skills determine how and how much they understand – representations mean different things to people with different expertise.
3.3 Our Research on Design Communication Between 1992 and 1997 we carried out an ethnographic study of design processes in over 25 knitwear companies in Britain, Germany and Italy, which was originally intended as a requirements analysis for an intelligent design support system. This found that the communication between knitwear designers and knitting machine technicians was a major bottleneck, and identified a variety of factors contributing to communication breakdown [17]. The most important were the difficulty of expressing designs clearly and unambiguously in the available representations, and failure to recognise communication problems. These findings have informed our studies of engineering design. In 1999 we interviewed 23 experienced designers and design managers in a study of the customisation process in helicopter design [18]. In 2000 we conducted a series of interviews in an automotive company studying design planning. Communication was not the primary target in any of these studies, but the aspects of design we studied depended on communication. While knitwear design is far less complex than engineering, with smaller teams and a much simpler product, it exhibits many of the characteristics of engineering design [17]. In knitwear design we could observe all aspects of the design process and see people communicating in different situations. It was only possible to talk to a small fraction of the participants in the engineering processes, who were all specialists in their fields. None communicated with colleagues in as many different situations as did the knitwear designers. In all the organisations we studied designers did some joint problem solving with other designers. For example all the knitwear designers in a company jointly work out a colour scheme for each season. But they develop conceptual designs for garments independently, and hand them over to technicians in the form of inaccurate, incomplete and inconsistent specifications; they only negotiate over the design when the technicians perceive that they cannot interpret the specifications accurately [17]. In the engineering companies we have studied, designs are handed over to colleagues with different technical expertise. The handovers are discussed in formal meetings, but often only by the bosses of the people doing the designing; if people knew each other, conflicts were sometimes resolved through local negotiation, but communication across organisational fields was always problematic [see 19]. What aspects of design the researcher studies is not independent of how a design process works. What gets done individually and what is done in meetings depends on the particular problem, but also on the structure and culture of the organisation. In large organisations and complex products, the complexity of the process inevitably leads to more handovers and information transmission between people who don’t know each other. We have studied design cultures in which passing on specifications and other asynchronous communication is important. Had we looked solely at communication or at processes where the major participants design together in meetings, our study would have used methods more similar to the sociological studies mentioned above and come up with more similar findings.
4 Dimensions of Communication Situations The situations in which designers interact vary in a large number of ways. The dimensions of variation we list below are not orthogonal: common situations have related values along a number of the dimensions. These different situations can create different types of communication breakdown; they influence how designers behave and what they need in the way of computer support for their collaborative activities. Form of Communication ○ Place: Participants are face-to-face vs participants are geographically remote. ○ Time: Communication is interactive in real time vs communication is asynchronous. ○ Size: Interaction between pair ↔ interaction between many. ○ Identity: Recipients are known (conversation, private notes) ↔ recipients are unknown (record keeping, subcontractors to be found, open audience). Form of Task ○ Objective of task: Generation of ideas or alternative solutions vs convergent problem solving vs decision making from alternatives vs acquisition or imparting of pre-existing information. ○ Division of decision-making: Joint problem solving ↔ negotiated handover ↔ sequential problem solving. ○ Hierarchy of decisions: Different participants’ tasks are of equal importance ↔ some tasks are subordinate to others. ○ Duration: Interactive or communicative activity is brief ↔ activity is extended. ○ Information type: Facts, proposals, specifications vs opinions or judgements or prognoses vs problem-solving strategy advice (see section 3.2). ○ Time pressure: Task is time critical ↔ task is not urgent. Subject Expertise ○ Equality of expertise: Participants have equal levels of expertise ↔ Some participants are more knowledgeable than others. (One important interaction type is apprentice consults more experienced colleague.) ○ Balance of Expertise: Participants have shared expertise (and use the same concepts and can interpret each other’s terms and representations) ↔ participants have complementary expertise. ○ Mental representations: Participants conceptualise topic in similar terms ↔ different terms. ○ Familiarity: Participants know each other ↔ participants cannot make assumptions about others’ knowledge. ○ Context: Participants share contextual information ↔ participants have different (or no) knowledge of the context. Tool Expertise ○ Competence with groupware: Experienced frequent user (skilled and comfortable with using the medium) ↔ novice or infrequent participant. Organisation ○ Hierarchy: Participants at same level of hierarchy ↔ participants have different status. ○ Interest: Participants inside same company ↔ participants working for different companies.
○ Security: All information can be shared ↔ some information must not be shared (for instance in dealings with suppliers, or with people without security clearance). Representation of information ○ Medium: Speech, gestures, hand drawn sketches, hardcopy printouts of text files or CAD models, web pages, shared files, physical objects such as prototypes… ○ Form of information: Text, data plots, tables, diagrams, code, photographs… ○ Notation: Some fields have alternative notational conventions for the same information.
5 Interaction Scenarios Some of these dimensions determine most of the characteristics of an interaction situation. They define common interaction scenarios, which turn up in many different industries. These scenarios reflect typical work situations, which require their own computer support. One way to classify scenarios is by the way that the inputs to the tasks of the participants are related. •
Handover. One person undertakes a design task and finishes it is far as possible, then passes on the design to another specialist in a different aspect of the design, through a written or oral specification. The expectation is that the next person will do what is required within the specification rather than advancing the design by changing the specification. The participants are often co-located but communication is asynchronous. Later tasks are often seen as subordinate, so that two-way negotiations are excluded. For example, knitwear designers give their technicians specifications, without much discussion unless problems occur [17]. Such over-the-wall sequential design processes are still quite common in engineering, especially when designs are handed over to suppliers or contractors.
•
Joint Designing. A group of people work on one problem together. Typically they work at the same time in the same room. Individuals might work on parts of the problems, but they have easy access to each other and discuss issues as they occur. Joint designing is typically done by groups of people with similar expertise, who are solving a problem that concerns all. The team members usually share a lot of background knowledge and awareness of context, and often get to know each other well. They can talk to each other spontaneously and get rapid feedback on their communication acts [see 8, 9, 10, 15]. For example knitwear designers work out colour schemes as a group, because they all use the same colour scheme. In engineering designers often work jointly during conceptual design when even a complex problem is addressed by a small group.
•
Interface Negotiation. In concurrent design various people from different fields of expertise work on a design at the same time. Their tasks have mutually dependent inputs. To achieve full concurrency, they need to work with estimates of parameter values and negotiate to achieve mutually consistent solutions to their individual problems. In reality most processes give priority to some tasks and decisions, and stagger the beginning of the tasks. It is well recognised that concurrent design processes work best with colocated dedicated project teams. Communication occurs informally, through one-to-one conversations as well as in meetings.
Episodes of interaction can have a variety of purposes, even within a meeting with a different primary purpose. The types of discussion listed below can be about most of Minneman’s nine classes of subject matter (see section 3.2).
•
Request for information. Designers frequently find they need more information, and usually their main source is their colleagues. Pure information request is more likely to occur in design handover or concurrent situations than in joint designing sessions.
•
Negotiation for clarity and negotiation of constraints. The participants in a discussion must make sure that they understand each other’s positions – that is, achieve compatible interpretations of the situation. This often requires understanding the constraints the others must meet, to understand what the constraints on their own activities should be. So negotiation for clarity often leads to a negotiation over constraints. This is particularly important when designs are handed over (not necessarily in a linear process) from one specialist to another who is doing an equally important task independently.
•
Idea generation. In many design process that are essentially sequential, idea generation is undertaken as a joint activity in a meeting, because designers need each other’s input before committing time and resources to any particular solution. Designers often reuse ideas from past designs or other sources; how much they refer to visual props depends on how much they need to explain ideas with reference to their sources.
•
Conflict resolution. Meetings are often set up to resolve conflicts between elements of a design. This is typically done through real-time discussion. Conflict resolution situations vary according to whether there is an authority capable of arbitrating or imposing a decision on conflicting parties.
•
Decision making. Much design comprises an exploration of possibilities followed by a decision on which avenue to follow. Decisions need to be made about what trade-offs to make, and often in conflict resolution, as well as about concepts. If individuals make decisions on their own, they have to justify them (see below). In meetings decision can be made jointly or by individuals higher up in the hierarchy.
•
Justification. Designers must often justify their solutions or decisions, either orally in meetings or in reports. The recipient cannot be assumed to have the same knowledge as the person who has to justify the solution. Justifications may be made to equals, bosses or outsiders; and the explanations must be pitched to the recipients’ understanding. Specific justifications are often necessary in handover activities.
Each individual engages in most of these communication situations as part of their normal work, possibly with the same people. Designers use different channels on different occasions to convey different kinds of design information. For example a designer might engage in joint problem solving with his boss in a face to face meeting involving conversation and sketches, when they are negotiating over the constraints on a particular problem. The designer then works on his own using a CAD system. When he has a question has sends an email message or picks up the telephone. Later he has to return to the boss to justify the design that he has come up with in another face to face meeting.
6 Implications for Engineering Groupware No single approach to supporting communication is sufficient to handle the richness and variety of possible communication acts. Hence we should be suspicious of any assumptions that supporting formal meetings, or record management and transmission, or real-time interaction, is going to meet all the information transmission needs of a distributed organisation. However we can draw some lessons from studies of design communication.
•
As almost all design conversation refers to visual props, shared workspace systems are vital for distributed communication. Such systems should support sketching and gesturing as well as the use of formal documents and CAD models.
•
Much communication centres on contextual information, especially about past designs and competitor products, and relies on all participants being familiar with this context. Support for rapid retrieval and display of archive information could make shared workspace systems much more effective. Here support for communication overlaps with long-term information management.
•
Quick, casual, low-effort interactions are at least as important as organised meetings and formal information management – restricting them is potentially catastrophic. Hence it is essential for meeting support systems to support easy and rapid initiation of brief unscheduled interactions, especially if integrating informal interaction with information management activities is a desired objective.
•
Graphic representations and models not only communicate information across divisions of expertise and responsibility, but also play an important role in structuring design processes [14]. Access to and control of representations through information management systems will affect the design process. If using a particular model or representation is essential, how can it be used in communication scenarios for which the system was not designed?
The range of alternative situations in which designers interact indicates that there is scope for computer support for cooperative designing that goes beyond efficient document retrieval or video conferencing technology and shared workspaces. Important issues include translation between alternative notations and graphic representations, enabling the less-informed members of a group to establish the context of the discussion quickly and efficiently, supporting awareness of other participants in remote interactions, and maintaining security. Acknowledgements The authors are grateful to the EPSRC, GKN Westland Helicopters Ltd and Lotus Engineering Ltd for their support of this project, and to the engineers who participated in it. References [1]
Frost, R., “A Suggested Taxonomy for Engineering Design Problems”, Journal of Engineering Design, Vol. 5 No. 4, pp. 399-410, 1994.
[2]
Ullman, D., “A Taxonomy for Mechanical Design”, Research in Engineering Design, Vol. 3, pp. 179-189, 1992.
[3]
Ullman, D., “A Taxonomy for Classifying Engineering Design Decision Problems and Support Systems, Artificial Intelligence in Engineering Design, Analysis and Manufacturing, Vol. 9, pp. 427-438, 1995.
[4]
Kaplan, S.M., Tolone, W.J., Bogia, D.P. and Bignoli, C., “Flexible, Active Support for Collaborative Work with ConversationBuilder”, Proceedings of Computer Supported Cooperative Work ’92, ACM Press, Toronto, Canada, 1992, pp. 378-385.
[5]
Medland, A.J., “Forms of Communications Observed During the Study of Design Activities in Industry”, Journal of Engineering Design, Vol. 5, pp. 243-253, 1992.
[6]
Minneman, S.L. and Harrison, S.R., “The DrawStream Station: a tool for distributed and asynchronous chats about sketches and artifacts”, Proceedings of HCI’99, Munich, Germany, 1999.
[7]
Bly, S.A. and Minneman, S.M., “Commune: a shared drawing surface”, Proceedings of the Conference on Office Automation Systems, Boston, MA, 1990, pp. 184-192.
[8]
Bly, S.A., “A Use of Drawing Surfaces in Different Collaborative Settings”, Proceedings of Computer Supported Cooperative Work ’88, ACM Press, Portland, OR, 1988, pp. 250-256.
[9]
Tang, J.C., “Listing, Drawing and Gesturing in Design: A Study of the Use of Shared Workspaces by Design Teams”, PhD Thesis, Department of Mechanical Engineering, Stanford University, Stanford, CA, 1989. Xerox PARC report SSL-89-3.
[10] Minneman, S.L., “The Social Construction of a Technical Reality: Empirical Studies of Group Engineering Design Practice”, PhD Thesis, Department of Mechanical Engineering, Stanford University, Stanford, CA, 1991. Xerox PARC report SSL-91-22. [11] Cross, N.G., Christiaans, H.H.C.M. and Dorst, K. (editors), Analysing Design Activity, John Wiley, Chichester, UK, 1996. [12] Eckert, C.M. and Stacey, M.K., “Sources of inspiration: a language of design”, Design Studies, Vol. 21, 523-538, 2000. [13] Bucciarelli, L.L. “Designing Engineers”, MIT Press, Cambridge, MA, 1994. [14] Henderson, K. “On line and on paper”, MIT Press, Cambridge, MA, 1999. [15] Brereton, M.F., Cannon, D.M., Mabogunje, A. and Leifer, L.J., “Collaboration in Design Teams: Mediating Design Progress through Social Interaction”, in [11], pp. 319341. [16] Huang, W. and Wei, K.K., “Task as a moderator for the effects of group support systems on group influence processes”, European Journal of Information Systems, Vol. 6, pp. 208-217, 1997. [17] Eckert, C.M., “The Communication Bottleneck in Knitwear Design: Analysis and Computing Solutions”, Computer Supported Cooperative Work, 2001. [18] Eckert, C.M., Zanker, W. and Clarkson, P.J., “Aspects of a better understanding of changes”, Proceedings of ICED’01, PEP, Glasgow, UK, 2001. [19] Eckert, C.M., Clarkson, P.J. and Stacey, M.K., “Information Flow in Engineering Companies: Problems and their Causes”, Proceedings of ICED’01, PEP, Glasgow, UK, 2001. Dr Claudia Eckert Engineering Design Centre Department of Engineering University of Cambridge Trumpington Street Cambridge, CB2 1PZ, UK Phone:+44 (0)1223 332 662 Fax:+44 (0)1223 332 662 Email:
[email protected]
Dr Martin Stacey Department of Computer and Information Sciences De Montfort University Kents Hill Milton Keynes, MK7 6HP, UK Phone: +44 1908 834936 Fax: +44 1908 834948 Email:
[email protected]
