... has been adopted and elaboratored by many Scandinavian (e.g. Bødker and ..... A boundary object (Star, 1989) is a shared object of working which crosses the .... In M. S. Ackerman, Ed., Proceedings of CSCW '96, New York, ACM Press, ...
Describing Team Work with Activity Theory Phil Turner and Susan Turner HCI Research Group, Napier University, Edinburgh The proposal that activity theory might serve as a model or theory for computer supported cooperative working remains an open question. Here we demonstrate the usefulness of activity theory in elucidating a series of small group software design meetings. The structure and dynamics of the meetings are presented, as are the work processes of transformation and mediation. Finally, we show how a contradictions analysis can support the derivation of requirements on a system to support the design meetings. Keywords: activity theory, case study, cooperative working, scenarios, software design, requirements.
1
Introduction
In the last decade or so, the focus of much research activity in the broad domain of human computer systems has moved from the single user at the computer to groups of people interacting with and through a variety of technologies (Myers et al, 1996). In parallel with this shift, attention has moved from the usability laboratory to the rich context of everyday work. However, while research and development for individual human-computer interaction has been strongly informed by cognitive accounts of human information processing (e.g. Card et al, 1983), as yet no single account of collaborative working has gained widespread acceptance. This is despite the considerable body of literature in the study of work, with researchers drawn from a variety of disciplines including sociologists, anthropologists, psychologists and even computer scientists. As a number of commentators have observed, it may be this very diversity which has prevented the creation of a comprehensive and widely accepted account, theory or model of work (e.g. Rogers and Ellis, 1994) - much less collaborative work. This paper discusses the adequacy of activity theory in this domain. Our motivation stems from the observation that in recent years a number of prominent researchers have suggested that activity theory might serve as a framework for computer supported cooperative work (e.g. Bødker, 1991; Kuutti and Arvonen, 1992; Kuutti and Bannon, 1993; Kuutti, 1996; Nardi, 1996). However it remains far from clear how this might be acheived in practice. Activity theory stems from the work of Vygotsky and his students Luria and Leont’ev, and from its origins in psychology and education has recently gained ground in many other domains, including the study of work (e.g. Engeström, 1995; Engeström et al; 1997; Engeström, 1999a), information systems (e.g. Christiansen, 1996; Heeren and Lewis, 1997; Hasan, 1998) and organisational theory (e.g. Blackler, 1993, 1994 & 1995). Finally, our own recent research has included investigations of the use of activity (theory) structured scenarios in the diagnosis of IT uptake issues (Turner et al, 1998) and the use of activity theoretic concepts to structure ethnographically gathered data (Turner et al, 1999). While the word theory is misleading, activity theory does provide a description of context which can be used as a starting point from which to describe work. As Verenikina and Gould (1998) have observed, Vygotsky derived many of his ideas from the analysis of productive work, an activity that is intrinsically social. In much the same way, computer supported cooperative work (CSCW) is typically defined in terms of groups of people working together towards a common goal supported by technology. So the mapping between activity theory and CSCW appears to be very natural. However activity theory has not been adopted by the CSCW community for reasons which might include the obscurity of its terminology, what is sometimes perceived as Marxist political baggage and perceptions that it supports only description rather than a more practical 1
approach to the analysis or design of systems to support cooperative working. Our approach is not to test explicitly the premise that activity theory could be adopted as a theory for CSCW; instead we have set ourselves the more modest target of using elements of the activity theory corpus to describe team working within a series of software design meetings and from there to consider possible links to CSCW. But first we must place activity theory in the broader context of other competing accounts of work. The following tour of the main contenders is necessarily brief, and excludes those accounts or approaches which consider tasks in isolation from their context (e.g. traditional task analysis) and those which are primarily prescriptive rather than descriptive (e.g. Taylorist approaches). The categorisation employed, of course, is only very approximate and the categories are not intended to be orthogonal. 1.1 Design-driven approaches Accounts of work which fall into this category are explicitly motivated by the need for descriptions as a precursor the design of better work systems. While most are founded on a theory to a greater or lesser extent, their main purpose is to inform design, rather than to understand the workplace or develop theory as aims in themselves. Participatory design techniques (e.g. Greenbaum and Kyng 1991) generally include descriptions of work as a focus for interaction between designers and systems users. These may relate to current practice or possibilities for new systems, depending on the stage of the analysis and design process. However, while there are many participatory design techniques to support the description of any particular instance of work, rather fewer are based on an underlying model. A notable exception to this general observation is Bødker’s use of activity-structured scenarios (Bødker and Christiansen, 1997), one of the stimuli for our current investigations. The Contextual Design method (Holtzblatt and Beyer, 1996) draws upon a range of theoretical developments and field study results to provide a comprehensive method for contextually grounded software design. The method has been widely promoted as a practical means of achieving customer-centred software products. Inter alia, a number of interlocking work models are constructed: a flow model of roles, responsibilities and communications between them, a sequence model for tasks and actions; an artefact model of resources created, used and modified; a cultural model of the organisation and finally a physical model of the work environment. A recent example of a strongly theoretically based account that is nonetheless oriented towards design is Vicente’s Cognitive Work Analysis (Vicente, 1999). Cognitive Work Analysis (CWA) has its origins in the work of Rasmussen and draws on the theoretical foundations of cognitive engineering. The method is primarily targeted at those domains with complex, dynamic environmental constraints, typical examples including nuclear plants and operating theatres. The approach includes five complementary analyses: the functional structure of the work domain; control tasks which must be undertaken to achieve work goals; strategies to cope with task demands; social organisation and cooperation (broadly allocation of responsibilities for tasks and communication between roles); and worker competencies. Together the analyses provide a very full description of the work domain under study, although they may require extension for more internally-driven work such as scheduling (Higgins, 1998). 1.2 Theory-driven approaches Here we include accounts where the description and understanding of work is an end in itself, or where such an endeavour is a means to the testing and explication of theory. Ethnomethodological and ethnographic studies (e.g. Bentley et al, 1992; Hughes et al, 1994; Benford et al, 1997; Pycock and Bowers, 1996; Dourish and Button, 1998) do not rest upon a generic model of work, but rather share an underlying philosophical approach to the study or work in context. They typically produce richly descriptive accounts, reflecting their sociological and anthropological lineages (respectively), whose contents and structure will vary from study to study depending on what is observed. Indeed, a strongly ethnographic approach is fundamentally incompatible with preconceived structures or models. Similarly, ethnographers generally have 2
little to say about systems design, although there is now one study which proposes a bridge from rich ethnographic data to the systems design language UML (Viller and Sommerville, 1999) and others have suggested a role for the ethnographer as a facilitator for the designer (e.g. Twidale et al., 1994). While Hutchins (Hutchins, 1995a and 1995b) is often credited with the concept of distributed cognition, activity theorists (e.g. Cole and Engeström, 1993; Pea, 1993) have argued that activity theory itself is an account of distributed cognition (and distributed intelligence and tool use) and thus claim primacy. These claims aside, Hutchins’ account of cognition incorporates social and organisational perspectives, the primary insight being that cognitive processes and the representation of knowledge may both be distributed between multiple human actors and artefacts. It is also distinguished by its emphasis on the role of external representations (Rogers and Ellis, 1994). The elements of the cognitive system include human beings and artefacts, representations of information which may be both internal and external to the human actors, and the relationships between these elements as they work to achieve the systems goal. Many tasks involve the co-ordination of representational states, both internal and external, whereby multiple representations are combined, compared, derived from each other or made to correspond (e.g. Hutchins and Klausen, 1996). A distributed cognition approach thus offers great potential in terms of how socially shared activity achieves its goals, although there is little explicit structure to support systematic modelling. Finally, in this brief review of theory driven accounts of CSCW, we conclude with a brief note about Kjeld Schmidt’s contribution. Schmidt has been primarily interested in coordination mechanisms and what he describes as the articulation of work (e.g. Schmidt and Simone, 1996). Schmidt and his colleagues have recognised that central to successful CSCW is the issue of supporting the articulation of that work. Computer Supported Cooperative Work is defined as a interdependent group of actors working to change the ‘common field of work’ which is recursively related to the articulation of that work (i.e. the work needed to make the cooperative work, work which in itself is potentially cooperative). 1.3 In summary In summary it will be seen that each of the perspectives briefly outlined share the general shift from narrow HCI to considering technological support for human activity in real work contexts. However none have been presented or offered as a comprehensive account of cooperative work or CSCW per se (Schmidt’s contribution notwithstanding). We now move on to discuss how activity theory contributes to this endeavour, with particular reference to its potential as a model of collaborative work.
2
Activity Theory
Perhaps the first observation to be made of activity theory is that it not a theory. Instead it is better regarded as a body of thought which has been found relevant not only to psychology and education, but more recently in the understanding of work in organisations and several other fields. It developed from the ideas of the Russian psychologist Vygotsky (1896-1934) and his successors. More recently, Engeström (e.g. Engeström, 1987; Cole and Engeström, 1993; Engeström, 1995) has extended these ideas to include a model of human activity and methods for analysing activity and bringing about change. Most authors would agree that the core features of activity theory, more fully described as CHAT – Cultural Historical Activity Theory – comprise a recognition of the role and importance of culture, history and activity in understanding human behaviour. Other authors do, of course, emphasize different aspects of activity theory variously reflecting their individual research needs and the dynamic, evolving nature of activity theory.
3
The flavour of activity theory employed here draws primarily upon the contemporary work of Engeström1 which has been adopted and elaboratored by many Scandinavian (e.g. Bødker and Christiansen, 1997; Bardram, 1998a and 1998b), American (e.g. Nardi, 1996), Australian (e.g. Hassan, 1998) and British researchers (e.g. Blackler 1993, 1994, 1995). Engeström’s account of activity theory is probably the most dominant formulation in use in the study of information systems, HCI and CSCW research. In such research there is perhaps a greater focus on the role of activity per se rather than history and culture. Reflecting this, Engeström has formulated three basic principles, building on the work of earlier activity theorists, which are widely used and cited within the activity theory community. These are, in no particular order, (a) activities are the smallest meaningful unit of analysis (originally identified by Leontev); (b) the principle of selforganising activity systems driven by contradictions and (c) changes in activities (and by extension the organisation hosting them) as instantiations of cycles of expansive learning. 2.1 The structure of an activity Central to activity theory is the concept that all purposive human activity can be characterised by a triadic interaction between a subject (one or more people) and the group’s object (or purpose) mediated by artefacts or tools. In activity theory terms, the subject is the individual or individuals carrying out the activity, the artefact is any tool or representation used in that activity, whether external or internal to the subject, and the object encompasses both the purpose of the activity and its product or output. Subsequent developments of activity theory by Engeström and others have added more elements to the original formulation and these are: community (all other groups with a stake in the activity), the division of labour (the horizontal and vertical divisions of responsibilities and power within the activity) and praxis (the formal and informal rules and norms governing the relations between the subjects and the wider community for the activity). These relationships are often represented by an activity triangle. Thus activities are social and collective in nature. The use of activity triangles is widespread in the activity theory literature but it must be remembered that this is only a partial representation of an activity. The triangle should be regarded as a nexus, existing as it does in a continuum of development and learning and in turn masking its internal structure. Within the activity are the individual actions by which it is carried out. These are each directed at achieving a particular goal mediated by the use of an artefact. Actions, in turn, are executed by means of operations: lower level steps that do not require conscious attention. Activity theory is perhaps unique among accounts of work in placing such a strong emphasis on the role of individual and group or collective learning. Vygotsky’s work on developmental learning has been a major influence on the thinking of Engeström, who has extended the idea to encompass collective learning which he has termed expansive learning (Engeström, 1987). Engeström has demonstrated the usefulness of expansive learning with its cycles of internalisation, questioning, reflection and externalisation in the development of activities in a variety of domains (Engeström, 1990; Engeström, 1997; Engeström, 1999b). The drivers for these expansive cycles of learning and development are contradictions within and between activities. While this is something of a departure from Vygotsky, it has proved particularly valuable to HCI and CSCW researchers (e.g. Holt and Morris, 1993; Bai and Lindberg, 1998; Turner et al. 1999 Lui, Maxfield, and Dew, 2000). We now consider contradictions in more detail.
For a detailed account of Engeström’s formulation of activity theory please see Engeström (1987) Learning by expanding: An activity-theoretical approach to developmental research. Helsinki: Orienta-Konsultit.; or Cole and Engeström’s (1993) A cultural-historical approach to distributed cognition. In G. Salomon (ed.) Distributed Cognitions – Psychological and educational considerations, Cambridge University Press.
1
4
2.2 Engeström’s description of contradictions Activities are dynamic entities, having their roots in earlier activities and bearing the seeds of their own successors. They are subject to transformation in the light of contradictions. Those contradictions found within a single node of an activity are described as primary contradictions. In practice, this kind of contradiction can be understood in terms of breakdowns between actions or sets of actions which realise the activity. These actions are typically poly-motivated, i.e. the same action executed by different people for different reasons, or by the same person as a part of two separate activities and it is this poly-motivation which may be at the root of subsequent contradictions. The next category of contradictions are those which occur between nodes and are described as secondary contradictions. Tertiary contradictions may be found when an activity is remodelled to take account of new motives or ways of working. Thus they occur between an existing activity and what is described as a ‘culturally more advanced form’ of that activity. A culturally more advanced activity is one which has arisen from the resolution of contradictions within an existing activity and may involve the creation of new working practices (praxis) or artefacts or division of responsibilities. Finally, those occurring between different co-existing or concurrent activities are described as quaternary contradictions. From this, it can be seen that a complex and continuing evolving web of contradictions may emerge. Primary and secondary contradictions in an activity may give rise to a new activity which in turn spawns a set of tertiary contradictions between it and the original activity and this is may be compounded by quaternary contradictions with co-existing activities. 2.3 Mediation and transformation Another key concept in activity theory is the role of mediation by artefacts. Artefacts are created by people to mediate and, to an extent, control their own behaviour. Artefacts are developed over time and are local to the community which created them, so that they have an historical and cultural lineage. A further interesting observation is that artefacts are crystallisations of actions (Bardram, 1997), that is, artefacts may embody procedures and routines (e.g. a paper-based form or a plan). Wartofsky (1979) has suggested a hierarchy of artefacts which has been extended by Engeström (1990). Primary artefacts or what artefacts are used to identify and describe objects; secondary (or how) artefacts are used to guide and describe processes and procedures; tertiary (or why) artefacts are used to diagnose and explain the properties and behaviours of objects. More recently, Engeström has added quaternary (or where to) artefacts which mediate expansive learning (the interested reader is directed to Engeström, 1987 for a detailed exposition of expansive learning). 2.4 Co-ordination, co-operation and reflective communication Engeström (1997) has adopted Fichtner’s three level model of subject-object-subject interaction which distinguishes among between co-ordination, cooperation and co-construction. The routine or scripted flow of work is described as co-ordination. The script is “written rules and places or tacitly assumed traditions” (ibid, p372) and is used to co-ordinate the actions of individuals. Cooperation is that mode of interaction “in which actors actively balance and integrate their actions” (ibid, p372). Here the emphasis is on maintaining the focus of work on the shared object of that work. Finally, reflective communication (or co-construction) is “where the community reconceptualises an entire activity” (ibid, p373). To conclude this very brief introduction to activity theory and developmental work research we echo the words of Engeström and Miettinen who, in discussing the need for a unit of analysis for interacting with complex systems, conclude: To be able to analyze such complex interactions and relationships, a theoretical account of the constituent elements of the system under investigation is needed … Activity theory has a strong candidate for such a unit of analysis in the concept of object-oriented, collective and culturally mediated human activity.
p.9 Engeström and Miettinen, 1999
5
We now move to the case study, which we will use to demonstrate the application of activity theory.
3
The software design case study
The work setting for this is a software house, TTM Ltd (which is, of course, a pseudonym) and the data we present is in two forms, foreground and background. The foreground data, which is the focus of this work, comprise a video record of a series of design meetings which a particular team of three software designers held as a part of their project work at TTM. Background knowledge of TTM is drawn from the authors of this paper who are former TTM employees (the first author for 8 years, the second of 4 years). Thus it is fair to say that the authors were well acquainted with the organisational culture and working practices at TTM. 3.1 TTM Ltd TTM was involved in both commercial software development and research and development (R&D) work, with much of the latter being supported by the European Union’s Framework programmes. The company was organised into three divisions each with a divisional manager. The technical workforce of approximately 50 people was supplemented by administrative and sales staff. Working practice was quite informal despite the company being ISO quality accredited. Project teams ranged in size from as little as one person up to perhaps five or six. Each project team had a project leader (or manager – depending upon the size or importance of the project or the seniority of the assigned staff) who reported to a departmental manager. At any time, TTM might have had 10 or so research projects running. Mallard Mallard (again a pseudonym) was a European-supported whiteboard whiteboard consortium of multinational companies working collaboratively with the specific aim of improving the visualisation of evolving software products by providing Peter Roger an integrated development process and supporting tools. This case study concerns work of the team local to TTM, comprising a project manager Matt (Roger), and two software engineers members, Peter and Matt. Their particular responsibility was a configurable graphical animation workbench for the visualisation of specification and design notations (graphical animator hereafter). Our study began camera shortly after the specification of the animator had been completed and the team was working on the high level Figure 1 design: the output of this phase was to be a design document for consideration by the wider consortium. Work during the weeks 6
of the study comprised periods of individual work interspersed with team meetings. The team's desks were located close together, in part of a large open-plan office, allowing frequent informal communication between team members, particularly Matt and Peter. However, prolonged wholeteam discussions tended to disturb other people, so these were held in a meeting room and provided the data for this study. This small meeting room was equipped with a large table and chairs and two unfixed whiteboards. Figure 1 is an illustration of the layout of the meeting room. Five Mallard design meetings were video-recorded and form the basis of the ensuing analysis and discussion. The team was debriefed after each meeting to ascertain the general purpose of the session, its conclusions and the consequential actions. A transcript of each tape was made. In addition to the intensive analysis of the meeting record, each team member was briefly interviewed at the end of each working day to establish what tasks he had been engaged upon, and what communication he had had with other team members. The camera was positioned to take in the whiteboards and the team members as they sat around one end of the meeting room table. Apart from an early meeting (not recorded), which the second author attended for familiarisation, no extra persons were present during meetings except when changing tapes. 3.2 The use of artefacts and media in the design meetings During the meetings a number of different artefacts and media were in use. A pair of whiteboards were used as a display of design elements which had already been developed, either in an earlier meeting or by an individual in preparation for a meeting - individuals could add material to the whiteboard before the others arrived, or transfer it from notes during the meeting, explaining as they drew. The whiteboards were also used as an exploratory tool for sketching out new ideas and modifying old ones; and as a bridging mechanism between two temporally separate parts of a design meeting. Printed documents, both formal project documents and individual work-in-progress, sometimes with annotations made in document reviews outside meetings, were used for reference to specifications and to earlier versions of the design. Of the five meetings analysed, three included the current design document itself as a main topic of discussion. However, as well as providing a basis for discussion, documents were also an important workspace for the individual’s own active design work. Documents were also annotated with modifications and suggestions during the meeting. Individual notebooks/pads were also used for comments and notes. These were often very detailed and sometimes provided a shared resource for activities after the meeting. Analysis of the data shows a rapid sequence of switches in focus between media. While individual copies of documents, and of course notebooks, were by default private, occasionally their owners would move them to the public arena to amplify an idea. Moreover, every so often, a team member looked over his neighbour’s shoulder to point out a feature in the other’s notes for comment. At the end of the meeting the resources reverted to their status of records, either to serve as a basis for the next stage of individual or team work, or to be transformed into more permanent design documentation as working papers, or eventually, official project documents.
4
The activity analysis
The background material just presented provides a rich description of the TTM team at work. However, its very richness presents a problem: how to tease apart the different strands of the work being undertaken so that we can describe each one systematically in relation to the other elements. Here we have used activity theory as an orienting-schema. The core activity of software designing may both be seen in its wider setting (the description of which, for reasons of space, we have kept to a minimum) and as a means of providing the context for its constituent actions by which it is realised.
7
4.1 A note on methodology Vygotsky’s work in the 1920’s, was in part, the result of a reaction against the then dominant behaviourist paradigm formulated by psychologists such as J. B. Watson. While behaviourism stressed that human behaviour is a result of conditioning, Vygotsky placed an emphasis on the higher psychological process which behaviourism denied. This raises methodological problems for this kind of analysis. Indeed we are very conscious that have had to rely on the observation of the overt behaviour of the participants in the design meetings, in common with other activity theory researchers in this field. Of course, these observations have been supported by debriefing episodes wherein those being observed were asked as to their intentions and motivation. However this remains a largely unresolved methodological issue for activity theory. 4.2 The wider context of the design activity In practice activities are manifest as relatively stable institutional patterns of behaviour, and unlike Athena do not spring, fully-formed ab initio: instead, they have a cultural context and a history (of their development). Thus any activity necessarily embodies the constraints and benefits of its development and the design activity here is no exception. Interestingly this activity has two distinct lineages: the first is inherited from TTM’s history, praxis and policy in participating in collaborative European projects; the second, lies with the fact that Mallard is a follow-on project from an earlier successful (sic) Esprit project with the same project consortium. While we have abundant data on both these lineages, not all of it is immediately germane to this discussion. However it is worth noting that most of the project’s praxis arose from a mixture of these histories, from the regularity and nature of the wider project’s meetings; project document format standards and naming conventions to the use of TTM’s quality system and management reporting protocols. This is also true of the choice of artefacts used across the Mallard project (Sparc architecture, Solaris, GNU C++ compiler, choice of email platform and so forth). Thus the design activity is a product of its own historical development. 4.3 The structure of the activity What follows is the instantiation of the design work in terms of the canonical structure of an activity. This is intended to provide the wider context for the finer grained analysis at the action level. Subjects The project team comprised Roger, Peter and Matt, who had worked together on the project for some months. Peter and Matt were both experienced designers with effectively identical roles. Roger also worked as a software designer but had the additional responsibility of being the project manager. Artefacts The team’s use of design documents, whiteboards and notepads and so forth have already been described in section 3.2 Object In a narrow, local sense, the object of the activity was to design and ultimately implement a graphical animator. From a broader perspective, TTM were interested in adding the animator to its portfolio of products, as was the project consortium. Praxis Firstly there are the formal rules governing the consortium as a whole, embodied in such devices as the project contract, the collaboration agreement and the European Union’s own abundant rules and regulations. Add to this a variety of local, formal rules and norms at TTM including its ISO 9000 quality system. As already noted, there was a significant shared history of co-working within the project, and many of the rules governing the ways in which consortium meetings were held and informal contact between partner companies had arisen from the ways in which the earlier project had been run. Locally, the team’s mode of working embodied a lack of formal procedures within the group and little overt structuring of the work undertaken within meetings. 8
Community
Division of labour
The broader (local) community of this activity included the software engineers and designers working at TTM who would be expected to contribute to the project if called upon to do so. Other TTM staff, ranging from finance, the quality team and management also had a stake in the project. Other members of the community were the Mallard project partners. The division of labour reflected the makeup of the community and is again usefully divided into local and consortium. At the local level, the most significant feature is Roger’s role as project manager with all the rights and responsibilities that entails. At the consortium level, the division of responsibility was embodied in the project contract and the collaboration agreement.
While the above does describe who is doing what to whom and why, it remains too high level in itself to understand the dynamics of the design activity This requires an action level analysis. 4.4 An action level analysis As noted earlier, an activity is realised by a set of actions. Unlike the activity itself, these are directed at achieving a particular goal as distinct from the overall object (actions have the classic Vygotskian triadic structure of subject – mediating artefact – goal). We have identified the following actions (of course, there may be others which were not captured on video): updating fellow team members, discussing design options, seeking clarification or questioning, recording the design, and managing the design meeting. We now describe two exemplars of these in more detail, updating fellow team members and discussing design options and the switch between them. Note: Verbal protocols Throughout this section protocol fragments from the transcripts of a design meeting have been used to illustrate the activities under analysis. These fragments are drawn from a single design meeting which occurred in the middle of the sequence under study. This was at a time when the designers were developing and refining ideas and beginning to think about how they were to be presented to the outside world. The original protocols are annotated to show the concurrent use of gesture and design media. For reasons of simplicity and space, these annotations have largely been omitted here, except where the use of design media is the main topic of discussion. (Occasionally in the protocol fragments the word indistinct will appear in italics. This is a reference to a break in the transcript where what a speaker has said cannot be determined.) Action I: Updating fellow team members Between meetings, the Mallard team at TTM worked independently to develop parts of the design. Sometimes this resulted in the production and distribution of short working papers; otherwise ideas were captured in notebooks or drawn up on the whiteboard just before the next group session. Even if a paper had been circulated it was not always read in advance of the meeting. Such independent work therefore resulted in the meeting action of ‘updating fellow team members’. Updating occupied a substantial proportion of meeting time. Typically, each designer in turn would stand up and verbally brief the others, using any or all of the range of artefacts identified, indicating the current focus of the argument with explicit verbal reference or gesture. While updating was in progress, the speaker would not generally be interrupted. Comments and questions usually waited to the end of a segment of exposition. While (apparently) listening, other team members might make notes or consult other materials. Protocol fragment 1 has Roger pointing at the whiteboard to begin updating his co-designers with his latest ideas for his part of the design.
9
Speaker
Roger
Action gestures at whiteboard
gestures at whiteboard
Roger
Looks through document Points at document
Verbal protocol OK. I thought I'd pretty-print that [laughter]. Probably haven't lost too much that was interesting. Construction on the left hand side but I've left a little bit of space around construction because we may well end up breaking it down [Peter: mm hm] into models of state, configuration type things, [Peter: yeah] [Matt: yeah]. Control, well yes, control, scheduling is permeating all layers as a result of some of the things we've said [Peter: mm hm] user model, value, evaluation of value, only goes up as far as the abstract model, I haven't taken that through [Peter: yeah] and then there's execution events whose flavour changes as we go up [Peter: yeah] and everything else whose flavour changes, we really ought to indistinct. So what I propose we do is partly on the basis of this original breakdown in structuring that we should be able to find a home or an equivalance for functional things in the mapping layers [Peter: mm hm] like parser, concrete to abstract controller – we had one of those but we haven’t got a name for it now [Peter: mm hm] and we ought to be able to put in some things …. Protocol fragment 1
This action has ultimately the goal of creating a common understanding, evidence of which can be found in the murmurs of agreement or acknowledgement (depending on the tone of the mm hm) from those being updated. Finally, the action of updating is an example of coordination as it is part of the scripted or normal execution of work. As the overall design process is distributed across individual work and team design meetings, the resulting design is also necessarily distributed. A consequence of this is that these multiple representations must be transformed into a centralised and unified form to form the basis of an agreed understanding. Subsequently, of course, this unified, centralised representation does again become re-distributed. Protocol fragment 2, included with its original annotations, illustrates this process and shows a design representation distributed between project documents and the whiteboard. Speaker Peter
Action gestures at document and then looks at the whiteboard gestures at document
Verbal Protocol Right. All I'm thinking about here is in some way you've got to define a given set of events, and when you've got those events, what the animation is. I mean, OK you might define it in terms of these frames things – whatever, who cares. [Roger: laugh] Somehow you've got to handle that and you've got a set of rules to work out what you're going to do. [Roger: mm hm] The more composite the events get the more complex that set of rules is and the more you're going to have. [Roger: mm hm] So there's a balanced trade-off between the granularity of the event structure if you like and the complexity of defining animation rules. Protocol fragment 2
Switching between actions This action level analysis can also illustrate the dynamics of a design meeting. Actions in the shape of challenges or questions are used to switch from updating to co-designing. Sometimes a challenge or question is ignored or answered without a subsequent change in a activity, but in many instances they do serve as a nexus. Protocol fragment 3 sees the murmurs of agreement while Roger is talking (updating the team) slowly decline until a pivotal question is asked by Matt, “So part of the information in the filter will be the thing that is using that filter?”. Roger tries to continue his updating by clarifying his previous statement, but the tone of the meeting now switches from updating to discussing design options. This continues until Roger ends the process with “Pass. It might be.”. (Note – actions have been removed from the next two protocol fragments for the sake of clarity.) 10
Roger updating others: (an instance of coordination)
Speaker Verbal protocol Roger
Peter Roger Matt Roger Matt Roger Peter Roger
You have a number of mappers [P: mm hm], say from the kernel model you have a number of interested parties [M: mm hm] and the kernel doesn't care what those interested parties do with the information, [M: mm hm] but one of them could be mapping up to the user model, to do some filtering; one could be writing a Petri net level execution block for debugging, turning it into ASCII as it went and filtering some things out so it could be a debug type thing. [P: mm hm] Another filter or another thing could be … probably not the animation because there's a different level of communication with that but it could do any number of things and the kernel [M: mm hm] doesn't care what they're doing [M: mm hm]. and the kernel spits out all its events to a set of thing Switch to co-designing: (an instance of cooperation) I thought it would be indistinct interest and doesn't care what they are So part of the information in the filter will be the thing that is using that filter? Well … the filter's actually in the mapper, you know. It's the mapper and the mapper uses the filter but you have more than one mapper … more than one thinger listening to the kernel. So is the filter a separate thing to the mapper? Pass. It might be Protocol fragment 3
In the next fragment of protocol (protocol fragment 4), Roger can be observed interrupting himself clarifying a point, “Already we’ve missed an opportunity …” which again switches him from updating to discussing designing options. Speaker Roger Peter Roger Peter Matt
Verbal protocol Already we’ve missed an opportunity for reducing a lot of messages by not having it actually in the kernel. So maybe when its actually in the kernel the kernel can have a set of listeners and a filter which doesn’t applies to those listeners. It seems to make more sense for whoever's listening to filter out what they don't want to listen to. Roger initiates another co-designing But it increases the traffic episode: Yeah but (an instance of reflective communication.) Also if you looked at that idea then the person who's interested in information, say he's a user at the concrete model layer, [P: mm hm] then if it's his responsibility to filter out then what you'll get, you'll get everything right up to that top level. Protocol fragment 4
Action II: Discussing design options As an example of continuing design work, we next have Roger, Matt and Peter discussing the element mapper for the design animator. The same range of media is involved in this action as in the last but here the emphasis is on building on and extending existing knowledge. Ideas (design options or alternatives) are suggested, debated and decisions made or implicitly deferred. The whole tenor of the group work is one of informality with little or no apparent structure. Protocol fragment 5 shows that the argument draws upon a shared understanding of past states of the design, as well as developing ideas further.
11
Speaker
Action
Matt
Peter
Roger
Gets up
Verbal Protocol Wouldn't this idea you were talking about beforehand ... indistinct … Couldn’t you have a monitor within the animator which says, I want monitor operations on this element map, particular kinds of operations on this element map which might be connected with events. I think that sounds a bit confusing and complicated actually. I thought what we were talking about before was we had an event coming in here, the abstract user model would take it, it would do things to its own state and then pass it on. So the sort of thing you've got going up is an indistinct value object in its own right, or possibly.... I mean that if we looked at the example we worked through yesterday or perhaps the day before that, about two different approaches on the conversion of animation events, execution events and animation events … Protocol fragment 5
Having established the common ground from which to proceed the team works together to agree a decision (protocol fragment 6). Speaker Peter Roger Peter Roger Peter
Action
Drawing
Verbal Protocol That's true actually. And we ought to be also actually mapping values into a stable form. So we could possibly have it on the element map in that form, convert element or something like that? Yeah, OK. Ok, I'm convinced. So the animator talks to that. Protocol fragment 6
Again, Roger is observed pointing at parts of the design on the whiteboard and gesturing towards the working papers to suggest an alternative. As already noted, the praxis in this action is much closer to ‘brain-storming’ with the designers frequently interrupting each other. At this early point in the meeting, Roger is outlining a design proposal which is quite deliberately rough. The provisional nature of his proposal can be seen in those parts of protocol fragment which we have underlined. Speaker
Roger Matt Roger
Action
Verbal Protocol Yeah structure… but it's using the same information whereas … we'll look at the words later. I think this may be more precise or a subset of it. Points at Are these two circles on the way up? Or have you deliberately done them on whiteboard the way up. You indistinct the values, but you need to know what elements those values map to. Drawing on Well, that's that. Sorry, that was meant to be up and that was meant to be whiteboard down. [Matt: right] So what I'm about to do, perhaps I'll do this in a different colour. I mean this is not decent syntax there's something like that giving us that [Peter: yeah] [M: mm] and on the way back down I assert we, basically we can say, you know, its a Chinese Restaurant, 3 number 7s [P: yeah] but I guess we can say in principle that it's an element idea [Peter: yeah] that gets mapped, we're not interested in values. Protocol fragment 7
As the meeting progresses, this part of the design crystallises and is recorded.
12
Speaker
Action
Roger
Verbal Protocol So this means - I'd like to record somewhere that this functionality belongs in one particular place or another, or perhaps it should be in the text associated with the abstract user model. Could we agree that's where it should be? Does that seem sensible? Protocol fragment 8
This process is typical of the design transformations in all the meetings observed. A further transformation of the design is from something local to the design team at TTM to an output for external consumption by the wider project. Such transformations result in the creation of boundary objects. A boundary object (Star, 1989) is a shared object of working which crosses the boundaries of different workgroups and as such acts as a vehicle for communication about work. Individuals in different workgroups necessarily interpret these objects in different ways according to their individual and group needs (Mark, 1997). Protocol fragment 9 shows the team presenting their design-as-boundary-object. Speaker Roger Peter Roger Peter Roger
Action
Verbal Protocol Well, we can just say at the moment that we’ve made an assumption, but we are aware it is an assumption and requires validation or whatever. [Peter: yeah] They won’t even have thought about monitors in that sense. Indistinct languages But it’s breakpoints generalised, among other things. Lots of other things. But we won’t tell them about all those. [laughter] Protocol fragment 9
This concludes the action level analysis.
5
Discussion and conclusions
While we are reluctant to extrapolate too far from just one case study, the material above suggests that Engeström’s formulation of activity theory offers a means of elucidation of the design meetings; allows consideration of the dynamics of the meetings as the design develops and the designers switch between different modes of collaborative workings and helps to explicate the role of the various tools and media used by the designers2. This action level analysis is, of course, closely related to Wertsch’s (1995) observation that the proper unit of analysis is mediated action. While Wertsch has stressed the importance of individuals performing actions in context, he has distanced himself from the notions of the collective nature of human activity, emphasising instead sign-mediation and the interactional aspects instead (ibid). Adopting such an approach to understanding cooperative work has, consequently, not been adopted here. Instead we have again adhered to Engeström’s approach which we believe is closer to Vygostsky’s original formulation. However, it is also apparent that the inclusion of insights from other perspectives may be necessary for a complete account. For example, we are conscious that an activity description does not deal elegantly with mutual awareness, a central issue for cooperative working and one which is well-supported by the conceptual framework of distributed cognition. Changes to the activity theory corpus may be needed to deal this and present an interesting avenue for further conceptual research. A more structural issue concerns the granularity and composition of an activity: for example, it is not entirely clear how the local activity at TTM can be related Since this particular case study does not contain any overt instances of individual or group learning (save in the most general sense of the designers learning about the design as they develop it) we cannot reach any conclusions as to the usefulness of this aspect of activity theory.
2
13
meaningfully to the larger project using the current syntax of activity theory. A further point is that it is far from clear how activity theory could deal with Schmidt and Simone’s work on coordination and the articulation of work. While recognising some of the shortcomings of activity theory in this context we are also acutely aware that we have been selective in our application of it to the analysis of the design meetings. Key concepts such as learning, internalisation – externalisation among others have not been addressed for the want of space. 5.1 Activity theory compared with other accounts of (cooperative) work We now return to those other accounts of work reviewed at the beginning of this paper. To start with the most developed and comprehensive design-oriented approaches, a readily apparent conclusion is that activity theory does not provide comparable ready made descriptive techniques. That being said, the potential descriptive scope is broader than CWA's focus on event-driven domains, and affords consideration of aspects of learning and development over time which are less explicit in both CWA and Contextual Design. Further, in contrast to both these other approaches, an activity model of work is fundamentally cooperative, thus lending strong support for the description of such work. Finally, unlike Contextual Design (but not Cognitive Work Analysis) an activity is strongly driven by a single theory of human behaviour, rather than pragmatically derived from insights from a number of sources. How far this is significant in practice remains a matter for speculation: conceivably, a set of activity descriptions may be more cohesive. Turning to other strongly theoretically motivated approaches, ethnography's rejection of predetermined, generic models necessarily means that no meaningful direct comparison may be made. However, it is clear that an ethnographic approach does provide a rich source of data for the construction of a structured activity description. As for situated action and distributed cognition, we concur with Nardi's conclusions (Nardi, 1996) that while both provide valuable insights into behaviour in the real world of work, situated action's reluctance to engage with the subjective and intentional nature of human action, and distributed cognition's treatment of people and artefacts as functionally equivalent limit these approaches as generators of useful descriptions . However, both provide valuable insights which could extend an activity description, particularly at the action and operational levels. 5.2 From accounts of team working to CSCW At the outset of this paper, we stated that our intention was to demonstrate the usefulness of activity theory in understanding purposive team work. This we have done. To make the link to CSCW we must indicate how a link between activity theory and systems design can be made. Here we introduce one such route, namely, a contradictions-driven approach to reasoning about requirements systematically. Contradictions manifest themselves as disturbances or non sequitors in the free running of an activity. To illustrate the possible use of contradictions we begin with the following observation: it is apparent from the video record that Matt is largely excluded from the exchanges between the other two team members. Other evidence suggests that this was largely because he acted as the group’s primary note taker. This may be indicative of a secondary contradiction between the object (producing a software design) and the way in which work has been divided among the team, and it is reasonable to conclude that this issue might hinder the efficient production of the software design. Thus examining the contradiction has surfaced an issue worthy of further exploration (reminiscent of the first stages of Soft Systems Methodology). A practical way of investigating this further is by means of a scenario. The following scenario is intended to reflect and explore a resolution of this contradiction: let us suppose… … we were to introduce meeting capture technology to replace the need for one member of the design team to act as a recorder. The meeting capture technology would automatically record all design options explored, decisions made and so forth. The technology would also act as central repository for the storage for the design itself, together with personal notes, and records. The entire team could then concentrate on the design process 14
without the distraction of having to keep notes. If the meeting capture system could be shown to be successful it could then be shared with the rest of the project consortium or used as a means of inducting a new team member. However, as activity theory is essentially an holistic approach to the study of work with its assertion that an activity is the smallest meaningful unit of analysis, we should in turn construct a contradictions model of the new activity embodied in the scenario. Some of the emergent contradictions arising from the above scenario are as follows: 1. At the subject node there is the potential for improved team cohesion, with all three designers concentrating on the design process, rather than Matt being distracted note taking. 2. The introduction of meeting capture software offers the opportunity to help simplify and rationalise of the range of artefacts used in the meetings. 3. The object of the activity is to create a software design for the graphical animator. The resolution of the secondary contradictions should result in a more effective design effort. 4. In terms of the rules and norms governing the conduct of the design meetings, fewer interruptions to the flow of the meeting to make an explicit record of the design, should again contribute the design process. 5. In a less immediate sense, the use of the meeting capture software may act as a validation site for its use for the broader communities at TTM and the project consortium, visibility being a contributor to technology acceptance (e.g. Bikson and Eveland, 1996). 6. Better use of Matt’s time as a designer during the meetings, however: 7. The operation of the software itself, unless completely automatic and unobtrusive, may interfere with the object of producing the software design. The above list shows the utility of contradictions modelling as an adjunct to scenario-based requirements elicitation (e.g. Carroll, 1995). This simple analysis of the possible contradictions is intended to be illustrative only, as the consequences for concurrent activities have not been explored. Indeed, the above set of contradictions are just one of many possible worlds. It would be expected that after the possible sets of contradictions had been identified, such devices as futures workshops would be used to agree which contradictions would be resolved and which suppressed (Engeström 1999). 5.3 In conclusion We observe that with the extensions discussed above, activity theory affords a powerful descriptive account of cooperative work, but requires further development and operationalisation to be usable as a method by non-activity theorists. A final promising possibility is the potential, via structured scenarios and contradictions, for a link to made to requirements and therefore to design.
6
References
Bardram, J. E. (1997) Plans as situated action: an activity theory approach to workflow systems. In J. A. Hughes, W. Prinz, T. Rodden and K.Schmidt (Eds.) The Proceedings of the Fifth European Conference on Computer Supported Cooperative Work, Kluwer Academic Publishers, 17-32. Bardram, J. E. (1998a) Scenario-based Design of Cooperative Systems. In F. Darses and P. Zaraté (Eds.) The Proceedings of COOP ‘98, INRIA. Bardram, J. E. (1998b) Designing for the dynamics of cooperative activities. In The Proceedings of the CSCW’98, ACM, 89-98. Beyer, H. and Holtzblatt, K (1998), Contextual Design: Defining Customer-Centered Systems, Morgan Kaufmann Publishers Inc., San Francisco. Bikson, T. K. and Eveland, J. D. (1996) Groupware Implementation: Reinvention in the Sociotechnical Frame. In M. S. Ackerman, Ed., Proceedings of CSCW '96, New York, ACM Press, 428-437. Blackler, F. (1993) Knowledge and the theory of Organizations: Organizations as Activity Systems and the reframing of management, Journal of Management Studies, 30(6), 863-884.
15
Blackler, F. (1994) Post(-)modern organizations: understanding how CSCW affects organizations, Journal of Information Technology, 9, 129-136 Blackler, F. (1995) Activity Theory, CSCW and Organizations, In A. F. Monk and N. Gilbert (Eds.) Perspectives on HCI – Diverse Approaches, Academic Press, London. Bødker, S. (1991) Activity theory as a challenge to systems design. In H.-E. Nissen, H. K. Klein and R. Hirschheim (Eds.) Information Systems Research: Contemporary Approaches and Emergent Traditions, Elsevier Science Publishers B. V. (North-Holland), 551-564. Bødker, S. and Christiansen (1997) Scenarios as springboards in CSCW design. In G.C. Bowker et al. (eds) Social Science, Technical Systems and Cooperative Work, Laurence Erlbaum Associates, Mahwah, NJ. Bødker, S. and Grønbæk, K (1998) Users and designers in mutual activity, In Y. Engeström and D. Middleton (Eds.) Cognition and Communication at work, Cambridge University Press. Card, S. K., Moran, T.P. and Newell, A. (1983) The psychology of human computer interaction, Lawrence Erlbaum Associates, Hillsdate, NJ. Carroll, J. M. (1995) The scenario perspective on system development. In J. M. Carroll (Ed.) Scenario-Based Design: Envisioning Work and Technology in Systems Development, John Wiley, New York. Christiansen, E. (1996) Tamed by a Rose. In B. A. Nardi (Ed.), Context and consciousness; activity theory and human-computer interaction, MIT Press, Cambridge, Mass, 175-198. Cole, M. and Engeström, Y. (1993) A cultural-historical approach to distributed cognition. In G. Salomon (ed.) Distributed Cognitions – Psychological and educational considerations, Cambridge University Press. Dourish, P. and Button, G. (1998) On “Technomethodology”: Foundational relationships between ethnomethodology and systems design, Human Computer Interaction, 13, 395-432. Engeström, Y. (1987) Learning by expanding: An activity-theoretical approach to developmental research, Helsinki: Orienta-Konsultit. Engeström, Y. (1995) Objects, contradictions and collaboration in medical cognition: an activitytheoretical perspective, Artificial Intelligence in Medicine, 7, 395-412. Engeström, Y. (1998) The tensions of judging: Handling cases of driving under the influence of alcohol in Finland and California, In Y. Engeström and D. Middleton (Eds.) Cognition and Communication at work, Cambridge University Press. Engeström, Y. (1999a) Expansive visibilization of work: an activity theoretic perspective, CSCW, 8(1-2). Engeström, Y. (1999b) Innovation learning in work teams: Analysing cycles if knowledge creation in practice. In Y. Engeström, R. Mittinen, and R-L Punamäki (Eds) Perspective on activity theory, Cambridge University Press. Engeström, Y. and Miettinen, R. (1999) Introduction. In Y. Engeström, R. Mittinen, and R-L Punamäki (Eds) Perspective on activity theory, Cambridge University Press. Engeström, Y., Brown, K., Christopher, L. C. and Gregory, J. (1997) Coordination, cooperation, and communication in the courts: expansive transitions in legal work. In M. Cole, Y. Engeström and O. Vasquez (Eds.) Mind, Culture and Activity, Cambridge University Press. Greenbaum, J and Kyng, M. (eds.) (1991). Design at Work: Cooperative Design of Computer Systems. Lawrence Erlbaum, Hillsdale, NJ. Hasan, H. (1998) Activity theory: a Basis for the Contextual Study of Information Systems on Organisations. In H. Hasan, E. Gould and P. Hyland (Eds) Information Systems and Activity Theory: Tools in Context, University of Wollongong Press. Heeren, E. and Lewis, R. (1997) Selecting communication media for distributed communities, Journal of Computer Assisted Learning, 13, 85-98. Holt, G. R. and Morris, A. W. (1993) Activity Theory and the Analysis of Organizations, Human Organization, 52(1), 97-109. Holtzblatt, K. and Beyer, H. (1996) Contextual Design: Principles and Practice, in D. Wixon and J. Ramey (Eds.) Field Methods Casebook for Software Design, John Wiley & Sons, New York. Hutchins, E. (1995a) Cognition in the Wild, MIT Press. Hutchins, E. (1995b) How a cockpit remembers its speed, Cognitive Science, 19, 265-288. Hutchins, E. and Klausen, T. (1996) Distributed Cognition in an airline cockpit. In Y. Engeström and D. Middleton (Eds.) Cognition and communication at work. Cambridge University Press, 15-34 Kuutti K. and Bannon, L. (1993) Searching for unity among diversity: exploring the ‘interface’ concept, in Proc. Of the Human Factors Conference in Computing Systems, INTERCHI’93, ACM, New York. Kuutti, K. (1995) Work Processes: Scenarios as Preliminary Vocabulary. In J. M. Carroll (Ed.) ScenarioBased Design: Envisioning Work and Technology in Systems Development, John Wiley, New York, 19-36. 16
Kuutti, K. (1996) A Framework for HCI Research. In B. Nardi (Ed.) Context and Consciousness, MIT Press, Cambridge, MA. Kuutti, K. and Arvonen, T. (1992) Identifying Potential CSCW Applications by Means of Activity Theory Concepts: A Case Example. In J. Turner and R. Krauts (Eds.) The Proceedings of CSCW ‘92, ACM Press, 233-240. Lui, Y., Maxfield, J. and Dew, P.M. (2000) Investigation of task performance for interaction with a virtual environment for assembly, in R. Holland (Ed.) Proceedings of the 7th UK VR-SIG Conference, Strathclyde University, September 19-20th, pp. 53-64. Mark, G. (1997) Merging multiple perspectives in groupware use: intra- and inter-group conventions. In the Proceedings of Group ‘97, ACM Press, New York. Myers B., Hollan J., Cruz I., Bryson S., Bultermann D., Catarci, T., Citrin, W., Glinert E, Grudin J, Ionnadis Y, Jacob R, John B, Kurlander D, Olsen D, Pausch R, Shieber S, Shneiderman, Stasko J, Strong G, Wittenburg K. (1996) Strategic directions in human-computer interaction. ACM Computing Surveys, 28(4), Dec. 1996, ACM, USA. 794-809 Nardi, B. (1996) Some reflections on the application of activity theory. In B. Nardi (Ed.) Context and Consciousness, MIT Press, Cambridge, MA. Pycock, J. and Bowers, J. (1996) Getting Others To Get It Right: An Ethnography of Design Work in the Fashion Industry. In M. S. Ackerman, Ed., Proceedings of CSCW '96, New York, ACM Press, 219-228 Rogers, Y. and Ellis, J. (1994). Distributed Cognition: an alternative framework for analysing and explaining collaborative working. Journal of Information Technology, 9(2), 119-128. Schmidt, K. and Simone, C. (1996) Coordination Mechanisms: Towards a Conceptual Foundation of CSCW Systems Design, Computer Supported Cooperative Work: the Journal of Collaborative Computing, 5, 155-200. Star, S. L. (1989) The Cultures of Computing, Blackwell, Oxford. Turner, P. and Turner, S. (1998) Structuring scenarios using activity theory, In P. Wright and R. Fields (Eds.) The Proceeding of the First International Workshop on Understanding Work and Designing Artefacts, York, September 1998. Turner, P., Turner, S. and Horton, J. (1999) From description to requirements: an activity theoretic perspective, Proceedings of Group’99, ACM Press. Twidale, M., Randall, D. and Bentley, R. Situated evaluation for cooperative systems. In The Proceedings of CSCW ‘94, ACM Press, 441-452. Verenikina, I. and Gould E. (1998) Cultural-Historical Psychology and Activity Theory. In H. Hasan, E. Gould and P. Hyland (Eds) Information Systems and Activity Theory: Tools in Context, University of Wollongong Press. Vicente, K.J. (1999) Cognitive Work Analysis, Laurence Erlbaum Associates, Mahwah, NJ. Viller, S. and Sommerville, I. (1999) Coherence: An approach to representing ethnographic analyses in systems design, Human-Computer Interaction, 14(1-2), 9-41 Vygotsky, L. S. (1978) Mind in society: the development of higher psychological processes (English Translation Ed. M. Cole): Cambridge MA, Harvard University Press. Wartofsky, M. (1979) Models: representation and scientific understanding, Reidel Publishing Company, Dordrecht, Holland. Wertsch, J. V. (1995) The need for action in socio-cultural research. In J. V. Wertsch, P. del Rio, A. Alvarez (Eds.) Sociocultural studies of the Mind. Cambridge: CUP.
17
