In analyzing the use of information technology (IT) artifacts in support of problem solving / decision making tasks, it became apparent to me that prior research had tended to treat distinct spheres of activity as uniform aspects or extensions of a single medium of interest. Such an orientation treated artificial models on a par with actual situations and / or participants' conceptualizations. To accept this perspective requires a tacit presumption of deterministic inter-referentiality among these disparate domains. To act upon such a presumption requires an additional tacit presumption that this inter-referentiality could be preserved and manipulated through informational exchanges. These presumptions are indicative of the objectivistic / cognitivistic orientations I was seeking to overcome.
Mike Robinson (1989; 1991) had already touched on this problem with his construct of double level language , which distinguishes two interactional vehicles or modes observed in group work. The formal level consists of structured modes of interactivity, usually guided by the model or rules embodied in (e.g.) a software application. This level '...is essential as it provides a common reference point for participants. A sort of 'external world' that can be pointed at, and whose behavior is rule governed and predictable.' (1989, p. 56) The cultural level denotes '...a language that is actually spoken by a community of people'. Conversations at the cultural level may involve '...understanding, interpreting, and changing 'items' at the formal level...', '...procedural and annotative activities...', and '...also any other social or interpersonal aspects (relevant to the 'problem' or not) that the participants wished to introduce.' (Ibid. ).
My work with actual decision makers had frequently revealed disparities between (e.g.) actual situations, the models used to represent them, and actors' conceptualizations of either. In other words, the presumed uniformity noted at the beginning of this document was not apparent in concrete activities. Robinson's 'double level language' was useful for sorting out participants' interactional modalities, but it didn't really provide a tool for linking these distinctions to the task and the information artifacts. As a means for (1) sorting out the disparities observed in actual practice and (2) generalizing upon Robinson's double level language construct, I employed Maturana and Varela's construct of 'domain' -- relabeled 'venue' -- to differentiate among distinguishable contexts for, and sets of, activities. The term venue , chosen in accordance with its legal connotation of 'jurisdiction for action', denoted a setting or circumscriptive medium for action.
The resultant venue framework was originally devised to illustrate and analyze issues concerning the interplay of distinguishable domains of (inter-)action in IT-supported decision making processes. As such, the introduction of the venue framework was framed with respect to decision support systems (DSS) for individual users. The totality of the human user, the DSS, and the problem setting (e.g., a task) is termed a decision environment . Figure 1 below illustrates a single-user decision environment.
The scenario sketched out in Figure 1 is one in which the decision maker has direct engagement with both the task and depictive venues. This need not be the case. For example, if the decision maker's engagement with the task domain was conducted only via an intermediate depictional device (e.g., a control panel), the arrangement would be different. In such a case, there would be a depictive venue intervening between the task venue and the decision maker's cognitive venue. This intermediate depictive venue could well be separate from another depictive venue (as illustrated) which the decision maker engages in the course of decision processing. In any such case, the composite decision environment is differentiable into three venues:
All 'connection' or 'correspondence' between the task and the depictive venues operates through the decision maker's cognitive venue. This is an applied illustration of autopoietic theory's principle that it is only through an observer whose cognitive domain intersects two (or more) distinct domains of interactions that such an observed correspondence or connection is discerned (cf. Maturana and Varela, Autopoiesis and Cognition , 1980). A further implication of Maturana and Varela's theory is intended: that the intersection (cognitive) domain within which the observer engages both the other domains circumscribes a realm of congruence (in terms of interactional potential, hence in terms of referential potential) which in turn circumscribes the character of interactions under this intersection.
In shifting from single-user DSS to collectively-used group decision support systems ( GDSS ) , the triadic venue breakdown of Figure 1 becomes deficient as an analytical tool. Since GDSS are intended to support multiple decision makers, interpersonal interaction becomes an additional area of concern. A new 'dimension' is introduced in shifting focus from individuals to groups -- one which cannot be assigned to one or more of the 3 venues delineated for a single-user DSS in Figure 1. Conversations occur in a 'real world' medium, but this medium is not necessarily isomorphic with the task venue as defined above. Similarly, conversations may be treated in terms of symbolization, but they are not necessarily confined to a symbolic framework isomorphic with the depictive venue as defined above. They reciprocally derive from and affect the cognitive venues of the conversants, but they are not some extended, group cognitive venue.{1} To address GDSS requires adding one more venue:
A four-venue delineation of a multi-participant interaction is illustrated in Figure 2 above. The discursive venue is a common domain of interaction for the participants. Although task-oriented interactions in the discursive venue are referentially related to the task and depictive venues, any interconnection is realized only to the extent that it is realized through one or more of the participants' cognitive venues. Should this seem unduly restrictive, it is justified by the evidence I've accumulated from observational studies of task-directed interactions that the accomplishment of a goal is itself often contingent upon the personal judgement and orientation of individual participants. Where the task goal is a determination (e.g., plan approval) and the task venue is a meeting, this disjunction is (somewhat facetiously) illustrated in Figure 3 below.
In its initial application setting, the venue framework circumscribes the settings or contexts within which the decision environment's ordering relationships are realized. Selecting one of the venues focuses attention on the entities and actions constituting it, thus circumscribing the vantage for observation. This differentiation of a decision environment into venues has the reciprocal effect of making all other entities, events, and phenomena less 'transparent' to observation. Phrased another way, the adoption of a particular venue 'vantage' imposes a relevant 'foreground' from which non-included elements (or roles associated with elements) recede. For example, the depictive venue entails a vantage on media, symbols, transformations on the symbols, etc. For an observer in the 'depictive vantage', the cognitive venues of participants and the issues deriving from the task venue are opaque to the extent (and because) they are specified (delineated) with regard to entities and relationships not subsumed within the observer's given vantage.
The four venues outlined above (cf. Figure 2) do not exhaust the possibilities. Additional venues could be specified as analytical needs warrant. One example would be a 'social venue' -- a setting for interactions not limited to the decision process, intersecting and influencing the decision environment via the task and discursive venues (themselves socially manifested). Such a venue would subsume actors engaging each other via 'roles' -- e.g., conducting their interaction with regard to status, power, expectations derivative from job descriptions, etc. A variation on this social venue would be a composite 'activity venue' in which the participants serve as allopoietic components of a task system. It is my position that current implementations of activity theory (cf. Nardi (ed.), 1991) address themselves to such a composite domain of study. Another example would be the 'physical venue' of the interactional setting itself, intersecting and influencing the decision environment primarily via the depictive and discursive venues. Such influence has been the subject of much research on the design and implementation of group meeting facilities.
The venue framework has been applied to studies of interactivity within work groups employing IT, as well as to functional analyses of groupware deployments. It provides a basis for organizing and clarifying observations and analyses in CSCW and HCI research settings. By differentiating among the domains of interaction underway, the venue framework gives observer / analysts a device for sorting out events with respect to their potentially (more to the point -- typically) distinct significations. For example, distinguishing between the discursive and depictive venues allows an analytical differentiation between task burdens for communication with other decision makers versus task burdens for wrestling with the IT support tools.
Finally, the venue framework has provided a valuable tool for critically analyzing the cognitivist bias inherent in GDSS implementations to date (cf. Whitaker, 1994). Cognitivism's computational paradigm presumes a measure of descriptive equivalence between the depictive and the cognitive venues. A cognitivist perspective circumscribes its subject activity as symbol processing, thus justifying the equivalence it draws between the data on the computer screen and the 'concepts' in the user's head. GDSS users concentrate on directly entering and reviewing symbolic representations, because these are assumed to parallel their own ideation. Second, cognitivism's computational model of cognition presumes a measure of operational equivalence between the depictive and the cognitive venues. Rationality in the decision maker is portrayed as logical manipulations and quantitative evaluations over symbols, explaining the equivalence drawn between group text generation (and manipulation) using GDSS and a 'rational decision process'.
Third, cognitivism's computational paradigm presumes a measure of both descriptive and operational equivalence between the the depictive and the discursive venues. 'Information flows' (data streams) in the depictive venue (e.g., bit streams in a LAN) are presumed analogous to interpersonal communications in the discursive venue -- i.e., that communicability is dependent on symbolizability. This omits consideration of many non-symbolic aspects of communication -- a troubling factor noted by Swedish IT professionals using the University of Arizona's Collaboration Management Room (Docherty, 1992). This presumed correspondence explains why (1) GDSS users are often allowed to interact only through their computer interfaces and (2) why GDSS researchers have seemed willing to treat as seamless the extension of their experience from co-located facilities to distributed scenarios.
NOTES:
'...a new branch of cognitive science devoted to the study of the representation of knowledge both inside the heads of individuals and in the world ...[and] ... the propagation of knowledge between different individuals and artifacts ...By studying cognitive phenomena in this fashion it is hoped that an understanding of how intelligence is manifested at the systems level, as opposed to the individual cognitive level, will be obtained.'(quoting N. Flor and E. Hutchins (1991, page number(s) not cited)
This definition summarizes my objections to 'distributed cognition' as delineated here -- it focuses upon a discernible quantum knowledge commodity which is (1) external to the embodiment of any particular observer, and (2) capable of straightforward exchange among the subjects of interest. These attributions are clear evidence of an (at least tacit) objectivistic and cognitivistic orientation of the sort I reject. To be fair, it must be pointed out that similar (at least tacit) orientations are discernible in the activity theory which Nardi preferentially selects over distributed cognition.
