This page offers but a brief sketch of some issues pertaining to the delineation of systems as wholes versus their delineation as networks of interacting consituents. The issues may seem a bit abstract, but I have found them to be obstacles in problem solving / decision making processes on a regular basis. The key distinction I've drawn between syn-referentiality and sys-referentiality has proven very useful in sorting out problem space ambiguities and providing contextual tracking mechanisms for facilitating such processes.
It has been common in scientific pursuits to categorize phenomena in terms of their levels of composition and/or participation within a hierarchical schema. Such schemata typically range from the 'micro' toward the 'macro' in scope. For example, in biology the canonical such hierarchy ranges through (e.g.) molecule, macromolecule, organelle, cell, organ, somatic system, organism, population, and species. Traditionally, a qualified uniformitarianism across the boundaries of such hierarchical levels has been tolerated (if not assumed). This attitude is the basis for the sort of reductionism I was seeking to overcome.
Maturana and Varela make it clear in a number of their writings that an observer's circumscription of a phenomenon is contingent on the context of her interaction with the phenomenon. More specifically, the distinctions employed by the observer in delineating the phenomenon circumscribe both what is observable and how observation can proceed. One dimension of variation in making distinctions is scale (in the sense of micro vs. macro level). As such, an autopoietic account of multiple levels in hierarchies must consider them as distinct domains of observational engagement. This is not to say that they are to be treated as totally isolated and mutually irrelevant -- only that the observer's descriptions of them are based upon (and qualified by) non-isomorphic distinctions which in turn serve to limit any isomorphism among the explanations grounded therein.
The issue of hierarchies and their validity has been largely neglected or blurred in systems theory. In my opinion, this is a result of the fact that the essential construct 'system' has never been adequately specified. We use the term 'system' (with wildly variable exactitude) to circumscribe objects of concern and prospective solutions in everyday problem solving exercises. When a factor of interest lies outside the scope of the circumscribed unit of reference, the conventional approach has been to slide into descriptions of 'information flow' and attribute performance variations to 'feedback loops' acting as threads mystically knitting together the 'system' to its 'environment'. The wholesale invocation of this strategy spelled the demise of classical systems theory and its devolution through cybernetics to systems engineering.
A close inspection of the literature will underscore the absence of a unified delineation of the core concept 'system'. Check any text on systems theory or systems analysis, and the most you will find is a 'laundry list' of attributes intended to summarize the characteristics of a 'system'. More commonly, the entire issue of definition is left to the reader's imagination. Von Bertalanffy (1968, pp. 9, 38, 55 ff.) takes this latter approach, specifying a system as nothing more detailed than '...a set of elements standing in interrelations.' This is precisely the same ambiguity and laundry list approach to be found in defining 'life' in biology texts.
The last time the notion of 'system' was an issue of deep concern was during the 1940's. Most detailed analyses of the basic concepts assumed that whole / part relationships were obvious and therefore non-problematical (cf. Feibleman and Friend, 1945). Andras Angyal's (1941) analysis of systems logic is perhaps the most perceptive of the period. He focused on the organization of a system rather than its composition, concluding that systems components are not significantly related to each other except to the extent their 'interrelations' are relevant to the whole system in which they participate.
Contemporary decision making is increasingly aimed at 'systems' of varying complexity and interrelatedness. In my observations of actual decision makers (individuals and teams), the framing of the problem has typically been the most critical step toward a solution. In some cases, the solution has been relatively straightforward once this was done. Once a system has been so 'framed' (as a model), the efficiency and effectiveness of subsequent analysis is dependent upon the extent to which the decision maker(s) can adhere to this model for the rest of the process.
The most common adherence problem I've noted in my experience is the need to address systems from two perspectives:
Decision makers who attempt to proceed from one perspective alone run the risk of missing, mistaking, or misjudging phenomena delineated from the other perspective. Working from an exclusively internal perspective can lead to 'not seeing the forest for the trees', and adhering to the external vantage can lead to global solutions with destructive local consequences. Unfortunately, I found no conceptual framework which acknowledged, much less solved, this sort of perspectival confusion. As a result, I had to work by analogy from an example of such dichotomous viewpoints. That example is outlined in the next section.
The issue of how autopoiesis can or should be applied to social systems is an ongoing topic of debate. For the purposes of this discussion, it's enough to note there are two primary approaches in applying autopoiesis to social systems. The first applies the formal aspects of autopoietic theory (e.g., organization; autopoiesis) to the social system itself. The second derives an explanation of the social system from the phenomenological aspects of the theory (e.g., the observer; languaging). These two approaches have demarcated the lines of debate over the years. In the sections below, some of the key issues to date will be reviewed. This overview will necessarily be brief and shallow. A more substantial review of these issues can be found in my 1995 WWW treatise on Self-Organization, Autopoiesis, and Enterprises.
Approach I: Enterprises as Autopoietic Systems
One approach in applying autopoietic theory to enterprises is to treat the enterprise itself as a unit system to which autopoiesis is ascribed. An enterprise in this view provides the static and the dynamic framework within which the presence and behaviors of its participating subsystems (including people) are realized. For this to work, the enterprise must be analyzable as a coherent network of objects and processes. Because such analyses are the hallmark of many social and organizational science studies, this has been the approach by which most writers have attempted to apply Maturana and Varela's ideas to enterprises. The best-known proponent of this approach is the German sociologist Niklas Luhmann, who fit autopoiesis into his ongoing systemic analysis of society (e.g., Luhmann, 1982; 1984; 1986). An overview of the salient points is given in Bednarz (1988). A detailed presentation and analysis of Luhmann's ideas can be found in Mingers (1994), and von Krogh and Roos (1995) discuss him throughout their book.
If human beings are the constituent elements of a social system, how can the social system (itself) be considered self-reproducing in terms of the humans? Luhmann's solution is to search for some constituent element other than humans. The rationale is that '...social processes must correspondingly produce social components if the concept of autopoiesis is to be extended to the social domain with any validity.' (Bednarz, 1988, p. 61) Luhmann redefined social systems as being realized in a domain of 'communications'. In other words, the constituent elements of the social system are communications, and the required conditions for autopoiesis are met in terms of such communications. This has the advantage of describing the system in terms of its operational characteristics, independent from the specific participants in that system at any given time. Luhmann's approach is radical in the sense that it treats social systems solely in terms of 'communications', making the human participants peripheral components at most. His ideas are most persuasive in his specific application of the principles to the field of law (cf. Teubner, 1988), where highly-structured 'communications' are more easily considered as a network unto themselves than is the case for most enterprises.
Approach II: Enterprises as Emergent From Interactivity
The other main approach is to treat social systems as constitutively emergent from interactivity among their participants. This approach takes the individual and collective participants as the fundamental objects of interest. The 'enterprise' itself is addressed as the emergent phenomenon which is conventionally described as an 'organization'. Excluding Maturana and Varela themselves, the most cogent alternative to Luhmann's analysis of autopoiesis and social systems comes from another German sociologist -- Peter Hejl -- who provides a concise, detailed analysis with regard to systems theory in general and autopoietic theory in particular (Hejl: 1980;1981;1984).
Hejl's (1980;1981) starting point is a critique of prior attempts to define social systems as entities in and of themselves (e.g., sociological structuralism and functionalism). He sets out to explore the idea of society as "...the process in which individuals interact with one another and with their natural (real) environment under the primacy of self-preservation." (p. 176). In other words, what had since Durkheim been considered a stable or evolving structural entity (i.e., society as a unit object of which individuals are merely members) was to be analyzed as an emergent effect of individuals' mutual interactivity.
Hejl defines social domains as being generated through "...a process of mutual interactions and hence modulation which results in a partial parallelization of the interacting systems." (1984, p. 68) This is basically a variation on consensual domains invoking 'parallelization' rather than 'mutual orientation'. What others had viewed as a unit social system, Hejl defined as an instantiation of a social domain -- "...a group of living systems which are characterized by a parallelization of one or several of their cognitive states and which interact with respect to these cognitive states." (Op.cit., p. 70)
In Hejl's view, social systems are defined in terms of an intersection between their composite identity and the individual participants. He characterizes such phenomena as syn-referential, i.e.:"...constituted by components, i.e., living systems, that interact with respect to a social domain. Thus the components of a syn-referential system are necessarily individual living systems, but they are components only inasmuch as they modulate one another's parallelized states through their interactions in an operationally closed way." (1984, p. 75)
Syn-referentiality delineates system constitution as an emergent, even a secondary, effect of interactivity among its constituent components. Luhmann's account treats the system as a whole unit whose components are discernible only in terms of a uniform system-wide characteristic or factor (in this case, communication). By labeling Luhmann's approach 'sys-referentiality' for symmetry, we obtain the descriptive complementarity summarized in Table 1.
Luhmann's Model Hejl's Model (Sys-Referential) (Syn-Referential) Focal aspects employed from autopoietic theory: Formal aspects Phenomenological aspects (organization; closure) (coupling; languaging; observer) Basic unit of analysis: System S as a unit System S's components as units Elementary constituent: Communicative act Individual participants Status of social system: Autonomous / autopoietic Emergent / enacted Known by atomic form Known by apparent pattern Referential mode (cf. Robinson's double level language): Formal level language Cultural level language
As modes of observation / analysis, syn- and sys-referentiality have complementary foci. A syn-referential viewpoint (synspect) focuses primarily on the constituent units whose interactions (secondarily) realize a subsuming system S. A sys-referential viewpoint (syspect) focuses primarily on the entire unit system S whose overall character is (secondarily) describable in terms of its constituents' dynamics. The two viewpoints are complementary in the sense that one sets its boundary of observation (its 'horizon', to borrow the hermeneutic term) where the other's ends. A comparative sketch of the two perspectives applied to a simple system is given in Figure 1.
Figure 1: A System S Considered from Syn- and Sys-Referential Perspectives
Figure 1 shows a simple system S constituted by the interaction of two components, x and y. The synspect on S focuses on x, y, and the interactions between them. The subsumption of these elements into a circumscribed entity S is an allusion from this viewpoint. A syspect focuses on the entire circumscribed entity S. From this viewpoint x, y, and their pattern(s) of interaction 'recede' in terms of referential clarity. In other words, the composite S is 'over the horizon' in the synspect, and the componential elements x and y (as well as their interactivity) is similarly beyond grasp in the syspect.
The distinction between syn- and sys-referentiality (with respect to a given system S) is not isomorphic with the common distinction between 'top-down' versus 'bottom-up' approaches to system analysis and intervention. These latter terms pertain to the 'direction' in which analysis proceeds along a scalar dimension (e.g., hierarchical 'altitude'). They are not qualified with respect to intra-/inter-system boundaries along such a dimension. The syn-/sys distinction concentrates on such boundary conditions as limits to (sub)system delineation. As such, syn-/sys- boundaries can be construed as 'punctuating' the path(s) between 'top' and 'bottom' of a constitutional hierarchy.
In practice, the syn-/sys- duality is rarely acknowledged, and decision makers freely jump from sys- to syn-referential viewpoints without realizing that they are thereby addressing their problem via distinct (and therefore at most only partially interrelated) approaches. This is the second (constitutional level) form of Robinson and Bannon's (1991) flipover noted earlier. The relevance to decision making in (e.g.) management and design activities is that such profligate conceptual hopscotch both distorts the problem analysis and obscures (in the name of 'objectivity') any biases or value criteria implicit in the linkages drawn between the two vantages. This last issue is unavoidable, as noted by Varela (1979, p. 85):
'...[T]he establishment of system boundaries is inescapably associated with what I shall call a cognitive point of view, that is, a particular set of presuppositions and attitudes, a perspective, or a frame in the sense of [Gregory] Bateson ... or [Erving] Goffman...; in particular, it is associated with some notion of value, or interest. It is also linked up with the cognitive capacities ... of the distinctor. Conversely, the distinctions made reveal the cognitive capabilities of the distinctor.'
The disjunction between these viewpoints demarcates many dilemmas in enterprise management today. Just as it is impossible to simultaneously address a social domain as a whole and as a network of interactors, it is impossible to simultaneously analyze a task system as a formalized unity and as an interactional milieu. A sys-referential approach to such systems should only be undertaken with the acknowledgement that results may not be informative regarding participants and relations among them. A syn-referential approach should only be undertaken with the acknowledgement that results may not be informative regarding the overall task system and its relation to its environment.
I have participated in business process re-engineering (BPR) exercises in which pursuing an exclusively syn- or sys-referential line has resulted in an initial solution which had to be rejected when viewed from the other perspective. For example, in one case a participatory design (PD) process resulted in a workload allocation procedure made so flexible and desirable from the perspective of the workers (syn-referential with respect to the workplace) that it was tentatively adopted until it became apparent that the procedure detracted from overall organizational performance (sys-referential with respect to the workplace). Any number of cases illustrate the reverse -- where exclusive attention to overall organizational performance led to interventions with negative impacts on constituents and their interactions.
The flipover between synspect and syspect is not something that can (or should) be eliminated. Migration across the syn-/sys- boundary is necessary for fleshing out the decision makers' understanding of a problem and the ramifications of candidate solutions. My studies and facilitation of decision making processes indicate that establishing and maintaining referential / indexical focus is important for processual efficiency and effectiveness. This form of flipover was analyzed as a basis for identifying and tracking its occurrence during decision processes. The syn-/sys- distinction, employed as an organizing criterion, has proven useful in demarcating such focus initially and in tracing its effects over time.
NOTE: The syn-/sys- distinction was formulated in 1989 and heavily elaborated during 1990 / 1991. No documentation from that work is publicly available.