What Is Complexity Science?

Toward an Ecology of Ignorance

Will Medd
University of Salford, ENG


What contribution do the social sciences make to the question: What is complexity science? There is a temptation, a desire, to argue the case for a particular complexity science; to acknowledge that there are different perspectives of complexity science, and then to clarify what complexity science is really about. In doing so, we might emphasize complexity science in terms of what it offers for how we see the world, how we investigate the world; indeed, how complexity science offers particular ways of being in the world. This has been a particularly prominent desire for popular science writers (see for example Brockman, 1995; Casti, 1994; Glieck, 1987; Hall, 1992; Lewin, 1993; Cohen & Stewart, 1994; Waldrop, 1992) in their narratives about, to use Brockman's (1995) term, a paradigm shift toward an emerging third culture. And, in a similar way, the tendency within the social sciences has been to adopt what are posited as the key insights of complexity science, in particular emphasizing the importance of nonlinearity, self-organization, and emergence (see for example Byrne, 1998; Cilliers, 1998; Eve et al., 1997).

The problem of these approaches is the tendency to posit particular key insights of complexity science that are uncritically adopted. And yet, we know that complexity science is not only seen in different ways, it is performed in different ways in different places (see Thrift, 1999). Indeed, attempting to offer a definitive account of complexity science for the purposes of clarification (see for example Dent, 1999) can only add to the complexity of what complexity science is: it means to offer another account, another performance of complexity science. I want to propose an alternative approach because I think there is an important opportunity provided by the question What is complexity science?. The opportunity is twofold. First, the question prompts us to think about the assumptions we make about complexity science, and to think more critically about what might hitherto have been taken for granted. Second, that we can pose the question tells us something about how we might need to start thinking about complexity science and what work still needs to be done in understanding complex social dynamics. These questions arise explicitly in the encounter between complexity science and the social world.

For the social sciences, research does not only claim to draw on the complexity sciences; rather, it claims to contribute to the complexity sciences. Indeed, the argument is that there is a two-way process, and emphasis is placed on the potential for the complexity sciences to learn from the social sciences once the social sciences get going (Byrne, 1998: 17). Indeed, the social sciences certainly have got going. Authors from a wide range of disciplines are developing arguments about the implications for transforming our understanding of various aspects of the social world by drawing on and developing the complexity sciences (e.g. Byrne, 1998; Eve et al., 1997; Kiel & Elliot, 1996; Urry, 2000; Haynes, 1999). Of course, these arguments about the role of complexity science can be seen as metaphorical. Indeed, as Thrift (1999) argues, it is in part the metaphorical character of complexity sciences that has enabled them to travel to so many sitesfrom different scientific disciplines, to organizational consultants, to the New Age movement. However, there is more to complexity science than mere similes (see Khahil, 1996: 4-7), and yet to explore the different explanatory possibilities of complexity science for social dynamics, a central problem persists: In what sense can complexity science accommodate and explain the characteristics of the social world? The key problem lies in the assumptions we make about the social world if we claim that the models of complexity science can be representative of that world. It is by exploring these assumptions that we can make the most of the opportunity we are faced with to answer the question What is complexity science?.

I want to address the question by offering another performance in the social sciences, a performance that puts the possibilities of complexity science to explain social dynamics to the ultimate test: How can complexity science help us answer that question? As a social phenomenon, what would be involved in establishing whether complexity science is a complex adaptive system? What kind of phenomenon will we be describing and what assumptions will this entail? Emphasizing performance is to suggest that we cannot separate what complexity science is from how it is done. It suggests that we do not assume that complexity science is an entity, a phenomenon that exists in and of itself, which we then use, put into practice. It suggests instead that complexity science becomes, is made, through the relations in and through which it is performed, as a reality that is done and enacted rather than observed (Mol, 1999: 77). The fact that we can ask the question What is complexity science? arises from difference and dispute within complexity science; it implies that our answer needs to deal with differences and disputes. And this is the key test for complexity science: How can complexity science account for the complexities of the performance of complexity science?

There is an immediate problem facing me. I want to address the question What is complexity science? through complexity science, and I want to do so while not reducing complexity science to an abstracted singular conceptual framework. I want to maintain differences. However, I cannot maintain all differences. I cannot be comprehensive: complexity denies this possibility (Luhmann, 1995). I need to simplify. A central issue in examining the complexity sciences through the complexity sciences is to explore the ways in which this allows simplification. If we are to think about whether complexity science is a complex adaptive system, we need to explore the adequacy of the simplifications that this will entail. We need to examine the simplifications that complexity science suggests in order to know the world, and yet, hand in hand with this knowing is ignorance (Luhmann, 1998). We will have to ignore that which we cannot accommodate in our representations, that which we cannot see, and that which we do not know about. And so this is another question for our test of complexity science: How does it handle ignorance?


Our starting point for examining the contribution of the social sciences to the question What is complexity science? might have been to identify the complexity sciences as a fundamental challenge to the traditional linear programme in science as a whole and its ideas of certainty and randomness (Byrne, 1997: 1; see also Eve et al., 1997). We might have located complexity science within debates about the current status of knowledge in general, and knowledge of the social world in particular, arguing, for example, that complexity science is an affirmation of modernism, the basis for a reformulation of modernism, the basis for a foundational postmodernism. For example, the authors in Eve et al. (1997) suggest that complexity science represents a reconstruction of science and social science, while Byrne (1998; see also Reed & Harvey, 1992) argues from a critical realist standpoint that the complexity program contributes to absolute and unregenerate progressive modernism (Byrne, 1998: 158). By contrast, Cilliers argues that the complexity sciences do represent the postmodern condition, characterized by a multiplicity of representation and heterogeneous discourses that represent a recognition of complexity, but a condition that is relational, self-organized, not one of anything goes (Cilliers, 1998).

So, where should we begin our account of complexity science by drawing on the complexity sciences? Byrne (1998: 14-15), for example, argues that the complexity sciences are based on premises of holism rejecting reductionism to the parts, nonlinearity rejecting linear relationships, and historical irreversible evolution rejecting Newtonian conceptions of time. To establish complexity science as a complex system, the implication is that we would need to demonstrate that it is holistic, nonlinear and historically irreversible. Alternatively, Cilliers (1998: 3-5) offers a list of the different characteristics of complex systems. Complex systems have a large number of elements; interact physically and informationally in time; are rich in the sense that several elements affect several others; are nonlinear; have short-range interactions; are characterized by feedback; are open to the environment; operate far from equilibrium; are historical; and, finally, each element operates according to local information only (and is ignorant of the behavior of the whole). To establish complexity science as a complex adaptive system, we would need to examine how the phenomenon complexity science can be understood in these terms, in a similar way as Cilliers does with an example of an economy (Cilliers, 1998: 6-7). We do not have the space here to do this and I want to focus on one aspect of complexity science that is particularly central to the complexity science research program in the social sciences: emergence.

Emergence is central to complexity science (Holland, 1998; Kaufmann, 1993) and is of particular significance since it enables a move away from reductionist and teleological explanations of system dynamics (see Capra, 1996). Authors have recognized that emergence is not new to the social sciences. For example, George Herbert Mead was concerned with emergence in which the combination of elements with one another brings with it something that was not there before (Mihata, 1997: 31). This is consistent with complexity science ideas of emergence, where emergent phenomena are conceptualized as occurring on the macro level, in contrast to the micro-level components and processes out of which they arise (Goldstein, 1999: 49). However, the complexity sciences have focused on how global structure arises not just from local interaction but from interaction based on relatively simple rules (Mihata, 1997: 31). From the rules it is not possible to predict global dynamics, and the emergent phenomena are seen to have generative powers. For example, history and tradition become emergent determinants of how a society is self-organized (Turner, 1997: xvii; Lee, 1997: 21; Mihata, 1997: 36-7). And so complexity science, through a social science account of complexity science, could be understood as an emergent and self-organized phenomenon, one that is not predictable from the interactions of the underlying processes.

For the social sciences, then, the significance of nonlinear, self-organized emergence is that it offers the possibility of thinking about social systems in ways that avoid the limitations of models dependent on homeostatic principles, for example as dominated Parsonian sociology (Byrne, 1998; Baker, 1993; Reed & Harvey, 1992). With complexity science, change, for example, can be conceptualized through the model of dissipative structures emphasizing far-from-equilibrium dynamics in contrast to homeostatic stability (Byrne, 1998: 47). Similarly, the micro/macro relationships can be conceptualized in terms of the chaos/complexity framework, which, in contrast to rational choice models where collective action is conceptualized as aggregate behavior, emphasizes the emergent properties of interactive dynamics (Byrne, 1998: 48-51). And, addressing the relationship between agency and structure, Byrne argues that perturbations of far-from-equilibrium conditions can originate in the values and actions of humans themselves (Byrne, 1998: 50 quoting Reed & Harvey, 1992: 370). Within this framework, complexity science not only serves as a basis for demarcating the distinctive character of the social as an object of knowledge, but also allows for the reflexive, knowledge informed, reconstitution of the social order (Byrne, 1998: 51). The implication, then, is that complexity science can allow us to think of complexity science as an emergent social phenomenon, one that is constituted through reflexive, knowledgeinformed change.

We could continue this line of thought. We could develop a theoretical model based on complexity science to describe our social system complexity science. We could describe this system through a variety of concepts, for example the edge of chaos, self-organized criticality, co-evolution, fitness landscapes, etc. However, to do this in an explanatory rather than merely descriptive way, we need to establish the basis on which we can investigate complexity science as a complex system. This prompts the question: On what basis can we claim that these characteristics represent the reality of our system complexity science? How can we explore the social dynamics of complexity science through the models of complexity science? I want to suggest that we can make a distinction between two methodological approaches on offer from the social sciences: the study of complex social orders, and the study of complex social orderings.


One approach in the social sciences is to approach the social world as a complex social order, which means analyzing social dynamics through the emergent patterns represented by quantified variables. The intention is to observe the forms of social structures in relation to parameters which are characteristic of the system as a whole (Byrne, 1997: 3). This would mean that our understanding of the social system complexity science would need to be based on the patterns that emerge through the representation of different variables. The emphasis of this approach is on interpreting the dynamicsin our case the dynamics of complexity sciencethrough ex post observations of the underlying processes. In other words, we would not model the actual interactions of different agents and processes of complexity science, but, rather, explore the relationships emerging between variables, as a representation of the emergent outcomes of these dynamics.

One way of exploring emergent patterns of complex social orders is through nonlinear mathematical models (see for example Alisch et al., 1997; Brown, 1994; Cartwright, 1991; Kiel, 1993; Kiel & Elliot, 1996; Richards, 1990). We would quickly run into trouble here, however, because whatever variable or combination of variables we choose to represent our system of complexity science, we can be confident that we would not attain data with enough time points to draw clear conclusions from nonlinear mathematical modeling (see Byrne, 1998). Further, to set up a nonlinear mathematical equation we would need to establish the feedback relationships within the model, and yet part of the interest in emergent complex systems is to see what relationships emerge. To use these mathematical models, then, not only would we need to assume that the social system complexity science could be represented in a quantified form, but we would also have to assume a priori entirely deterministic relationships in the equations used to represent the processes underlying the emergent patterns.

Another way to study complex social orders has been developed by Byrne (1997; 1998). While he draws on a set of tools (Byrne, 1998: 29), his approach involves developing quantitative methodology that sees the quantitative as itself inherently qualitative (Byrne, 1997: 2). Byrne develops a methodology that draws on large quantitative data sets (e.g. the British General Household Survey) containing within them information generating hierarchical characteristics (emergence) and time series (processes of becoming). In other words, these data provide for the possibility of observing system changes where the forms of social structures can be examined in relation to parameters which are characteristic of the system as a whole (Byrne, 1997: 3). Applying this to complexity science might mean, for example, looking at a variety of variables at different levelsfrom individuals, to departments, to institutions, to nations, etc. over time. This methodology proposes that using quantitative modeling and statistical analysis, qualitative descriptions of change can be gained, and accounts of the social world can be generated for which the complexity sciences provide a suitable ontology and language. One problem here is finding the appropriate quantifiable variables through which to represent the dynamics of complexity science. A second problem is what assumptions of the underlying dynamics we would be making to assume that we were representing complexity science as a complex system in complexity science terms. This refers us to seeing complexity science as a complex social ordering.


To study complexity science as a complex social ordering means that our emphasis is agent based, modeling the processes of interactions of agents at a local level, and exploring the emergent global characteristics of the system (see for example Goodwin, 1994; Gilbert & Doran, 1994; Gilbert & Conte, 1995; Holland, 1998; Kaufman, 1993). In this method, the essential basis of understanding complex social orderings is to set up a model of the interactions of the system, and to explore the emergent dynamic under various parameter conditions. To explore complexity science, then, we would identify a series of interaction agentsperhaps researchersand with, to use Turner's words, successive tweakings of the variables and the connections among them, we can fit the model to resemble that of reality (Turner, 1997: xxvi). The important theme in this approach is that emergent systems are the consequence ofand can therefore by modeled throughthe interactions of the parts through which relational properties emerge. And so, in our endeavor to understand What is complexity science?, we would search for individual agentsperhaps researchers, institutions, or modelsthat are clustered together to form the larger-scale phenomenacomplexity science, or different approaches within complexity science.

The study of complex social orderings, in contrast to identifying variables to observe complexity science as a complex social order, emphasizes the interacting relationships between agents. While these models do involve emergent patterns of interrelationships, the underlying dynamics leading to these emergent interrelationships are predetermined by the modeler. This occurs even where claims are made that the models can learn and be trained (see for example Cilliers, 1998), because the way in which learning or training takes place is predetermined by the modeler. These problems are exacerbated when there is a need to locate parameters within which the interactions take place. Since an important aspect of complex adaptive systems is their local and contingent nature, for us to model in a way to capture the dynamics of the system, our model must repeat the system (Cilliers, 1998: 10), it must be as complex as the system itself (Cilliers, 1998: 58).1 This, of course, is not possible, and therefore to model our system complexity science we need to make various decisions, decisions about agents, their forms of interaction, and the parameters by which they are affected. In contrast to the study of complex social orders, therefore, in which we make ex post observations of the underlying dynamics, an approach based on the study of complex social orderings would require ex ante assumptions about those underlying dynamics.


In order to understand the phenomenon of complexity science through these two approaches, we have two separate problems linked by a similar issue: emergence. To understand complexity science as a complex social order, we need to identify variables to represent particular characteristics of the system. What variables could we use to understand the emergent, complex, self-organized and nonlinear dynamics of complexity science? Would it be the funding of research, the number of researchers, the number of institutions, the number of publications? All of these? Or the form of funding (private sector, state sector), the characteristics of researchers (which disciplines, gender, race/ethnicity), the type of institution (research, business), the type of publication? There are many possibilities and temporalities through which to construct our data. My point is that the performance of complexity science in this way requires us to determine what variables and temporalities we use to characterize the complex social order of complexity science. This is particularly problematic if we consider that the emergent dynamics may mean that new dynamics emerge that our variables do not capture. For example, while we might initially have modeled complexity science as a research system bringing together different academic fields, increasingly complexity science has become a form of management practice, one that indeed blurs the boundaries between research and management (Thrift, 1999). The problem is, therefore, that modeling through quantified variables may leave us ignorant of the emergent dynamics of complexity science, and yet emergence is a central aspect of complex systems, according to the complexity sciences.

Emergence is also a key to understanding the problem faced in understanding complexity science through the methods offered in the study of complex social orderings. Here, the problem is the assumptions that we must make about the underlying dynamics of a complex system. The emphasis of this approach, as we have seen, is on emergent dynamics from the interactions of the underlying agents and processes. Different approaches are possible (see for example Cilliers, 1998: 13-21). In rule-based models, our problem would be deciding, for example, the logical relationship between different rules, as well as establishing the meta-rules of the system. In connectionist models, while they allow for the possibility of self-organization (Cilliers, 1998: 19), we would have to train the model in its learning, and assume that in our system at the level of the individual neuron no complex behavior is discernable, but the system of neurons is capable of performing specific, complex tasks (Cilliers, 1998: 18). Could this allow us to understand complexity science? The implication would seem to be that while the system of complexity science may be complex, be able to perform complex tasks, the individual neurons (be they researchers, research institutes, or conferences, for example) would demonstrate no such complex behavior. Emergence in this case is restricted at different levels, for it is hard to see how we would then understand a research institute, for example, as a complex organization.

The problem of emergence becomes even more pertinent when we consider how we might link the study of complex social orders and complex social orderings. In our study of complexity science, our approach to complexity science as a form of complex social ordering might mean, for simplicity's sake, having a series of researchers interacting and learning in particular ways. We would have established the ways in which those interactions and learning would take place (even where learning takes place, we would have determined how this is done), emergent from which are some interesting complex nonlinear dynamics. If we wanted to study this system in terms of a complex social order, by contrast, we would be examining various variables of these researchers, and showing the complex patterns that emerge and that were thought to represent aspects of the complex nonlinear dynamics. Now, when we introduce some reflexivity into thisarguably one of the key characteristics of the social world in which processes are applied to themselves (Luhmann, 1995: 450-55; see also May, 1998)how could our model of the complex social ordering deal with the introduction of our study of the complex social order? How could our model accommodate this? Could we allow for new forms of interactions, of new learning, of new actors? The point here is this: What assumptions about the dynamics of complexity research are we making if we assume that our model of complex social orderings can repeat the system of complexity research? Can we even begin to identify a model of rules, behaviors, relations, or learning that would stand in response to our own performance of complexity science?

There is an argument that the problem with which we are confronted is essentially deterministic assumptions within complexity science. For example, while dissipative structures are described in terms of indeterminacy at the point of bifurcation, this indeterminacy is the problem of observation (Capra, 1996: 187). It is a problem of observation, not something that lies in the ontology of the systems. Indeed, complexity science might be better understood in terms of a delinking of determinism and predictability, as Kumminga argues of chaos theory (1990: 58). The structures of the various models are fundamentally deterministic. For example, in the rule-based models, rules (and meta-rules) determine the parameters of behaviors of the various agents within the model. Even in the connectionist models, the transfer function of each node, and the possible connections between nodes, are set up by the designer and do not change in time (and if they could, this would depend again on their setup by the designer; Cilliers, 1998: 16-18). While connectionist models can learn and can be trained, the conditions under which this is possible are determined by the designer.

The problem of determinism is highlighted in the scenario I presented earlier of introducing our model of complexity science into the dynamics of complexity science. It is an essential problem of modeling social dynamics (see Denning, 1990; Renfrew, 1987). There is a claim that social systems differ from physical systems because in social systems, perturbations may come from within the system itself (Reed & Harvey, 1992: 370; Byrne, 1998: 50). However, the problem with modeling and simulation is that the models of complex systems remain closed systems as far as the internal operations are concerned, and therefore the perturbation is represented as a variable. The models have no way of changing the operations within themselves unless programmed to do so (thus referring the closure to a second-order level).2 They cannot account for change within the relationships themselves, only their outcomes, and in this sense remain deterministic.

This problem of hidden determinism highlights a sense in which a rather reductionist approach to emergence is developed in the performance of complexity science. There is reductionism in so far as the properties and dynamics of the system are characterized as emergent from lower-level dynamics. Explanation of the emergent relations and the emergent system comes from interactions below operating according to particular rules, relationships, or variables. This form of reductionism in the extreme opens the way for approaches drawing on complexity science that advocate a biological basis for explaining social processes (see for example Smith, 1997). Reductionism also means that the limitation of determinism in explaining social dynamics becomes particularly clear. First, if it is the case that interacting individual agents constitute social systems, for example the economy (e.g. Cilliers, 1998: 7), the modeling process of complex systems needs to incorporate the ways in which those individuals interact. The problem of the modeling process, as I have argued, is that it has to assume a priori the conditions structuring those interactions, which refers human agency to rule-based dynamics (see Hatch & Tsoukas, 1997). Second, part of determining those dynamics refers to determining the way in which the properties of the whole feedback, and influence, those dynamics. For Cilliers, this is not necessary because agents in complex systems operate according to distributed representations where the parts are ignorant of the whole (Cilliers, 1998: 5). However, this seems to confuse representation with accurate representation. Human actors, as the process of making sense of complex systems makes clear, do reflect on the dynamics of the system as a whole, however inaccurate their representations may be, and this can inform their actions. There can be a relationship between the whole and the part, but how this occurs we cannot assume a priori, nor can we assume that it is static.

One response to this situation is to suggest that our understanding of social dynamics through complexity science needs to incorporate a more complex sense of human agency, one with the capacity to reflect upon, and to contradict, any mathematical description made of them (Castoriadis cited by Tsoukas, 1998a: 305). This more complex sense of agency means a move away from naïve assumptions about the homogeneity and rationality of social actors (Amin & Hausner, 1997: 5; Shackley et al., 1996: 215). As a variety of authors writing about complexity and the social world have suggested, what we would look to is maintaining ways of understanding human agency and being in terms of reflexivity, performativity, narrative, intersubjectivity, and poetics (see for example Chia, 1998; Dillon, 2000; Griffin & Shaw, 1998; Murray, 1998; Tsoukas, 1998a, 1998b). This sense of agency would enable us to think about the processes by which complexity science is performed in the social world. When we begin to explore the ways in which complexity science is performed in the social world itself, that is, the ways in which the phenomenon of complexity science emerges, what we see are different performances (see Shackley et al., 1996; Thrift, 1999).3 This suggests the limits of modeling per se, and leaves us back where we started with the problem of the basis on which we can claim that complexity science offers us a way of representing the social world.


To summarize my argument so far, I have tried to offer a performance of the way the social sciences have developed complexity science to make sense of social dynamics, and I have done this through a case study of a particular social phenomenon, complexity science. It has been a performance of the sorts of approaches and assumptions that we might need to make in such a study. While not a comprehensive account of all the possibilities, I have made a distinction between the study of complex social orders and the study of complex social orderings, and I have argued that both are problematized by emergence.

We are faced with two problems. First, there is the problem of reflexivity at the level of the individual (e.g., researchers responding to the research in ways we cannot predetermine) and at the level of the system (e.g,. the way complexity science might change as a response to models of itself). Second, we are also faced with the problem of trying to understand complexitythe complexity of complexity sciencewhile contributing to that complexity (see Luhmann, 1995, 1998 on this paradox). It seems that more work needs to be done if complexity science is to offer explanatory accounts of social dynamics, and I want to make some suggestions about what this needs to involve.

The essence of my starting point is complexitycomplexity referring to the impossibility of connecting all elements together, the impossibility of complete observation and representation (Luhmann, 1995). This is a problem for research, for knowledge production, but it is a problem characteristic of the social world in general:

Social life is de facto organized: we, as sentient beings, have no choice but to organize our world and our actions in it. The interesting questions are how we do it; what we do it for? (Tsoukas, 1998a: 292)

The problem is, who does we refer to? There is a temptation here to refer to human agency once more. A complex human agency maybe, but to understand the system of complexity science we would assume again that the emergent system can be understood as a global dynamic arising from the interactions of the parts. Unanswered, however, would be how do the agents of the social worldour world of complexity science become agents of that world? Looking at the processes of modeling gives us a clue. For simplicity's sake, let us imagine that as a modeler of the system complexity science, we would first structure relationships into the models that were deemed interactions of the system complexity science. Already, in the design of the model, before the agents do anything as such, they are to represent interacting individuals in the social world in such a way that they represent the particular system of complexity science. The clue is that this process occurs in the emergent system of the social world itself: agents of social systems are constructed as agents by that system. The difference between how this is done in an actual social system and the modelparallelling Capra's argument about living systems (Capra, 1996)is that the social system constructs the agents itself.

These arguments can be developed through Luhmann's (1990, 1995, 1998) work on social systems, which is based on conceptualizing complexity, emergence, and self-reference in relation to meaning and the social world. For Luhmann, complexity refers to the impossibility of complete observation and representation, which leads to the necessity of making distinctions and therefore selections. It is these selections that constitute system formation. In this context, emergence refers to the organization of selected relationships as differentiated from their environment. The emergent system is formed through a system/environment difference constituted through the selections. However, a key aspect of this is self-reference, referring to the way in which the system itself determines the selections. There is not the space to detail Luhmann's interpretation of these concepts here (see Luhmann, 1990, 1995); however, we can see the way in which complexity, emergence, and self-reference are implicitly built into the assumed ontologies proposed by complexity science: agents interact according to particular selective relationships (complexity reduction), differentiated from other possible relationships in their environment (emergence), where these selective relationships are structured from within the system (self-reference).

For Luhmann, the social world emergesin the sense that it is differentiated from other worldsas a consequence of the problem of double contingency, in which one system's action is dependent on another system's action, where the latter system's action is dependent on the former system's action. Since the complexity of each system means that neither system can know the other system, each requires a means to determine its action by reducing the complexity of possibilities of the other system's action. This refers to a coordinated selectivity constituted in communication that is not reducible to either system but, rather, is differentiated from the systems. However, communication is complete, constituted, back to front, which refers to the observation by the recipient of intended information communicated by the utterer. It is the constitution of communication that refers to the emergent social world as a form of structured complexity differentiated from other complex worldsa world that is not simply arising out of micro-level components, but is an emergent system constituting those components. How that constitution occurs needs to be the focus of our inquiry.

If we develop an account based on emergence as differentiation, what does this imply about my performance of complexity science? It suggests that the work I do with my computer, my thinking, my motives, etc. are not part of the social system complexity science per se. They might be part of my world and other worlds, but the world of complexity science is a world of communications. As a performance of complexity science in the social world, my work is not complete. It can only become a performance in so far as I succeed in a communication that contributes to complexity science. Developing Bhaskar's (1989) point that things are real in so far as they produce effects, and Luhmann's argument that the social is constituted in so far as communication is received, a different approach emerges. In Luhmann's account, the effect cannot be attributed to the event, for it depends on the system's responses to the event (see discussions on autopoiesis for this: Capra, 1996). One implication is that events do not have singular realities but become multiple. Putting this another way, if we can only assume the reality of an event through the relations in which it is constituted, and if these relations are many, then it implies the need to think about the multiple status of these realities (see Mol, 1999). This suggests that processes of understanding complex social dynamics need to explore the ways in which the realities of the dynamics are constituted in different ways, by different systems. This implies a dense interconnectivity that comes hand in hand with differentiation.

I want to make one final point. In looking at how realities get made, there is an important aspect of this process of dealing with complexity to consider: the fact that complexity refers to the impossibility of complete observation refers to the necessity of ignorance as a characteristic of claims made on the world (Luhmann, 1995, 1998). This refers to our problem of knowing the social world, but also to the problem in the social world, since the implication of ignorance is contingency and contingency means risk (Luhmann, 1995: 25). Ignorance becomes the productive dynamic of complexity systems; it is the motivating other of complex systems (cf. Dillon's, 2000, discussion of the different others in poststructuralism). In a sense, then, our focus on complexity becomes a question of how we can deal with ignorance in the world. Unless we can repeat the system and build a model as complex as the system itself (Cilliers, 1998: 10), we have to ignore, and an important part of this ignorance is that we cannot observe ourselves in the processes of observing. We cannot stand outside. In Luhmann's terms, we live in an ecology of ignorance (Luhmann, 1998) in which we produce complexity in trying to deal with complexity, new ignorance in trying to know. It is not just that ignorance is a problem, it also makes things possible. For how could we make any claims about our system complexity science if we had to account for everything? And yet we are producing that system, without such knowledge and representation. To deal with this, therefore, complexity science needs to begin to ask Luhmann's (1998: 78) question: How is ignorance dealt with?


This article is based on my PhD thesis funded by the Economic and Social Research Council. I am grateful to my supervisors and to all those in the Department of Sociology, Lancaster University who enabled that work to happen, especially Dan Shapiro and Paul Haynes, and also to my examiners, Dave Byrne and John Urry. I'm also grateful to the anonymous reviewer for very constructive feedback.

  1. There are some other key issues that would need to be considered here. First, in order to repeat the system, Cilliers argues that the model must first be able to store information. In terms of storing information, the system requires a representation of the environment in order to adapt to it. Cilliers argues that meaning (by which he refers to representation) in a complex system is conferred by the relationship between the structural components of the system itself Meaning is the result of a process, and this process is dialecticalinvolving elements from inside and outsideas well as historical, in the sense that previous states of the system are vitally important. The process takes place in an active, open and complex system (Cilliers, 1998: 11). Because meaning is constituted in this waythrough which elements only have meaning through distributed representation (1998: 11)Cilliers argues that the abstract level of meaning (the semantic' level) becomes redundant, or, rather, unnecessary, for the process of modelling the system (1998: 11). Second, the model must be its ability to generate change in order to adapt to the environment. This refers to self-organization (1998: 10), which implies that a system can develop a complex structure from fairly unstructured beginnings (1998: 12). This requires that the system must be able to adapt, it cannot have a rigid program, the system must be plastic (1998: 12). Having rejected the use of rule-based models, Cilliers claims that neural networks (or the connectionist model) are better able to offer understanding of complex systems. In particular, the connectionist models can be set up in order to learn and be trained (1998: 17-18).
  2. See discussion by Sayer (1992: Chapter 6), who explores the relationship of closed and open structures in the context of quantitative methods in the social sciences.
  3. Shackley et al. (1996) argue that while researchers may well be aware of limitations and may themselves be reflexive about the theoretical, methodological, and policy implications, when transferred to the policy realm the simplifications become more acute and will virtually automatically filter out the more contextualist and relativist components in favor of the more straightforward realist arguments (Shackley et al., 1996: 202). The consequences of these simplification processes, as my critique of complexity science has begun to indicate, lead to woefully inadequate representations of the full uncertainties and indeterminacies in the system in question (Shackley et al., 1996: 206).


Alisch, L. M., Azizighanbari, S., & Bargfeldt, M. (1997) Foreword: Chaos and social science, in R. A. Eve, S. Horsfal, & M. E. Lee (eds) Chaos, Complexity and Sociology: Myths, Models and Theories, London: Sage.

Amin, A. & Hausner, J. (eds) (1997) Beyond Market and Hierarchy: Interactive Governance and Social Complexity, Cheltenham, UK: Edward Elgar.

Back, K. W. (1997) The persistence of emergence, in R. A. Eve, S. Horsfal, & M. E. Lee (eds) Chaos, Complexity and Sociology: Myths, Models and Theories, London: Sage.

Baker, P L. (1993) Chaos, order and sociological theory, Sociological Inquiry, 63(2): 123-49.

Bhaskar, R. (1989) *Reclaiming Reality: A Critical Introduction to Contemporary Philosophy, * London: Verso.

Brockman, J. (ed.) (1995) The Third Culture: Beyond the Scientific Revolution, New York: Simon and Schuster.

Brown, C. (1994) Chaos and Catastrophe Theories, London: Sage.

Byrne, D. (1997) Complexity theory and social research, Sociological Research Update, 18.

Byrne, D. (1998) Complexity Theory and the Social Sciences: An Introduction, London: Routledge.

Capra, F. (1996) The Web of Life: A New Synthesis of Mind and Matter, London: HarperCollins.

Cartwright, T. J. (1991) Planning and chaos theory, Journal of American Planning Association, 57: 44-56.

Casti, J. L. (1994) Complexification, London: Abacus.

Chia, R. (1998) From complexity science to complex thinking: Organization as simple location, Organization, 5(3): 341-69.

Cilliers, P. (1998) Complexity and Postmodernism: Understanding Complex Systems, London: Routledge.

Cohen, J. & Stewart, I. (1994) The Collapse of Chaos: Discovering Simplicity in a Complex World, Harmondsworth, UK: Penguin.

Denning, P (1990) The science of computing: Modeling reality, American Scientist, 78 (Nov-Dec): 495-8.

Dent, E. B. (1999) Complexity science: A worldview shift, Emergence, 1(4): 5-19.

Dillon, M. (2000) Postructuralism, complexity and poetics, *Theory, Culture and Society, * 17(5): 1-26.

Eve, R. A., Horsfall, S., & Lee, M. E. (eds) (1997) Chaos, Complexity and Sociology: Myths, Models and Theories, London: Sage.

Gilbert, N. & Conte, R. (eds) (1995) Artificial Societies, London: UCL Press.

Gilbert, N. & Doran, J. (eds) (1994) Simulating Societies: The Computer Simulation of Social Phenomena, London: UCL Press.

Glieck, J. (1987) Chaos: Making a New Science, London: Penguin.

Goldstein, J. (1999) Emergence as a construct: History and issues, Emergence, 1(1): 49-72.

Goodwin, B. (1994) How the Leopard Changed its Spots, London: Phoenix Giants.

Griffin, D. & Shaw, P (1998) Speaking of complexity in management and practice, Organization, 5(3): 315-39.

Hall, N. (ed.) (1992) The New Scientist Guide to Chaos, London, Penguin.

Hatch, M. J. & Tsoukas, H. (1997) Complex Thinking About Organizational Complexity: The Appeal of a Narrative Approach to Complexity, paper presented to OMT Division for Presentation at the Academy of Management, Boston.

Haynes, P. (1999) Complex Policy Planning: The Government and Strategic Management of the Social Care Market, Aldershot, UK: Ashgate.

Holland, J. (1998) Emergence: From Chaos to Order, Reading, MA: Addison-Wesley.

Kauffman, S. (1993) The Origins of Order, Oxford, UK: Oxford University Press.

Khalil, E. L. (1996) Social theory and naturalism: An introduction, in E. L. Khalil & K. E. Boulding (eds), Evolution, Order and Complexity, London: Routledge.

Kiel, L. D. (1993) Nonlinear dynamical analysis: Assessing systems concepts in a government agency, Public Administration Review, 53(2): 143-53.

Kiel, L. D. & Elliot, E. (eds) (1996) Chaos Theory in the Social Sciences: Foundations and Applications, Michigan, University of Michigan Press.

Kumminga, H. (1990) What is this thing called chaos? New Left Review, 181(May/June): 49-59.

Lee, M. E. (1997) From enlightenment to chaos: Toward nonmodern social theory, in R. A. Eve, S. Horsfall, & M. E. Lee (eds), Chaos, Complexity and Sociology: Myths, Models and Theories, London: Sage.

Lewin, R. (1993) Complexity: Life on the Edge of Chaos, London: Phoenix.

Luhmann, N. (1990) Essays on Self Reference, New York: Columbia University Press.

Luhmann, N. (1995) Social Systems, Stanford, CA: Stanford University Press.

Luhmann, N. (1998) Observations on Modernity, Stanford, CA: Stanford University Press.

May, T. (1998) Reflexivity in the age of reconstructive social science, International Journal of Social Research Methodology , 1(1): 7-24.

Mihata, K. (1997) The persistence of 'emergence,'” in R. A. Eve, S. Horsfall, & M. E. Lee (eds), Chaos, Complexity and Sociology: Myths, Models and Theories, London: Sage.

Mol, A. (1999) Ontological politics: A word and some questions, in J. Law & J. Hassard (eds), Actor-Network Theory and After, Oxford, UK: Blackwell.

Murray, P. J. (1998) Complexity theory and the fifth discipline, Systemic Practice and Action Research, 11(3): 275-93.

Reed, M. & Harvey, D. L. (1992) The new science and the old: Complexity and realism in the social sciences, Journal for the Theory of Social Research, 22: 353-80.

Renfrew, C. (1987) Problems in the modelling of socio-cultural systems, European Journal of Operational Research, 30: 179-92.

Richards, D. (1990) Is strategic decision making chaotic?, Behavioral Science, 35: 219-32.

Sayer, A. (1992) Method, in A. Sayer, Methods in Social Science: A Realist Approach, London: Routledge.

Shackley, S., Waterton, C., & Wynne, B. (1996) Imagine complexity! The past, present, and future potential of complex thinking, Futures, 28: 201-25.

Smith, T. S. (1997) Nonlinear dynamics and the micro-macro bridge, in R. A. Eve, S. Horsfall, & M. E. Lee (eds), Chaos, Complexity and Sociology: Myths, Models and Theories, London: Sage.

Thrift, N. (1999) The place of complexity, Theory, Culture, Society, 16(3): 31-69.

Tsoukas, H. (1998a) Introduction: Chaos, complexity and organization, Organization, 5(3): 291-313.

Tsoukas, H. (1998b) Forms of knowledge and forms of life in organized contexts, in R. Chia (ed.), In the Realm of Organization: Essays for Robert Cooper, London: Routledge.

Turner, F. (1997) Foreword: Chaos and social science, in R. A. Eve, S. Horsfall, & M. E. Lee (eds), Chaos, Complexity and Sociology: Myths, Models and Theories, London: Sage.

Urry, J. (2000) Mobile Sociology, British Journal of Sociology, 51(1): 185-203.

Waldrop, M. M. (1992) Complexity: The Emerging Science at the Edge of Order and Chaos, London: Viking.