Reviewed by
Mario Paolucci
ThinkingGolem, Piazza Prati
degli Strozzi 21, 00195 Roma, Italy.
The purpose of this book is to promote and support the application of complex systems theory to organisation theory. There are good motivations for trying to do so; complexity is advocated by many as a tool for bridging the gap between more traditional methodologies, employed in the social sciences in general, and the more formalised and mathematically rigorous ("hard") sciences. After strong growth in the second half of the '80s (see for example Anderson et al. 1988), followed by a period of relative calm, complexity and complex (adaptive) systems are now rallying their strength. The field is currently entering a new period of expansion, as evidenced by the significant presence of EC funded projects and networks in the field (Exystence, COSI and so on). The wide applicability of complexity theories and the need for formalisation felt in organisation theory makes this an interesting and potentially fruitful match. For example, the journal Emergence published a discussion about this interaction in its first number which came out at almost exactly the same time as Marion's book (McKelvey 1999).
The book begins by introducing the building bricks for understanding complexity, starting with the suggestive assertion that "Ancient Hebrews and Charles Darwin agreed that life resulted from the sifting of Order from Chaos" (p. xi). The approach is to apply complexity to the different organisation theory disciplines, as they are discussed through the rest of the book. In the introduction, the first abstraction discussed is that of the "Edge of Chaos" (EOC), a beautiful metaphor that also inspired the title of the book. The EOC is the middle point between two extremes. The first is that of conservative, integrable systems where everything is predictable and well defined given the initial conditions. The drawback is that there is no space for adaptability or flexibility, characteristics of social systems and more generally of living systems. The second extreme is that of chaos, in the sense used in the theory of dynamical systems. Here, initial conditions do not form an acceptable basis for predictions and the only order that can be found operates at a higher level, through the presence of structural characteristics such as attractors. In Marion's words:
"[c]haotic order per se is actually too violent, too changing to describe much that goes on among living beings. Complexity theorists ... argue that life tunes Chaos's intensity down a bit to a transition band between Chaos and predictable stability called the Edge of Chaos. Dynamics in this band are still Chaotic but they also possess characteristics of order. Full-blown Chaotic Systems flit a bit too readily from novelty to novelty; living systems need to consolidate gains. Predictable, stable systems,by contrast, possess none of the panache needed to create new order or even to respond adaptively to creative environments. Complex Systems lie between these poles, at the Edge of Chaos, and they have both panache and stability sufficient to serve life." (p. xiv)
The mathematical level of the presentation is very simple and the author seems to excuse himself every time he cannot avoid showing an equation. This fact characterises the book as oriented more towards organisation scientists who want to have a first look at complexity issues than to the complexity theorist studying organisation science. Indeed, in my opinion, a complexity theorist will find that the presentation of organisation theory is often no more than a brief reminder and the lack of data and formalisation makes it difficult to develop any applications on the basis of what has been read. By contrast, the basic presentation of complexity seems to be quite accurate and will probably stimulate an organisation theorist to further reading.
In the first chapter, along with a short story about chaos theory, Marion proposes a thought experiment called Einstein's Island. Would Albert Einstein develop the theory of relativity if born on a desert island with full access to printed knowledge but without the interaction of other researchers? The purpose of the experiment is to induce the reader to think about the importance of connectivity in our society - what would become of us without our networks of friends and relatives, our network of economic partners for everyday transactions, our network of information providers? Do the networks around us merely provide resources or they do affect our thinking in a deeper way? To what extent, and in what sense, can we endorse the thesis of Berger and Luckmann (1966) that "Man is a social product"? The answer is postponed until the end of the book. The chapter then concludes with a list of questions to be addressed, all of which are of the strongest interest for a social scientist: "Where do informal groups, cliques, fads, rumors, organizational myths, marked demand, riots, social movements, and new paradigms come from, and why do they tend to appear suddenly?" (p. 9)
In the second chapter, after a very basic and orthodox introduction to chaos, one of the main theses of the book is stated, that "nearly all social activity can be metaphorically described with a strange attractor (or variation thereof) ... many social behaviors can be physically described with an attractor" (p. 22). The meaning of the word "described" is critical here and in the rest of the book. Indeed, at least two different levels of description can be employed: the metaphorical level and the concrete level, the latter implying the possibility of building mathematical or simulation models of social activity.
The fact is that, while the metaphorical level is useful to challenge our intuition, locked as it is into the linear understanding of linear phenomena, the real advantage would come from the application of explicit models; such applications are still lacking and rare. The author is aware of "the paucity of study to support it, but Social Chaos Theory is, after all, a developing hypothesis." (p. 22, my italics). This position is susceptible to an obvious weakness: the risk of using the metaphorical description alone. But the deeper problem here is whether the application of metaphor is really helping in the understanding of social phenomena or just providing a fancy label for things that were already known? Precisely because the ideas of chaos, complexity, the EOC and attractors are so appealing and potentially share many desirable features, the use of these words can degenerate into hype and equally spurious support for contradictory theories. This risk is emphasised in McKelvey (1999) which suggests the rigorous use of formal simulation models to prevent the application of complexity to organisation science degenerating into a fad. Indeed, a more concrete and less metaphorical approach is necessary for prediction. Generally complexity could fail to become a proper science. Prediction is our best means of distinguishing science from pseudo-science and complexity risks becoming one of a long line of highly mathematical "theories of almost everything" that have gripped the imagination of scientists in this century (Horgan 1995, Phelan 2001). Failing the test of prediction, the result is a theory "explaining" everything - but with "explanation" reduced to the application of a uniform label to everything, including contradictory results, and consequently in the loss of any meaning for the concepts and tools involved. I will return to this point in the conclusion.
In chapter three, the emergence of global patterns from low-level interactions is discussed, passing through Darwinism, cellular automata (with the inevitable Game of Life), emergence and even more specialised topics like solitons (stable waves in non linear dissipative systems). Curiously enough, the systems that give rise to solitons are completely integrable (the opposite of chaotic) despite being an acceptable metaphor for some complex social phenomena.
Chapter four presents an interesting historical analysis of the consequences which derive from the success of Newton's linear theories. While paving the road for most science as we know it today, Newton's legacy (because of its vast objective success) has also been a constraint on further theory development, biasing it thenceforward towards linear, time-reversible systems. One consequence of this is that teleological reasoning has become ingrained in western thought; the linearity of Newton's laws led to an implicit assumption that for each effect a proportional cause can be found. This supposition does not fit well at all with well-known "catastrophic" changes that occur in social contexts, from the stock crash in 1987 to the riots following the 1992 Rodney King verdict in Los Angeles. Thus, physics could be found guilty of having attracted the social sciences towards an over simplistic path; only with the recent study of non linear dynamical systems has the formal apparatus of hard sciences recovered the potential to explain such phenomena.
Chapter five introduces open systems theory which became popular in the 1960s, suggesting that organisation theories cannot be confined to the insides of organisations, but must take the environment into consideration using a holistic approach. Several examples are discussed and the issue of feedback is introduced, giving rise to discussion of Brian Arthur's theory of increasing returns and David's discussion of inertia via frozen accidents like VHS dominance, QWERTY keyboards and (!) gas versus steam engines (Arthur 1994, David 1985). In the conclusions, the idea of a Complex Adaptive System (CAS) is introduced in the spirit of the Santa Fe Institute as an "adaptive, interactive network of actors ... structured by physics and teleology ... refined by selection". Teleology here refers to the ability of social systems to maintain goals and display intentional behaviour.
Chapter six lists the four main branches of organisation theory as they developed from open systems theory. These four theories are divided into two groups: prescriptive theories, focusing on prescriptions for the effective leader (Structural Contingency and Resource Dependence) and organic theories which argue that activities and structure are determined more by the environment than by the actions of leaders (Enacted Environment and Population Ecology).
In what follows, Structural Contingency Theory is discussed in terms of the CAS approach; the main point is that there is a compromise between the slack needed to adapt to the contingencies of the environment (which must be obtained by internal differentiation) and the co-ordination, conflict and integration problems caused by that differentiation. In one of the studies quoted, the result was that "the most productive companies were also the most conflictive of the organizations studied" (p. 86).
Chapter seven introduces Kauffman's networks of N units governed by boolean rules. In these networks, the main parameter is the number of connections (K) that determines the number of attractors appearing in the dynamics. For K between 5 and N, the number of attractors is quite large (of the order of N) and very connected (the system can easily move from one attractor to the other if perturbed). For K=1 the attractors are of big dimension and essentially disconnected while for K=2 the attractors are roughly of the square root of N in number and connected but require strong perturbations to produce transitions. Moreover, by tuning another parameter (P) which measures bias in the connections, the number of connections needed to have (in Marion's words) "useful order", can be shifted from K=2 which is a number of connections hardly plausible in a social setting. The chapter goes on by examining, with reference to a study by Blau and Schoenherr (1971), organisation size in terms of number of positions, levels and sections, looking for the square root principle mentioned above. The data fit is not impressive, but it seems to follow some kind of exponential law with an exponent of less than one. In Marion's words, we "... fully realize that science demands a more rational explanation than this mildly mystical one".
It is worth mentioning that, to test this hypothesis, Marion had to resort to manual point sampling from the published graphs in Blau and Schoenherr's work; original data were not available. This is just another paradigmatic example of how difficult it can be to obtain real data in the social sciences. This contrasts with what happens in the physical sciences where replication of experiments is easier and allowing for data availability is an established practice.
Chapter eight starts with a description of the Gaia hypothesis and its apparent conflict with evolutionist thinking. It then settles on a discussion of inter firm alliances and interdependency, with a nice example of inter firm co-ordination in Japan. McKelvey's basic unit for organisations, the comp, a template that dictates the structure of formal organisations, is introduced as a special case of the meme; as such, the comp is subject to the same evolutionary forces as the meme is.
In this chapter, at least a couple of points are debatable. The first is a sentence asserting that "... Dawkins argued ... that the Gaia hypothesis was fatally flawed because there were no worlds competing for survival in the model" (p. 114). There is no explicit reference and a short search did not discover this assertion in the work of Dawkins. This critique could be equally applied to nearly any of the natural sciences and in my opinion does not make sense at all; one is left to wonder if what Dawkins argued was at least more elaborated. A reference would have helped. The second point, a considerable slip of the pen, is the claim that two organisations competing for the same fixed amount of resource can be modelled as a "zero sum game, the type described by Prisoner's Dilemma" (p. 117). In fact, the PD is one of the prototypes for the class of non-zero sum games.
Chapter nine introduces the debate on determinism and causality based on a presentation of the Enacted Environment theory of organisations. This theory claims that "causation is too complex to be controlled ... we take credit for change after the fact" (p. 137) and that generally we rationalise events as if they had been planned deliberately when in fact we act according to passion or just by chance.
In Chapter ten, the coupling of systems is studied, according to the thesis that "fit systems are neither tightly coupled nor loosely coupled, rather they exist in an in-between level of coupling. That is, they exist at the Edge of Chaos" (p. 154).
Chapter eleven deals with natural selection and its application to population ecology theory. Fitness landscapes are described along with the thesis that the number of optima (peaks in the fitness landscape) increases with connectivity K and that high K brings about a large number of low fitness peaks and very few highly fit ones (chaotic catastrophe). If the fit peaks are very small, then the forces of evolution are no longer able to guarantee upward motion, since the number of disadvantageous mutations grows much larger than the number of good ones (error catastrophe). Of course, the solution to be found in the middle.
In chapter twelve, a very simple study of the logistic equation is presented. The subject does not allow the complete avoidance of mathematics, but the reader is advised that the discussion occurs in the spirit of the following phrase, used to comment on the conclusions: "It may be mathematical, but if you think about the whole thing conceptually, it makes perfect sense" (p. 201). Some more details are given on competition and on r-strategies and k-strategies.
In Chapters thirteen and fourteen, Marion discusses the problems of change, resistance to change and organisation dynamics in unstable environments and/or under perturbation. The theory of catastrophes is introduced starting, as usual, from the dynamics of sand slides and the power law distribution. Kauffman's work is then discussed again with the addition of S different species of actors connected by reciprocal dependence. The strength of this dependence is measured by a parameter C, obtaining the so called NKCS model. This model, originally built for the explanation of biological niches, is applied here to McKelvey's comps argued to be their organisational equivalents. The thesis here is that networks will evolve towards the EOC, since the extremes of order and chaos are unstable; the first because it is incapable of finding deep minima, the second because it will amplify perturbations, eventually jumping out of the deeper minima that it was able to find. To illustrate this point, a short but complete history of the microcomputer is presented, emphasising how its sudden appearance had in fact been presaged for years, while a sequence of necessary technologies (cathode tubes, microprocessors, memories, computer language logic) were set in place one by one. Their coming together marked a revolution, but would have been impossible were all the pieces not available for integration. More generally, is argued (again following Kauffman) that if systems have low internal complexity (K), the best results are obtained by large organisations with few divisions and with centralised decision-making. If instead the internal complexity is high, it is better to have smaller patches with less emphasis on hierarchical control and more on internal competition. This result, while well-argued in Kauffman's terms, was already present in the literature (Marion quotes Corvin 1987). While the network theory could have provided more detail, we are still in difficulty because of the problem of measuring parameters like K in real cases. Unfortunately abstract general results like the imperative "Organizations must learn to couple themselves at the Edge of Chaos" (p. 225) have a scent of hype about them.
Chapter fifteen should have been the place to introduce useful and applicable instruments to allow the researcher to experiment by himself with the ideas introduced in the book. Instead, this chapter seems to be little more than an introductory chapter on chaos, discussing Poincaré maps, prey-predator systems, Heiles equations and even PCA, but lacking any strong connection to organisation theory. There is no innovation in the application of the prey-predator model to organisation theory by simply asserting that a variable can represent the number of predator firms instead of the number of foxes. Strangely, this chapter seems to forget every specific organisation theory that was discussed in the previous chapters and the EOC is nowhere to be found!
The book concludes with an epilogue on Einstein's Island. The relevance of connections and networks to the functioning of complex systems should have made the answer obvious for most readers; I will not spoil the surprise for the rest.
As I observed at the beginning of this review, Marion's book appropriately promotes the application of an interesting (and relatively new) paradigm, that of complexity, to an area that has a strong need for new conceptualisations. I completely agree with the desirability of this discussion and the general merits of his proposal. Moreover, Marion recognises that this work is still in its earliest phase and that there are not yet striking results available. However, I have some reservations about his specific approach. As I said before, the lack of mathematical and formal depth does not encourage the reader to look at more formal models, a step that is necessary in my opinion. Of course, the endemic lack of data (recalling the procedure Marion had to follow with regard to the work of Blau and Schoenherr), one of the problems affecting the whole domain of social science, does not make things easier.
Another missing ingredient, whose absence I found surprising, was the application of simulation in the social sciences. Simulation is hardly mentioned at all and the concepts of "agent" and "agent-based simulation" are completely missing. Neither "simulation" nor "agent" has an entry in the subject index. This absence might also explain the feeling of non-relatedness caused by reading the later chapters. An unfolding multi-agent simulation could better have filled the role of a concrete example for the theory, as has already been suggested by some voices in the field. For example, Goldspink (2002) argues that "... [c]omputer based simulation is increasingly argued to constitute an important method for studying phenomena arising from organized complexity". The strength of the book is in its clear introduction to complexity themes, in a language that would not discourage (and might possibly attract) managers and organisation students. Its main weakness is the "suspended" feeling that one has after reading it. The reader is left with little indication of how and where to go next in the field, especially if they want to avoid what McKelvey (1999) calls "questionable scientific standards in organisation science" (p. 6)..
ReferencesANDERSON P. W., K. J. Arrow and D. Pines, editors, 1988. The Economy as a Complex Evolving System. Addison-Wesley, Redwood City, CA.
ARTHUR W. B. 1994. Increasing Returns and Path Dependence in the Economy. University of Michigan Press, Ann Arbor, MI.
BERGER P. L. and T. Luckmann 1966. The Social Construction of Reality. Doubleday, New York, NY.
BLAU P. M. and R. A. Schoenherr 1993. The structure of Organizations. Basic Books, New York, NY.
CORVIN R. G. 1987. The Organizational-Society Nexus: A Critical Review of Models and Metaphors. Greenwood, New York, NY.
DAVID P. A. 1985. Clio and the economics of QWERTY. American Economic Review, 75:332-337.
GOLDSPINK C. 2002. Methodological implications of complex systems approaches to sociality: Simulation as a foundation for knowledge. Journal of Artificial Societies and Social Simulation, 5, <https://www.jasss.org/5/1/3.html>.
HORGAN P. 1995. From complexity to perplexity. Scientific American, 272:74-79.
MCKELVEY B. 1999. Complexity theory in organisation science: Seizing the promise or becoming a fad? Emergence, 1:5-32.
PHELAN S. E. 2001. What is complexity science, really? Emergence, 3:120-136.
Return to Contents
of this issue
© Copyright Journal of Artificial Societies and Social Simulation, 2002