lunes, 8 de octubre de 2012

Hierarchy and general systems

Hierarchy and History in Simon's "Architecture of Complexity" 

Philip E. Agre 

Department of Information Studies 
University of California, Los Angeles 
Los Angeles, California 90095-1520 
USA 
pagre@ucla.edu 
http://polaris.gseis.ucla.edu/pagre/ 

Journal of the Learning Sciences 12(3), 2003. 



1 Hierarchy and general systems 

Herb Simon came to artificial intelligence from organizational studies in New Deal-era public administration, and only now, it seems, after Simon's sad passing in early 2001, are we in position to place this development in historical context. His dissertation, published in 1947 after wartime delays as Administrative Behavior, remains a landmark in the empirical study of organizations. Departing from the prescriptive tradition that preceded it, Administrative Behavior looked beneath the official representations to discover the complementarity between human cognition and institutional context. Although he was a liberal Democrat and active in political causes, Simon did not frame his research on public organizations as a political project [1]. Yet in bringing empirical evidence to the study of organizational rationality, he was contending with the central political questions of his time. 

It was his successive attempts to formalize the limitedly rational processes of organizational decision-making that led Simon, together with Allen Newell and other colleagues, to found the field of artificial intelligence in the 1960's [2]. The interaction between individual cognition and organizational structure that had been central to the argument of Administrative Behavior fell away in Newell and Simon's attempts to build computational models of individual problem-solving. But themes of structure remained central to the project in another way, in the day-to-day technical work of model-building [3]. The early problem-solving models were extraordinary innovations in the context of their times, since they required the modelers to invent much of what we now take for granted as computer science. 

In an epochal 1962 paper entitled "The Architecture of Complexity", Simon attempted to draw together what he had learned about the structure of complex systems [4]. ("The Architecture of Complexity" became widely influential when it was published with three 1968 lectures under the title of "The Sciences of the Artificial" (Simon 1969). I will cite that version. I will not discuss the expanded second edition of "The Sciences of the Artificial" published in 1981, or the third edition published in 1996.) I will argue that we can look at "The Architecture of Complexity" now with forty years of hindsight as a study in the historical conditions under which intellectual problems become visible. Much as the early work of the cognitivist movement had been framed as a response to the hegemony of behaviorism, "The Architecture of Complexity" is framed as a response to general systems theory. It was a complex response that operated on two levels. 

On a substantive level, general systems theory promised to explain the vast and rather nebulous phenomenon of "complexity" in terms of the order that "emerges" from a dynamical system of interacting elements. Simon likewise proposed to explain the phenomena of emergent complexity, but despite his friendly tone the unifying theme of his explanation -- hierarchical structure -- conflicted with the spirit of general systems theory. The theme of hierarchy is capable of being compatible with general systems theory, for example in theories that describe hierarchies of dynamic systems, each interacting with those above and below it in the hierarchy. Indeed, Simon had been influenced by such models; he gave particular credit to the work of Richard Goodwin (1947; see Simon 1991: 376-378). But for the Simon of "The Architecture of Complexity", hierarchy meant a static hierarchical structure. He was most interested in this structure's genesis (as we will see), but not in the structure's moment-to-moment reproduction [5]. This conception of hierarchical structure was obviously already a familiar theme in the study of organizations, but even more important for the emergence of artificial intelligence was the influence of Chomsky's (1957) theory of the hierarchical structures of natural language syntax. What is more, themes of hierarchy arose naturally from Newell and Simon's practical experience with computational system-building. Simon had every reason to regard hierarchy as an organizing theme for a science of complexity. 

This, perhaps, is why Simon found general systems theory so congenial on the level of methodology. General systems theory was "general" in the sense that it promised to explain phenomena in an extremely wide range of fields. To this end, it developed a vocabulary that can fairly be called "vague". What, after all, is a "system"? The term can be given any number of mathematical definitions, but in its reference to the world of empirical phenomena its great value was precisely its ability to direct attention in an organized way to the phenomena in any field. Once a "system" has been discerned within the materials of a particular discipline, one can set about paraphrasing them in terms of an idealized mathematical theory, most especially that of cybernetics (see Bowker 1993). This kind of generality need not be a bad thing; many heuristic devices are valuable despite their lack of analytical precision. Nonetheless, the great danger of this procedure is that, having discovered "systems" everywhere, vagueness can be mistaken for universality. The illusion of universality can then lead researchers to overgeneralize, and to foreclose intellectual alternatives. 

Just as the general systems theorists were impressed by their ability to discern "systems" wherever they looked, Simon had been struck by the recurrence within his own work of themes of hierarchy. Having then set about looking for hierarchy in the world, he found it everywhere. And so, in "The Architecture of Complexity", he proposed it as a universal principle of the structure of complex things. The conflict between general systems theory and "The Architecture of Complexity" is a conflict between explanatory schemas -- self-organization and hierarchy -- that are intrinsically vague, but precisely for that reason can seem universal. Now, it is entirely possible that one of these competing schemata really is the more powerful, the more heuristically valuable, the more widespread in the real empirical world. But because Simon's text deemphasizes the conflict between his own theory and the general systems theorists', the issue is not fully joined. By juxtaposing the two schemata here, I want to forestall judgements of universality and ask what is really at stake in choosing one rather than the other. 

Simon provides many examples of hierarchy [6]. He mentions organizations (1969: 87), of course, while emphasizing that hierarchy need not imply top-down relations of authority (8Cool. The basic structure of matter is hierarchical, in that molecules are made of atoms, which are themselves made of putatively elementary particles (87). Astronomers can treat stars or galaxies as elementary units (87). Books are divided into chapters, and then into sections, paragraphs, clauses, phrases, and words (90). The case of music is similar (90). The segmentary structures of societies -- individuals within families within tribes within nations -- are likewise hierarchical (8Cool. And problems can be solved more easily when they can be decomposed into subproblems whose solutions can be combined into a solution to the problem as a whole (95-96). 

Even though his overall topic was the "sciences of the artificial", Simon emphasized above all that hierarchy was something profoundly natural. He viewed hierarchy as a general principle of complex structures -- and not just of particular complex structures but of complexity in general. Hierarchy, he argued, emerges almost inevitably through a wide variety of evolutionary processes, for the simple reason that hierarchical structures are stable. To motivate this deep idea, he offered (1969: 90-92) his most important example of hierarchy, a "parable" about imaginary watchmakers named Hora and Tempus. According to this story, both watchmakers were equally skilled, but only one of them, Hora, prospered. The difference between them lay in the design of their watches. Each design involved 1000 elementary components, but the similarity ended there. Tempus' watches were not hierarchical; they were assembled one component at a time. Hora's watches, by contrast, were organized into hierarchical subassemblies whose "span" was ten [7]: he would combine ten elementary components into small subassemblies, and then he would combine ten subassemblies into larger subassemblies, and these in turn could be combined to make a complete watch. 

The difference between the two watchmakers' designs became crucial in the larger context of their lives as artisans. Customers would call them on the phone, and when the phone rang they would be compelled to abandon their current assembly, which would fall apart. These interruptions did not bother Hora, who lost at most ten units of work: whatever subassembly he happened to be working on. Tempus, however, would lose (on average) much more: an entire partly-assembled watch would fall apart, and not into modular subassemblies but back into its elementary components. If interruptions are at all common, it can easily be shown that Hora will complete many more watches than Tempus, whose business will surely fail to prosper. 

The watchmakers' story deserves close attention. It is striking, first of all, that Simon chooses to illustrate a story about evolution using an example that involves designers -- and watchmakers, no less [8]. The whole point of biological evolution is that it does not require a designer, and the watchmaker was a common post-Newtonian metaphor for God in a mechanistic universe [9]. Of course, Simon means to use the word "evolution" in a generalized way; he is not referring solely to Darwin's theory. Nonetheless, it is worth asking how far the story does generalize. His point, first of all, cannot be that naturally occurring complexity comes about entirely by assembling previously existing modular components. Some modularity is certainly found in complex organisms, for example in the cell as a modular component of plants and animals (Simon 1969: 8Cool, but very little biological structure above the cellular level is assembled from previously independent modular components [10]. The hierarchical structure of a book consists of preexisting components on some levels (letters, words, some of the phrases), but not on others (the great majority of sentences and paragraphs, and except for anthologies all of the sections and chapters) [11]. Nor did the modern corporation acquire most of its hierarchical structure by assembling existing components; rather, its leaders discovered through trial and error the need to impose such structures upon it (Chandler 1962). Modular industrial products are sometimes designed as complete architectures, and sometimes designed with standardized interfaces that invite others to introduce complementary products, but rarely are free-standing, independently designed products assembled post hoc into modular systems [12]. Simon (1969: 93) does make clear that "in spite of the overtones of the watchmaker parable, the theory assumes no teleological mechanism". "Direction", he asserts, is provided by "nothing more than survival of the fittest -- that is, of the stable" (ibid). But the equation between fitness and stability is forced: fitness in real ecosystems is more than a matter of not falling to pieces. The figure of the watchmaker functions (no doubt unintentionally) to make the equation between fitness and stability more plausible than it really is. 

The role of the designer in the watchmakers' story is peculiar for another reason. In introducing his engagement with general systems theory, Simon remarks that: 


In [complex] systems, the whole is more than the sum of its parts, not in an ultimate, metaphysical sense, but in the important pragmatic sense that, given the properties of the parts and the laws of their interaction, it is not a trivial matter to infer the properties of the whole. In the face of complexity, an in-principle reductionist may be at the same time a pragmatic holist (1969: 6Cool
He is saying two things here. Less problematically (though controversially at the time), he is affirming that hierarchy allows a theorist to choose a level of explanation, for example treating atoms as if they really were atomic. But beyond that, he also embraces the systems theorists' assertion that the properties of wholes cannot easily be predicted from the properties of their parts. Yet this is nearly the opposite of the way that engineers such as Hora -- or Simon himself -- use hierarchy in practice. For working engineers, hierarchy is not mainly a guarantee that subassemblies will remain intact when the phone rings. Rather, hierarchy simplifies the process of design cognitively by allowing the functional role of subassemblies to be articulated in a meaningful way in terms of their contribution to the function of the whole. Hierarchy allows subassemblies to be modified somewhat independently of one another, and it enables them to be assembled into new and potentially unexpected configurations when the need arises. A system whose overall functioning cannot be predicted from the functionality of its components is not generally considered to be well-engineered. 
Other anomalies arise when the watchmakers' example is compared to the example of organizations. Both are hierarchical. But a watch, like a molecule, is hierarchical in a particular way: assemblies physically contain subassemblies, and in this way the hierarchy is arranged in space. Elementary components that are far apart in terms of the graph structure of parts and subparts are typically far apart in spatial terms as well; in any event they are likely to be far apart in causal terms [13]. Simon captures this intuition with his concept of "nearly-decomposable systems" (1969: 99-103). Yet his argument is unstable here. He is not simply choosing to study certain complex structures that happen to be nearly decomposable; rather, he is arguing that the structures that arise in the world tend to be hierarchical (and thus nearly decomposable). This is precisely what makes the argument profound: it is not simply a prescription for engineers but a sweeping empirical claim that the universe in general is organized as if it had been engineered. But in order for the story about hierarchy-through-evolution to apply to human organizations, hierarchy must be highly adaptive for those organizations, and not just a transient accommodation to technological limits. Chandler's (1962) history makes clear that functionally differentiated organizational hierarchy in the corporation arose in large part from the need for clear lines of authority. But little follows from this: clear lines of authority might be necessary precisely because of the limitations of organizational technologies. Organizational hierarchy is not well-reputed nowadays; in any event, the case for hierarchy is not closed [14]. 

What can be learned from these puzzles? The lesson is surely not that hierarchy is a useless idea, or that hierarchies are not prevalent in many parts of the natural and social worlds. The point, rather, is that hierarchy is a somewhat more diverse phenomenon than the universal ambitions of Simon's theory require. It is understandable that Simon should want to idealize away some of this diversity; after all, he had genuinely made important discoveries about the role of hierarchy in the operation of complex organizations and the design of complex artefacts. Nonetheless, to place Simon's discoveries in context, it is important that we back away from some of his idealizations. 

2 Hierarchy in history 

How, then, can Simon's arguments be understood in their historical context? I believe that Simon's theory was very much a product of its time. The year that he wrote "The Architecture of Complexity" (1962) might be seen in retrospect as the high-water mark of the classical hierarchical organization. The late 19th century had established the transportation and communication infrastructures that made possible the economies of scale that paid for engineers and other professional information workers (Chandler 1977). Engineers had then invented the techniques of managerial rationalization, whereupon the "human" side of management had been discovered as an object of empirical study by Simon himself and codified prescriptively by Barnard (1962 [1938]) and Drucker (1945). The Cold War, finally, had given the hierarchical organization a new cultural role, as well as providing the research funds to develop scientific and computational theories of it. 

At the same time, many developments that we now take for granted were largely unimagined when Simon wrote. Computer networking hardly existed. The collapse of communism was literally unimaginable, and so was the collapse in cultural respect for hierarchy that accompanied it. In 1962, organizations could still be conceptualized as stably self-contained organisms whose components did diverse things. Managers still served largely as conduits for information, managerial accounting was poorly developed, outsourcing was not practical in most areas, business cycles were severe, and large inventories compensated for poor communications along supply chains. 

Simon identified some of the problems and articulated the themes that could become visible in this historical context. In terms of the conflict of explanatory schemas that I identified earlier -- self-organization versus hierarchy -- Simon caught sight of the schema that was ascendant in the time and place where he worked. Research on emergent complexity in dynamic systems was certainly current in the 1960's, but it did not have the mathematical or computational resources to generate a research program of the same breadth. Just as Minsky and Papert's Perceptrons (1969) had the effect (intentionally or not) of aborting the research program of emergent complexity in neural networks, "The Architecture of Complexity" effectively captured the theme of complexity and bound it to a research program centered on conceptions of hierarchical structure. The 1960s made hierarchy easy to think, and the central role of hierarchy in the era's thinking about computation was part of that larger trend. 

Simon's use of the watchmaker metaphor gives the question a decided political edge. Not only did clockwork metaphors figure in theological disputes, but (according to Mayr (1986)) the contrast between clocks and servomechanisms symbolized for early modern Europe the contrast between Continental themes of authority and English themes of liberty. Technology has long been part and parcel of political culture. The point is not that Simon and his allies were programmatically promoting hierarchical structures of authority in society. Indeed, many of them, especially the anarchist Chomsky, would certainly have rejected the thought. The point, rather, is that the patterns that Simon discerned became visible within the larger context of the times. If we wish to retreat from Simon's commitment to themes of hierarchy, the remedy is not an equally thoroughgoing commitment to the theme of self-organizing dynamic systems, much less the idealized market model with which dynamic systems are now culturally associated. And even though Simon opposed neoclassical economics with its imputations of unbounded rationality to "economic man", going so far late in life as to endorse organizations over markets as explanatory units for economics (Simon 1997), it would be anachronistic to ascribe to him an ideological project of making state control of the economy into an ontological principle of the universe. The ideological equation between hierarchy and government, and between dynamic systems and markets, was not well-developed in the era of "The Architecture of Complexity", and the idea of a world-historical conflict between hierarchies and markets (as, for example, in Fukuyama (1999)) would have made little sense in the context of the large corporation and the Cold War. Hierarchy in those days meant not only the regulatory and welfare state but the church and the military, as well as the corporate firm. Nor is the opposite extreme of emergent self-organization ex nihilo any more viable intellectually; indeed, the whole point of the most intellectually serious proponents of self-organization, such as Hayek (1948), is that the libertarian economy requires a robust institutional framework before the miracles of self-organized complexity can reliably occur [15]. 

To understand the place of the hierarchy and self-organization schemata in historical context, then, it helps to see them not only as warring enemies but as loosely articulated constituents of a single formation -- as two sides of a very old coin. Cultural fashion swings back and forth between the two sides of the coin historically, both because of changes in reality (such as the development of the hierarchical capitalist firm) and because of changes in ideology (such as with the fall of communism). But if we back away from questions of fashion and focus, we can see in historical context the profound complementarity of the two schemata, as well as the genuine complexity that each schema attempts to paper over. 

Simon was hardly the first scholar to notice hierarchy in all things; as Lovejoy (1936) long ago observed, hierarchy was a pervasive intellectual and cultural theme of medieval Europe. Panofsky (1957) demonstrated the influence of hierarchy in medieval architecture. Studies such as these have described the medieval mind as living in a static universe. Yet, as more recent scholarship has demonstrated, late-medieval experiences of the market led to counterphilosophies of dynamism. Kaye (1998), for example, argues that the philosophers of 14th century Oxford and Paris were immersed in the emerging market societies of that era -- in their roles as university administrators they effectively ran large businesses -- and that this experience was reflected in a new intellectual picture of the universe as a dynamic equilibrium. The connection between economics and cosmology was never made explicit in their writing, but they were certainly familiar with such "modern" ideas as the evils of government intervention in the dynamic equilibration processes of the market. What is more, they applied their incipient mathematical analysis of dynamic equilibrium systematically to every question they could think of, including theological questions that no later scholar would imagine treating in that way. 

For Kaye, this significance of this discovery is that the decline of the static medieval worldview must be dated at least two centuries earlier than earlier scholars had assumed. Yet this picture of decline does not match the historical facts. Writing in the late 18th century, Edmund Burke founded modern conservatism by convening an ideological coalition between the aristocracy and church -- traditional social elites who wished to conserve a static social order of deference and authority -- and the emergent merchant class -- who wished to encourage a dynamic social order of commerce. Burke's project was not universally embraced in its day -- far from it. But it has not disappeared, and it is currently the ascendent political movement in the United States. Like any coalition, the conservative coalition is not entirely stable. The interface between the dynamic and static components of its ideology must be constantly reworked and constantly smoothed over. Some of Burke's followers emphasize his themes of liberty; others his themes of order; and sometimes the themes are combined in ways that downplay their intrinsic tension. In any event, the point is that the explanatory schemata of hierarchy and self-organization are not always at war. Their relationship is complex and variable. Both schemata are woven throughout Western culture, and both are capable of coming to the surface in a wide variety of ways when conditions are right. 

As an example, consider Coase's (1937) influential paper about "the nature of the firm". Coase asks a striking question: if markets are the most efficient way to organize economic life, and if competition selects for efficiency, why do capitalist firms exist at all? Why is the entire economy not organized according to Adam Smith's idealized picture of individual artisans? The answer, Coase argues, is that market mechanisms are not costless, and that organizations arise when productive activities can be organized more cheaply by a hierarchical organization than by market transactions. Coase's argument had little impact at first, but beginning in the 1970s, a movement arose of conservative legal scholars who viewed social progress in almost teleological terms as the progressive reduction of transaction costs, and thus (they argued) the ever more perfect approximation of ideal market conditions (Posner 1972). 

But this conclusion does not follow from Coase's argument. The whole burden of Coase's paper, after all, was not to argue that transaction costs tend to disappear, but that they exist at all, and that they have structural consequences for the economy. Technologies such as the telegraph or the Internet that reduce transaction costs do not, by Coase's argument, necessarily reduce the size of organizations. Although Coase does not use these terms, he argues that the balance between organizations and markets its determined by the relationship between transaction costs (the costs of coordinating activity in the market) and organizing costs (the costs of coordinating activity in the firm). New technologies typically reduce both types of costs, and North and Wallis (1994: 613) observe that "the firm is not concerned with minimizing either transaction or transformation costs in isolation: the firm wants to minimize the total costs of producing and selling a given level of output with a given set of characteristics". 

My suggestion, then, is that phenomena of hierarchy and self-organization are not mutually exclusive, and that neither one is necessarily destined to win a world-historical battle against the other. Although they are analytically distinct and should not be conflated, they nonetheless coexist, in both ideology and in reality, and they are likely to continue coexisting in the future. From this perspective, the models of Simon and the general systems theorists -- all hierarchy or all self-organization -- are models of simplicity, not of complexity. Real complexity begins with the shifting relations between the two sides. 

Take, for example, the impact of computer networking on corporate organization since perhaps 1990. Phenomena such as delayering (the cutting-out of excessive layers of hierarchical organization) and outsourcing (buying from others every product and service that the organization has not specialized in making) have been attributed in part to the greatly enhanced capacities for relationship-at-a-distance that computer networking affords. Precisely because Simon's image of hierarchy is spatial, it does not fit well with the networked world, which collapses many types of distance (e.g., Frissen 1997). Organizations are not in fact nearly-decomposable: there is much to be gained by integrating the activities in the different functional divisions. The loss of clear lines of authority compels the organization's members to negotiate conflicting loyalties (Hirschhorn and Gilmore 1992), but pervasive computer networking greatly strengthens the tools of managerial control and long-distance coordination. These phenomena deconstruct both models -- hierarchy and self-organization alike. Hierarchy and self-organization both have their place in the networked world, but it is a different place, and a complicated one. 

3 Hierarchy and democracy 

What does this have to do with cognitive modeling? Let us recall the great insight of Simon's dissertation: the complex fit between individual cognition and its institutional environment. Simon wrestled with the foremost theme of his century, the place of rationalism in the complex relationship between the organizing principles of hierarchy and self-organization. The self-organization model teaches the value of a groundwork of rules to facilitate self-ordering, and the hierarchy model the value of structures that simplify cognition and life in general. Bringing these insights together, we see the need for a theory of cognition as a phenomenon of social life, embedded in institutions and relationships. Research in the tradition that Simon founded has returned to the computational study of organizations, and thus of individuals in organizational context (e.g., Carley and Prietula 1994). The permanent tensions between hierarchy and self-organization will surely emerge in this work in new and productive forms. 

Yet there is life beyond the collision between hierarchies and markets, and it lies in the values of democracy that Simon believed in. By initiating the empirical study of organizations, Simon discovered the living, breathing human actors who lie at their center. He saw something profound, that human agency and relationship are the ground of all social order, no matter how imputedly rational that order might be, and that human agency and relationship are reciprocally dependent on social order as well. Democracy requires hierarchies and markets, but more fundamentally it requires active citizens who can take up a vantage point outside those institutions, critically evaluating them and cooperatively participating in the processes of social choice that shape them. By retracing Simon's steps, we can learn something more about where this vantage point is located, and how to live in it. I did not always agree with Simon [16], but like all of us I am privileged to have learned about these things from him. 

Acknowledgments 

I appreciate the assistance of Mie Augier and Jean-Louis Le Moigne. 

Endnotes 

[1] In this, Simon contrasts with the more recent work on democratic governance of his once-collaborator James March. See March and Olsen (1995). Simon discussed his political activities, and his brush with Cold War blacklisting, in his autobiography (1991: 117-134). 

[2] See Newell, Shaw, and Simon (1960); Newell and Simon (1963); Newell and Simon (1972). 

[3] Edwards (1996: 250) rightly emphasizes this point. 

[4] The argument of "The Architecture of Complexity" also appeared in a more compressed form in Simon (1960: 40-43). Administrative Behavior includes a concept of "hierarchy of authority", but the concept of hierarchical structure is not central to the argument. In a late paper with Mie Augier meant to serve as an introduction to "The Architecture of Complexity", Simon extends his arguments in the light of more recent computational models of evolution (Augier and Simon in press). I am indebted to Mie Augier for sending me a prepublication copy of this paper. 

[5] For example, it is useful to contrast Simon's conception of structure with the autopoietic theory of Maturana and Varela (1980). Elsewhere in his 1968 lectures, Simon (1969: 24-25; cf. Agre 1997: 56) presents his famous parable of the ant whose complex path along a beach is determined jointly by the ant and the beach. Many authors have presented this as evidence of the interactionist nature of Simon's theory. Yet Simon announces a page later that he is only interested in cognition and not in embodied activity. The point of the ant story is that a general problem-solving mechanisms enters into unpredictable interactions with an abstract problem space -- a formal construct with no necessary relationship to the physical or social environment. On the other hand, it should be noted that Simon himself was certainly familiar with servomechanism models and had published a number of papers that employed them; see, for example, Simon (1952). In his autobiography (1991: 108-109) he credits his use of control theory to the influence of his father, an engineer. 

[6] He also acknowledges (1969: 94) that not all structures are hierarchical, giving the example of polymers, which are long chains of identical units. "However", he says, "for present purposes we can simply regard such a structure as a hierarchy with a span of one -- the limiting case". 

[7] The "span" of a hierarchy is the number of subordinates beneath each element. 

[8] Dawkins (1986) later played with this image when he referred to evolution as "the blind watchmaker". For the subsequent history of evolutionary metaphors in economic studies of organization see Nelson and Winter (1982) and Hodgson (1993, 1999). On evolutionary metaphors in the development of technology see Basalla (1988). 

[9] On the long Western tradition of likening the engineer to God, see Noble (1997). On the clockwork metaphor see McReynolds (1980). 

[10] To be sure, Margulis (1998) has subsequently argued that the cell's internal structure arose through the symbiotic combination of formerly independent organisms. In this case Simon's theory does apply. 

[11] Poerksen (1995) argues that bureaucratically rationalized language is increasingly composed of modular elements that lack the embedding in community and history that normally gives language its fullness of meaning. Yet Blair (1988) argues that an affinity for modularity is a central feature of American culture. 

[12] On the role of modularity in industrial design and strategy, see Baldwin and Clark (2000), Economides and Salop (1992), Langlois and Robertson (1992), Meyer and Lehnerd (1997), and Schilling (2000). 

[13] Simon (1969: 89-90) makes clear that hierarchies are defined in terms of interaction patterns, not spatial proximity. The point here is that the metaphor of hierarchy as a physical structure is spatial. 

[14] For numerous additional criticisms of the association between evolution and efficiency, and in particular the inference that hierarchical firms are efficient given their mere existence, see Hodgson (1991). 

[15] On the subsequent history of metaphors of self-organization in the social sciences, see Contractor (1999). 

[16] See Vera and Simon (1993) and Agre (1993; 1995; 1997: 54-57, 142-143). 

References 

Agre, P. E. (1993). The symbolic worldview: Reply to Vera and Simon. Cognitive Science 17, 61-69. 

Agre, P. E. (1995). The soul gained and lost: Artificial intelligence as a philosophical project. Stanford Humanities Review, 4, 1-19. 

Agre, P. E. (1997). Computation and Human Experience. Cambridge: Cambridge University Press. 

Augier, M. & Simon, H. A. (in press). The architecture of complexity and the role of organizations. In R. Garud, A. Kumaraswarmy, & R. Langlois (Eds.), Managing the Modular Age: Architectures, Networks, and Organizations. Oxford: Blackwell. 

Frissen, P. (1997). The virtual state: Postmodernisation, informatisation and public administration. In B. D. Loader (Ed.), The Governance of Cyberspace: Politics, Technology and Global Restructuring (pp. 111-125). London: Routledge. 

Baldwin, C. Y. & Clark, K. B. (2000). Design Rules, Volume 1: The Power of Modularity. Cambridge: MIT Press. 

Barnard, C. I. (1962). The Functions of the Executive. Cambridge, MA: Harvard University Press. Originally published in 1938. 

Basalla, G. (1988). The Evolution of Technology. Cambridge: Cambridge University Press. 

Blair, J. G. (1988) Modular America: Cross-Cultural Perspectives on the Emergence of an American Way. New York: Greenwood Press, 1988. 

Bowker, G. C. (1993). How to be universal: Some cybernetic strategies, 1943-70. Social Studies of Science, 23, 107-127. 

Carley, K. M., & Prietula, M. J. (Eds.) (1994). Computational Organization Theory. Hillsdale, NJ: Erlbaum. 

Chandler, Jr., A. D. (1962). Strategy and Structure: Chapters in the History of the Industrial Enterprise. Cambridge, MA: MIT Press. 

Chandler, Jr. A. D. (1977). The Visible Hand: The Managerial Revolution in American Business. Cambridge, MA: Harvard University Press. 

Chomsky, N. (1957). Syntactic Structures. 's-Gravenhage, The Netherlands: Mouton. 

Coase, R. H. (1937). The nature of the firm. Economica NS, 4, 385-405. 

Contractor, N. S. (1999). Self-organizing systems research in the social sciences: Reconciling the metaphors and the models. Management Communication Quarterly, 13, 154-166. 

Dawkins, R. (1986). The Blind Watchmaker. New York: Norton. 

Drucker, P. (1945). Management: Tasks, Responsibilities, Practices. New York: Harper and Row. 

Economides, N. & Salop, S. C. (1992). Competition and integration among complements, and network market structure, Journal of Industrial Economics, 40, 105-123. 

Edwards, P. (1996). The Closed World: Computers and the Politics of Discourse. Cambridge, MA: MIT Press. 

Frissen, P. (1997). The virtual state: Postmodernisation, informatisation and public administration. In B. D. Loader (Ed.), The Governance of Cyberspace: Politics, Technology and Global Restructuring (pp. 111-125). London: Routledge. 

Fukuyama, F. (1999). Death of the hierarchy. Financial Times, 12 June, I. 

Goodwin, R. (1947). Dynamical coupling with especial reference to markets having production lags. Econometrica, 15, 181-204. 

Kaye, J. (1998). Economy and Nature in the Fourteenth Century: Money, Market Exchange, and the Emergence of Scientific Thought. New York: Cambridge University Press. 

Hayek, F. A. (1948). Individualism and Economic Order. Chicago: University of Chicago Press. 

Hirschhorn, L., & Gilmore, T. (1992). The new boundaries of the "boundaryless" company. Harvard Business Review, 70, 104-115. 

Hodgson, G. M. (1991). Economic evolution: Intervention contra Pangloss. Journal of Economic Issues, 25, 519-533. Reprinted in Hodgson (1993). 

Hodgson, G. M. (1993). Economics and Evolution: Bringing Life Back into Economics. Cambridge: Polity Press. 

Hodgson, G. M. (1999). Evolution and Institutions: On Evolutionary Economics and the Evolution of Economics. Cheltenham, UK: Elgar. 

Langlois, R. N. & Robertson, P. L. (1992) Networks and innovation in a modular system: Lessons from the microcomputer and stereo component industries, Research Policy, 21, 297-313. 

Lovejoy, A. O. (1936). The Great Chain of Being: A Study of the History of an Idea. Cambridge, MA: Harvard University Press. 

March, J. G., & Olsen, J. P. (1995). Democratic Governance. New York: Free Press. 

March, J. G., & Simon, H. A. (1958). Organizations. New York: Wiley. 

Margulis, L. (1998). Symbiotic Planet: A New Look at Evolution. New York: Basic Books. 

Maturana, H. R., & Varela, F. J. (1980). Autopoiesis and Cognition: The Realization of the Living. Dordrecht: Reidel. 

Mayr, O. (1986) Authority, Liberty, and Automatic Machinery in Early Modern Europe. Baltimore: Johns Hopkins University Press. 

McReynolds, P. (1980). The clock metaphor in the history of psychology. In T. Nickles (Ed.), Scientific Discovery: Case Studies (pp. 97-112). Dordrecht: Reidel. 

Meyer M. H. & Lehnerd, A. P. (1997) The Power of Product Platforms: Building Value and Cost Leadership. New York: Free Press. 

Minsky, M., & Papert, S. (1969). Perceptrons: An Introduction to Computational Geometry. Cambridge, MA: MIT Press. 

Nelson, R. R., & Winter, S. G. (1982). An Evolutionary Theory of Economic Change. Cambridge, MA: Harvard University Press. 

Newell, A., Shaw, J. C., & Simon, H. A. (1960). Report on a general problem-solving program. In Proceedings of the International Conference on Information Processing, Paris, 256-264. 

Newell, A., & Simon, H. A. (1963). GPS, a program that simulates human thought. In E. A. Feigenbaum & J. Feldman (Eds.), Computers and Thought (pp. 279-293). New York: McGraw-Hill. Originally published in 1961. 

Newell, A., & Simon, H. A. (1972). Human Problem Solving. Englewood Cliffs, NJ: Prentice-Hall. 

Noble, D. F. (1997). The Religion of Technology: The Divinity of Man and the Spirit of Invention. New York: Knopf. 

North, D. C., & Wallis, J. J. (1994). Integrating institutional change and technical change in economic history: A transaction cost approach. Journal of Institutional and Theoretical Economics, 150, 609-624. 

Panofsky, E. (1957). Gothic Architecture and Scholasticism. New York: Meridian Books. 

Poerksen, U. (1995) Plastic Words: The Tyranny of a Modular Language, translated by Jutta Mason and David Cayley. University Park: Pennsylvania State University Press. 

Posner, R. A. (1973). Economic Analysis of Law. Boston: Little, Brown. 

Schilling, M. A. (2000) Toward a general modular systems theory and its application to interfirm product modularity, Academy of Management Review, 25, 312-334. 

Simon, H. A. (1947). Administrative Behavior: A Study of Decision-Making Processes in Administrative Organization. New York: Macmillan. 

Simon, H. A. (1952). On the application of servomechanism theory in the study of production control. Econometrica, 20, 247-268. 

Simon, H. A. (1960). The New Science of Management Decision. New York: Harper. 

Simon, H. A. (1969). The architecture of complexity. In The Sciences of the Artificial (pp. 192-229). Cambridge, MA: MIT Press. 

Simon, H. A. (1991). Models of My Life. New York: Basic Books. 

Simon, H. A. (1997). An Empirically Based Microeconomics. Cambridge: Cambridge University Press. 

Vera, A. H., & Simon, H. A. (1993). Situated action: A symbolic interpretation. Cognitive Science, 17, 7-48.
---