The Wizards of OS 
"Information wants to be Free"

The Market Logic of Information

Philip E. Agre 
Department of Information Studies 
University of California, Los Angeles

There can be no doubting that advanced information technologies will bring large-scale social change. The technology is advancing at an extraordinary rate in quantitative terms, and it will probably continue to improve by a factor of 100 each decade. Sheer computing power and communications bandwidth will be abundant. The hard question is what direction the consequent social changes will take. The technology is not only increasingly powerful but highly malleable. Its qualitative architecture is capable of evolving in many different directions, depending on who can assemble the resources to build it and secure the cooperation of others to use it. And that is a political and economic question, as much as it is technical. Social institutions will shape the technology, even as the technology provides raw material for the reshaping of institutions. 

Every new technology creates a vacuum of imagination, and the imagination of the West takes definite forms. One of these forms of imagination begins with a dualism between the corrupt world of the flesh and the purity of the spirit. In the Internet world this form of imagination gives us cyberspace, the technological equivalent of heaven, the purely abstract space-apart that is constructed from the mathematical ideals of computing. The virtual reality fad of the early 1990s was one version of cyberspace, and another is the concept of an online community. Of course, Western imagination is not all wrong, and there do exist virtual reality technologies and online communities. Nonetheless, experience is making clear that the idea of cyberspace is misleading, and that the uses of advanced information technologies do not constitute a space apart from ordinary reality. Quite the contrary, for most purposes advanced information technologies are deeply bound up with the rest of the world. Virtual reality is simply one end of a spectrum of applications, each of which embeds networked hardware and software into the world in a different way. It is the great philosophical virtue of ubiquitous computing technologies and spontaneous wireless networking protocols like Bluetooth to make this clear, and to help tear down the ancient mind/body divisions that have long narrowed our conceptions of computing and of ourselves. 

Notions of online community are equally misleading. People do form social bonds through new digital media. But communities are analytically prior to the particular technologies that their members use. The worldwide community of stamp collectors, for example, existed long before the Internet, and it conducts its collective life through many media besides the Internet. Its members even meet face-to-face. The same goes for the thousands of other communities of shared interest -- professions, associations, extended families, political parties, and so on. Use of the Internet might change the dynamics of these communities, and many striking stories of community change have been recorded. But Internet use is embedded in something larger. Even communities that form online often develop other means of interaction, for example by holding caucuses at conferences. In each case, the notion of cyberspace limits our vision by directing our attention to a small corner of a large phenomenon: the interaction and coevolution between new technologies and the institutional orders and ways of life in which they are embedded. 

Although the imaginative structures of cyberspace descend from a long philosophical tradition, their recent history is explained just as much by political economy. Simply put, cyberspace is a lobbying campaign. The notion that the computer industry seeks to avoid regulations and other government involvement is scarcely credible, given the large number of laws that it has successfully supported throughout the world. Government interference through draconian intellectual property laws is an example. Nonetheless, many other lobbying campaigns do draw on the imagery of cyberspace as a new and separate world, completely different in its workings from the old world of smokestacks and railroads, a kind of fragile spaceship that democracy could only harm. 

Yet at the same time, cyberspace has been imagined as a force that will bring about the perfection of the market that government is said to obstruct. By removing friction from the market mechanism, it is said, networking information technology will bring about something resembling the utopia of Adam Smith: an efficient market consisting of large numbers of producers and large numbers of consumers, none of whom possesses the kinds of economic or political power that we associate with the industrial era. This second form of imagination, it should be noted, is wholly incompatible with the first. Information is no longer expected to stand apart from the material world, but quite the contrary is expected to descend upon the world and reorder it according to another vision of heaven, this one derived from the dynamic, self-ordering equilibria of 19th century physics (Mirowski 1989). 

Despite the surface contradiction, the underlying pattern is clear enough: information, it is said, will impose its own order on our lives, which will henceforth participate in its transcendental purity. I propose to take seriously this idea of an order of information. If we are to take seriously the interaction of information technology and market economics, what order will the Internet be imposing on the world? It is a partial and loaded question, of course. It is altogether unclear that the world's people wish to have any such order imposed on them, and it is equally unclear that such an order could actually be imposed in practice. Nonetheless, technology and ideology alike are exercises in applied imagination, and precisely because imagination is a material force in the world it will be a useful exercise to recover the order of information that emerges from information technology and markets as they are practiced. 

As a window onto the way that business imagines the Web, let us consider the projects reported in a 9/18/00 Business Week supplement on 50 companies that the magazine's editors regard as "Web smart". All of these companies are using Web-based technologies to improve their operations. Some companies are using the Web to exchange order information with customers and suppliers. Others are using this same information to predict shortages or other market conditions, or to accelerate product cycles. A university is tracking its clinical trials. One group of manufacturers is sharing unused space their trucks; another manufacturer is gathering data from all the machines in its factory and making that data available to all of the machines' operators, for example so that bad product batches can be detected more quickly. With only a few exceptions, all of these projects can be described as instances of two overlapping visions, which might be summarized as follows: 

(1) Digital shadows (Agre 1994, Clarke 1994). Make the whole object continuously and interactively accessible to everyone who has a relationship with it. 

(2) Centers of calculation (Latour 1987). Make a whole population of objects likewise accessible to everyone. Continuously perform calculations on this population of objects -- scheduling, market-making, mapping, forecasting, cash flow, aggregate outcome measurement, data mining, etc. Make the results of these calculations interactively available to everyone who can use them, and accelerate action cycles accordingly. 

These prescriptions require a little explanation. 

  • "Object" here does not mean object-oriented programming, or physical object, or the pejorative sense of treating someone as an object, although any combination of those things might be present. Rather, it means that a person or thing is treated under some professioanl framing -- treating a person as either a student or a patient, for example -- and that the unit of analysis is the person or think together with digital representations and data about it. 

  • It is the "whole" object in that *all* such representations and data are made available for every important attribute of it. This can mean that digital "roots" have been grown into it -- a truck engine or factory machine, for example -- to record its operating parameters. It can also mean maintaining a history of it, for example the history of changes to an airplane or transactions on an account. 

  • "Accessible" means that this data is available on the Web, wirelessly or wireline, real time, all the time. It means tracking the object physically and/or bureaucratically. 

  • "Interactively" means that everyone with the appropriate authority can change or control the object remotely, either its virtual side or its physical side or both, depending on what makes sense for a given application. When the object is a person, the interaction can mean reaching "through" the virtual side and getting a synchronous connection to the person, for example customer service, a particular person who signed some information in a knowledge base, or a large number of people who might be assembled on a distance basis for a training course. 

  • "Everyone" can include the object's owner (monitoring one's car at a distance), the manufacturer (appliances "phoning home" for preventing maintenance or restock), or an employer (checking in on expenses incurred by drivers on the road). 

It remains to be seen, of course, which of these projects ever gets working, and which ones prove to be worth the trouble. Business Week's own evidence is sometimes equivocal, and its upbeat summaries surely gloss over a world of hidden agendas and messy details. None of the projects, moreover, reaches the full generality of the two models that I sketched. Nonetheless, a clear picture does emerge of the deep structure of a business vision of information technology, one that grows naturally from the project of rational ordering and control that has informed business computing practice from the start. It is a version of what David Gelernter (1991) calls the mirror world: making the whole world digitally available at every point in the world. This is the promise and illusion of Web search engines. 

But the business vision is actually more specific than the mirror world, and more suggestive, in its emphasis on the relationships that structure computing links. A family's relationship to its refrigerator, a manufacturer's relationship to each of the cars it has made, an investment firm's relationship to its clients -- each of these relationships has an institutional and informational architecture that can be inscribed in the architecture of the data links that join the two sides. By complicating the architecture of the data links, one can provide additional functionalities or increase the efficiency of existing ones. 

From the point of view of the individual, or of the artifact or organization, the change is striking. Relationships that may formerly have been episodic, their participants interacting only when jointly present or when talking on the phone, or through the arm's length of paper records that might not be up to date, are now to be continuous, always-on, 24/7. We should not conceive the change as discontinuous, much less as a rewriting of an underlying institutional logic. Nonetheless, the development of a portfolio of always-on relationships -- to people, organizations, and things -- does call us to revisit traditional conceptions of the person. No longer, for example, are particular relationships tied to particular places. Rather, everyone is tied to everyone else all the time. All of our potentially thousands of relationships compete for attention through wireless devices whose interfaces -- not based on the obtrusive ringing of a telephone, one supposes -- are as yet hardly imagined. 

Central to such a world are institutions and technologies that might be called switchboards: the practical means by which people establish, maintain, and evolve relationships. These might be market relationships of whatever structure, longer-term contractual relationships, professional ties, family relations, shared memberships in associations, and many others. Lower-level switchboards might have functionality no more complex than an Internet router or mailer, while more complex switchboards might layer on the functionalities of a marketplace, library, event notification, gaming environments, workflow tools, supply chain integration, geographic information and tracking, or many others. The architecture of these switchboards is obviously a matter of some concern from a legal and moral standpoint, as well as economic and technical. 

Identifying the technical basis for the maintenance of relationships, however, does not tell us the form that the resulting network of relationships will take. For that purpose, it will help to return to the second of the two business visions of cyberspace that I sketched at the opening: the friction-free electronic economy that will deliver us to the idealized market of Adam Smith. Let us leave aside the question of whether such an outcome would be desirable, and instead ask whether will actually happen. The argument for it is clear enough. It is based on Ronald Coase's (1937) epochal paper "The Nature of the Firm", which introduced the concept (though not the phrase) of transaction costs. Coase asked, given that Adam Smith's market of individual artisans and consumers is the economically optimal way to organize the economy, why do firms exist, much less large corporations? What is the economic reason for these often-substantial islands of centralized command and control? The answer, Coase suggests, is that the market mechanism itself entails costs. Buyer and seller must seek one another out, establish the quality of the goods, negotiate, handle money, monitor one another for honest performance of the deal, and perhaps even litigate about it afterwards. All of these activities are costly, and Coase argues that a transaction will be organized within the boundaries of a firm if it can thereby be done more cheaply than it could on the open market. This is likely to happen, for example, when two parties are likely to do a steady stream of business whose nature is hard to specify in advance, so that it is worth creating a stable relationship of employer and employee between them. The parties still incur organizing costs, but these might be less than the transaction costs that they would incur on the market. 

Coase's argument, then, lays the boundaries of firms around those regions of the economic landscape that are more efficiently organized that way. It is crucial that Coase's argument depends on the relative level of organizing costs and transaction costs. Despite the arguments of the cyber enthusiasts, it is not the case that new technologies, by reducing transaction costs, move the economy toward the idealized market. Businesses may even move toward technology-enabled business practices with higher transaction costs if they can make or save more money elsewhere (North and Wallis 1994). And a technology that reduces transaction costs may reduce organizing costs even more. Many authors have observed that Coase's argument is frustratingly indeterminate in practice, given that the relative values of transaction and organizing costs are so hard to measure or predict. Yet the overall direction of contemporary industrial organization provides powerful clues about the true picture. First of all, we are witnessing a historically unprecedented period of industrial concentration. Firms in both North America and Europe are merging at a rate of trillions of dollars per year. On the other hand, firms are also selling, spinning off, or outsourcing their non-core activities at a high rate as well. These developments may seem contradictory, in that the first one makes firms larger while the second makes them smaller. They are united by a single thread, however, and that is the role of economies of scale. Economies of scale are those forces toward industrial concentration that are created by the fixed costs of production. If a firm can sell its goods at a smaller unit cost by making a billion-dollar investment in a factory, and if the market is large enough to distribute that billion-dollar fixed cost while still underselling a competitor whose fixed costs are lower, then the market will become dominated by a smaller number of larger firms. 

Economic theory holds that economies of scale are limited, and that their magnitude determines whether the natural structure of the market is fragmentation, oligopoly, or monopoly. Economies of scale are a longstanding challenge to those who believe that the natural, unregulated state of the market is one of efficiency, because a sufficiently large oligopolist or monopolist can develop market power and extract economically inefficient rents. A rhetorical emphasis on technologies that reduce transaction costs tends to alleviate this embarrassment by presenting a plausible scenario by which the electronic economy can transcend industrial-age concentration. The argument, however, is mistaken. No Coasian analysis of the electronic economy is complete until we assess the impact of distributed information technologies on organizing costs and economies of scale. And in both cases the picture is clear. Simply put, distributed information technologies make it vastly easier to capture economies of scale and coordinate a large firm. 

The particular economies of scale that matter most here are the ones that derive from fixed costs in information. Because information can be duplicated very cheaply, nearly all of the costs associated with it are fixed. And firms experience many fixed costs of information, including software, policies, forms, regulatory filings, public relations and advertising, market research, and product designs. Computer networks can easily distribute these information objects or their inputs and outputs throughout a far-flung global corporation without regard for the number of operating units that have to be served. The key is that the operating units should all do the same thing. Information technology rewards scale, but only to the extent that practices are standardized. 

That is a major reason why companies merge: to capture economies of scale by laying off the information-producing head-office workers of the acquired firm. And that is why companies sell, spin off, or outsource non-core activities: to eliminate activities that do not enjoy economies of scale because they are different from the core activities that do. Transactions are complex beasts, embedded as they are in a matrix of longstanding relationships and practices (Hodgson 1988). It is hard to reduce transaction costs. Organizing costs, however, are regularly reduced by large percentages, simply by capturing economies of scale. And so, quite to the contrary of the "friction-free economy" model, it is altogether plausible that information technology is a primary contributor to the wave of industrial concentration that we read about each day in the business pages. 

This wave of concentration, to be sure, is not all bad. Economies of scale are still economies, and an industry that captures more such economies will pass them along to the consumer so long as the market remains competitive -- that is, until market power emerges. Conventional economists were largely unafraid of market power because they believed in diseconomies of scale: the various factors that limit economies of scale in practice. Economies of scale require standardization, but the world is not standardized. Consumers' tastes differ, as do cultures, legal systems, media, technical standards, and other features of the market environment. As globalization proceeds, however, all of these types of diversity are being reduced. English is spreading as a global language and technical standards are converging on a global basis, in each case because of the benefits of cross-border compatibility. Trade treaties are harmonizing many institutions, and economies of scale are driving a global homogenization of mass media. 

As diseconomies of scale are destroyed, it becomes more and more practical to run a globally integrated firm -- indeed, a global monopoly -- provided, again, that the firm maintains a strong focus, picking one activity and doing all of it for the whole world. The picture that results is a large collection of focused monopolies, each of them taking a precision "slice" through the world economy by means of global computer networks and by the grace of the standardized world that it both depends upon and helps to create (Bryan, Fraser, Oppenheim, and Rall 1999). 

This form of industrial organization can be understood in terms of the market dynamics of knowledge. It becomes economically practical to create, codify, practice, and extend a body of knowledge only once enough venues have been discovered where that knowledge can be profitably applied. Knowledge in this larger sense is not just the contents of books and heads and hands, but a global network of locations where the objects of that knowledge have been discovered, created, counted, measured, and brought into a practical alignment with the organization and its business plan. Everyone and everything in the world is criss-crossed with these knowledge-shaped slices, and every one of them is an "object" in the sense that I described earlier, bound digitally into an always-on network of relationships on a global basis. 

It does not follow from this picture of rampant standardization, of course, that the world is becoming utterly homogeneous, even if one believes in the economic logic of information in its pure form. Economic evidence suggests that nations and regions are specializing to a degree within the world system, evolving the local configuration of public and private institutions that supports a particular model of production within the overall picture that I am describing (Hollingsworth and Boyer 1999). 

More fundamentally, however, computer scientists know that the commonplace association between standards and homogeneity is entirely misleading. The Internet, for example, is a standard that provides a platform for the construction of a great diversity of activities. Networked computer applications, more generally, are organized into layers, each one built on top of others, and each providing a clearly defined interface for others to be built on top of it (Messerschmitt 1999). This layering scheme has engineering advantages for the management of complexity and the division of labor in system design. But it also has economic advantages for the capture of economies of scope: the ability to use a single asset, in this case a computer protocol and hardware and software conforming to it, for a variety of different purposes, each of which helps to pay off its fixed costs. Each protocol then tends to become dominant through the network effects of compatibility (Shapiro and Varian 1998), though the dominant protocol may or may not be owned, either legally or in practice, by a single firm. In this way, the logic of information is modularizing large parts of the world. Not all of it: a firm like Microsoft that does own a particular service layer (in Microsoft's case, personal computer operating systems and certain office productivity applications) can attempt to leverage that monopoly into other areas through nonmodular design. The market dynamics of modularity are complicated (Baldwin and Clark 2000, Schilling 2000). But on the whole, the logic of focused monopolies together with the logic of layered economies of scope are modularizing large parts of the world. 

Is this good or bad? In many ways it is good. If it is possible to outsource every possible aspect of a firm then it is easier to start a firm. Anyone with the ideas and connections can get moving quickly because they can focus on the particular knowledge that provides them with a competitive advantage. Building the global network of sites for the exercise of that knowledge -- whether through personal contact or the distribution of a product -- is still a major undertaking, and the building of a multitude of global professional networks was one of the great unwritting stories of the 1990s. This is an entirely different picture of the work of starting a firm than was commonplace even ten years ago, and it is still nascent and poorly understood. The new picture is also beneficial for the efficiencies that derive from the real-time wiring of global relationship networks, just as it is dangerous for the increased potentials for globalized market power. 

The greatest danger, however, lurks beneath the explosion of innovation and diversity that modular business and protocols make possible. Let us distinguish between two kinds of diversity: deep and shallow. Deep diversity is diversity that arises from independent evolution in unrelated and completely incommensurable institutional, cultural, and technical contexts. Shallow diversity is diversity that is generated from within a shared framework such as a grammar, a set of modules, or the settings of parameters. Shallow diversity can apply to the structure of different products, or it can apply to the infrastructures by which the products are designed, manufactured, and distributed. 

Deep versus shallow diversity is obviously a continuum, with deep diversity an impossible ideal at one end and shallow diversity a common but not inevitable case at the other end. Human biology, for example relating to the universals of language, limits deep diversity, as do millennia of encounters between cultures. Nonetheless, as a relative matter it is clear that many kinds of deep diversity are disappearing from the world. Indigenous languages and cultures are dying, audiences for local literature and music are absorbing "global" cultural influences, technologies and institutions are being harmonized, and so on. 

Meanwhile, economies of scale and scope are producing many strong drives toward shallow diversity. These drives allow a manufacturer to move away from a single monolithic standard to a degree of mix-and-match, and flexible manufacturing technologies, together with computer-aided design, make it possible to change products rapidly from within a standardized system of distribution. From this perspective, the world is not moving toward an extreme of homogeneity, but to a high level of shallow diversity. The world is still heavily standardized, but the standard applies not to the overall product but to the framework that is used to generate a shallow diversity of products. 

Shallow diversity is certainly adequate for many purposes. But a world without deep diversity would nonetheless be much poorer. The hyperefficiency that settles on particular frameworks of shallowly diverse production may be brittle and lacking in the capacity to shift to a qualitatively different model should circumstances require it. A world of shallow diversity might be caught in a local optimum of historical development: productive to be sure, but not nearly as productive as it could be if it shifted to a qualitatively different path of development. Many theories of institutional learning, such as that of Friedrich Hayek (1960), require different islands of evolutionary development to generate the deep diversity over which natural selection operates, but the whole purpose of globally integrated production networks is to absorb such islands and bring them into conformance with the standard. 

Lastly, a world without deep diversity would leave us poorer as human beings. Perhaps we will maintain always-on connections to everyone we know, but that will do us little good if none of those people knows or feels anything that is deeply different from what we know or feel in our own lives. It is only through the encounter with difference that we are able to question our own assumptions, and it is only through the encounter with difference that we can tell the difference between our own heads and the radical strangeness and challenge of the real world. In a world of shallow diversity, we will prosper and we will die. We must learn to value and conserve deep diversity, and we must learn what it would even mean to replenish what has been lost. 



Philip E. Agre, Surveillance and capture: Two models of privacy, The Information Society 10(2), 1994, pages 101-127. 

Carliss Y. Baldwin and Kim B. Clark, Design Rules, Volume 1: The Power of Modularity, Cambridge: MIT Press, 2000. 

Lowell Bryan, Jane Fraser, Jeremy Oppenheim, and Wilhelm Rall, Race for the World: Strategies to Build a Great Global Firm, Harvard Business School Press, 1999. 

Roger Clarke, The digital persona and its application to data surveillance, The Information Society 10(2), 1994, pages 77-92. 

Ronald H. Coase, The nature of the firm, Economica NS 4, 1937, pages 385-405. Reprinted in Oliver E. Williamson and Sidney G. Winter, The Nature of the Firm: Origins, Evolution, and Development, Oxford: Oxford University Press, 1991. 

David Gelernter, Mirror Worlds, or the Day Software Puts the Universe in a Shoebox, Oxford University Press, 1991. 

Friedrich A. Hayek, The Constitution of Liberty, Chicago: University of Chicago Press, 1960. 

Geoffrey M. Hodgson, Economics and Institutions: A Manifesto for a Modern Institutional Economics, Cambridge, UK: Polity Press, 1988. 

J. Rogers Hollingsworth and Robert Boyer, eds, Contemporary Capitalism: The Embeddedness of Institutions, Cambridge: Cambridge University Press, 1999. 

Bruno Latour, Science in Action: How to Follow Scientists and Engineers through Society, Cambridge: Harvard University Press, 1987. 

David G. Messerschmitt, Networked Applications: A Guide to the New Computing Infrastructure, Morgan Kaufman, 1999. 

Philip Mirowski, More Heat than Light: Economics as Social Physics, Physics as Nature's Economics, Cambridge: Cambridge University Press, 1989. 

Douglass C. North and John J. Wallis, Integrating institutional change and technical change in economic history: A transaction cost approach, Journal of Institutional and Theoretical Economics 150(4), 1994, pages 609-624. 

Melissa A. Schilling, Toward a general modular systems theory and its application to interfirm product modularity, Academy of Management Review 25(2), 2000, pages 312-334. 

Carl Shapiro and Hal Varian, Information Rules: A Strategic Guide to the Network Economy, Boston: Harvard Business School Press, 1998. 

Margaret Wertheim, Pearly Gates of Cyberspace: A History of Space from Dante to the Internet, Norton, 1999.