Internet: Towards a Holistic Ontology       Contents | Introduction | Chapters 1 3 | Conclusion | Notes | References | Appendices A B

Chapter 2:
Cybernetic Organism

Although Deleuze and Guattari named their central philosophical concept after the biological category of the rhizome, the previous chapter considered the Net to be biological only in structural terms.  The richness of the biological metaphor has not escaped popular writers, who have used organic imagery to describe the Net.  Rheingold for instance speaks of the Net as "propagating and evolving" like colonies of bacteria; entities on the Net as members of "entire ecosystems"; the parallel development of bulletin board systems as taking place "[l]ike real grassroots ... from the ground up ... self propagating ... difficult to eradicate"; and the rate of growth of Usenet as "biological — slow at first, and then exponential" (1994: 6, 107, 131 & 119).  Also, Rheingold mentions that in 1960, J. C. R.  Licklider, a key figure in the early development of the Net, wrote a paper titled "Man-Computer Symbiosis" — a phrase which treated computers as if they were a biological category (1994: 70).  In Licklider's description of this "symbiotic relationship", humans and machines operated as a single cybernetic system (Hafner & Lyon, 1996: 31, 34-5).  This concept of computers serving as an extension of human intellect was shared by Douglas Engelbart, another key figure (Hafner & Lyon, 1996: 78; Rheingold, 1994: 70).16

     A critical look at the use of such imagery reveals the possibility that the analogy between biology and technology extends beyond structure to genesis and development — the Net could be considered biological both as an entity and as a process.  It would perhaps be more prudent to say that the Net is like biology, rather than say that it is a kind of biology, a statement which would entail a broader definition of biology.  Instead of redefining biology, this chapter will borrow from cybernetics, specifically "the science of control and communication in the animal and the machine" founded the 1940s by Norbert Wiener and Arturo Rosenbleuth.  Cybernetics makes no informational distinction between machines and organisms, based on the theory that "the laws governing control are universal" (Beer, 1988: 197).  Cybernetics takes for granted that information flow is a function of adaptive control, a style of management which "is more like steersmanship than dictatorship", as is reflected in the Greek etymology of the term (1988: 197).  When we make the analogy between biological and technological systems, we can understand both as falling under the rubric of an as yet ill-defined evolutionary cybernetic theory, with much terminology borrowed from biology.  This chapter will explore such an evolutionary cybernetic theory, by examining the ideas of Rothschild, Dawkins, Lynch, Routt, and Kelly.

A mixed bag of metaphors

We begin with a collection of suggestions concerning parallels between the perpetuation of organisms and the propagation of information.  In Bionomics: Economy as Ecosystem, Rothschild calls for a paradigm shift away from the clockwork Newtonian "economy as machine", and towards a new understanding based on the biological analogy of "economy as an evolving ecosystem" (1990: 334-5).  To address the failings of outdated "orthodox economists", who continue to "assume that technology does not change" and who have as a result "lost all touch with economic reality" (1990: xii), Rothschild borrows the holistic approach from bionomics.  This is "the branch of ecology that examines the economic relations between organisms and their environment" to gain a perspective of how "the interplay of forces ... maintain[s] stability while spawning change" (1990: 335).  He observes the parallel by noting the common element of exchange:

Briefly stated, information is the essence of both systems.  In the biologic environment, genetic information, recorded in the DNA molecule, is the basis of all life.  In the economic environment, technological information, captured in books, blueprints, scientific journals, databases, and the know-how of millions of individuals, is the ultimate source of all economic life.  (1990: xi)

     While Rothschild concedes that his analogy is not perfect, he claims that it is powerful enough to be useful for explaining such complex economic phenomena as "competition, specialization, cooperation, exploitation, learning, growth, and several others", which are common to both ecosystems and market economies (1990: xiii).17  Both use "intricately linked feedback loops" to self-organize and to maintain a sensitive yet flexible balance (1990: xiv).  He uses this bionomic outlook to address economic concerns ranging from savings and investments to poverty, public education and pollution control.  If Rothschild is right, then the evident failure of Marxist economics is not due to a "poor implementation of a sound theory", but rather because "its core elements violate processes essential to the functioning of all living, evolving systems" (1990: 107).  He explains that Soviet-style socialism has failed because central state control breeds a large bureaucratic apparatus which is not only inefficient at managing complex tasks best left to decentralized market forces, but also imposes a heavy drain on resources, and ultimately becomes a parasitic organization because performance is not tied to reward (1990: 323-33).  In a market economy, feedback loops use prices as a signalling mechanism, so that in the absence of any directing authority, the market as an interconnected network is able to steer itself into creating value through mutual cooperation (1990: 323-33).  Such rewarding mutualism exists abundantly in natural ecosystems as symbiosis — a value-creating "positive-sum" process in which everybody benefits, contrary to the simplistic Marxist assumption that profit-making is a "zero-sum" game in which resources remain fixed in value and owners of capital profit at the expense of workers (1990: 324).

     If state control is so evil, then where does that leave the role of government?  Here Rothschild leans towards a laissez-faire style of economic management wherein:

The traditional notion of government's economic role — pushing the buttons and twisting the dials of society's economic machinery — is replaced by a vision of government as the astute cultivator of society's economic ecosystem, patiently nurturing the natural processes of growth.  (1990: xv)

In other words, while natural ecosystems run entirely on their own, economies still need to be supervised after a fashion.  This is why Rothschild refers to bionomics as nothing more than an analogy.

     In 1976, the computational biologist Dawkins proposed in The Selfish Gene that "in some respects, ideas (which he called 'memes') behave like genes, replicating themselves across brains the way genes do across bodies" (Wallace & Mangan, 1996: 232).  Wallace and Mangan add that communication technologies have amplified memetic reproduction across space and time (1996: 233).  These two proposals argue for a case of information making no distinction between human cognitive memory and communication technologies as replication media.  More recently, in Thought Contagion, Lynch has echoed these ideas: thoughts and beliefs, like viruses, propagate themselves across populations over time by "'programming' for their own retransmission" (1996a).  Borrowing from the Darwinian theory of "survival of the fittest", he claims that the fittest memetic combinations are most successful the more they can replicate in human host populations (1996b).  However, he has attempted to free his ideas of the constraints of metaphor by distilling a formal analysis of memetic behaviour in the language of symbolic mathematics, in an effort to arrive at a specific theory of memetics (n.d.).  It is still possible to argue that this theory has a place in a general evolutionary cybernetic theory.

     In the arena of cultural theory, Routt has observed that the ephemerality of specific popular artworks paradoxically gives rise to "durability on several levels", one of which is generic (1993: 134-5).  While Lynch has disavowed a "consumer theory of belief acquisition", preferring instead a memetic theory of "[r]ecursive transmissivity and receptivity" (1996c), Routt does not shy away from an organic metaphor of popular art consumption as a metabolic process essential to healthy human existence (1993: 135-6).  The fact that ephemeral popular art nevertheless "cannot rely on the compulsion of necessity to bring audiences to it", but must compete on the basis of freshness and novelty (1993: 137), seems to borrow from the Darwinian sources similar to those of Lynch.  In his Ph.D.  thesis on the same subject, Routt has even used terminology seemingly from the same theoretical neighbourhood as Dawkins and Lynch: "replication" and "massification" work with "planned obsolescence" (finite life-spans), producing continual freshness and novelty (adaptive mutation) and generic persistence (species survival) (1981).  The actual mechanics may be somewhat different from biological ones, but the conditions necessary for evolutionary processes to operate appear to be present in Routt's perspective as well, even if he makes no direct reference to Darwinism.

     The parallel between the organic and the technological becomes even more apparent when we consider information rendered in digital technology.  Dawkins explains in River Out Of Eden that all genetic life can be understood as essentially "a process of digital-information transfer" (Schrage, 1995).  Genetic codes, or "codons", propagate themselves down successive generations with virtually flawless exactitude, except where the evolutionary mechanism of mutation happens to step in (Dawkins, 1995, excerpted in Schrage, 1995).  The implication of digital electronic information and biological genes being 'genetic cousins', so to speak, is that analogy ends and identity begins.18  Identity makes a compelling case for information being subject to evolution; but even without a theory of identity, it might be plausible to use analogy alone to propose a unified evolutionary theory in which processes such as natural selection and mutation apply on both sides of the institutional divide between biology and information technology.  By studying both positive and negative analogies between the two, it might be feasible to seek an understanding of the general evolutionary operations common to self-organizing cybernetic systems such as the Net.

     Much of the remainder of this chapter, then, will take an approach borrowed from Kevin Kelly's Out of Control (1994), a non-technical but meticulous and interdisciplinary study of the convergence between biology and technology, centred on what he calls "neo-biological civilization" (1-a).19

Neo-biological technology

A neo-biological system, on Kelly's account, bears some striking resemblance to Deleuze and Guattari's rhizome, and can be described by several essential characteristics: it is decentralized; it is a network; it is the sum of connected, heterogeneous parts; it is nonlinear and complex; it evolves; it forms symbiotic or coevolutionary relationships; it adapts; it organizes itself; and its logic is counter-intuitive.  As in describing the rhizome, it will be no simple task detailing these attributes of neo-biological systems, as they are all interrelated parts of an integrated whole and do not work in isolation.  Nevertheless, Kelly has managed to compile a set of principles for creating neo-biological systems.  These he happily terms "The Nine Laws of God":

  • Distribute being
  • Control from the bottom up
  • Cultivate increasing returns
  • Grow by chunking
  • Maximize the fringes
  • Honor your errors
  • Pursue no optima; have multiple goals
  • Seek persistent disequilibrium
  • Change changes itself  (24-a)

     These instructions provide a clue as to how a complex theory of evolving cybernetic systems might be understood, and how it might apply to the Net.  We can hazard a line of inquiry along the lines of these categories, arranged according to a plausible causality:

• Distributed network structure
• Self-organization
• Adaptability
• Coevolution

Here we might bear in mind that these themes are more interrelated than discrete, and that the categorization is contrived for the purpose of study.

Distributed network structure

Kelly's neo-biological system is not a singular, atomistic whole, but a network of interconnected parts — a mirror image of Deleuze and Guattari's principle of multiplicity.  It begins with heterogeneous entities forming horizontal relationships with one another, relationships which grant no privileged authority to any position.  Kelly makes much of the synergy emerging from connections: "More is more than more; it's different"; "an emergent property", such as temperature, is a "group characteristic" absent from the individual (2-e).  Furthermore, increasing returns frequently come into play, so that more connections mean more synergy: "The total number of possible interactions between two or more members accumulates exponentially as the number of members increases" (2-e).  Deleuze and Guattari have mentioned this very phenomenon:

A multiplicity has ... determinations, magnitudes, and dimensions that cannot increase in number without the multiplicity changing in nature (the laws of combination therefore increase in number as the multiplicity grows).  (1987: 8)

     An important observation of such a network or multiplicity is that conscious direction is seldom if ever centralized in any one agency.  In nature, synergy commonly occurs whenever organisms, of their own accord, form relationships of mutually beneficial and mutually directed symbiosis and are richly rewarded for it.  In artificial realms, robotics engineers have been faced with practical limits when they try to build complex, centrally controlled systems (3-a).  When Rodney Brooks at MIT tried a different approach using distributed control, it worked so well that NASA incorporated the design into Sojourner, the first robotic rover to explore the surface of Mars (Coale, 1997).  This approach is called "subsumption architecture", a method of building complexity in a bottom-up direction:

  1. Do simple things first.
  2. Learn to do them flawlessly.
  3. Add new layers of activity over the results of the simple tasks.
  4. Don't change the simple things.
  5. Make the new layer work as flawlessly as the simple.
  6. Repeat, ad infinitum.  (3-c)

In one of Brooks' early designs — a cockroach-like robot — each robotic leg has its own microprocessor and is able to walk autonomously.  By sensing the positions and movements of the other legs, it coordinates itself with them so that "walking emerges out of the collective behavior of the 12 motors" (3-b).  Because control is not centralized but distributed, the robot does not get confused if one of its legs malfunctions, but is able to continue walking on the remaining legs, just as a real cockroach would.  It turns out that in nature, the ability of insects to walk is based on the same 'design' (3-b).  A new steering module may be added on top of the legs to control the overall direction and starting and stopping, without telling the legs how to do their jobs.  Even though the instructions for walking are, strictly speaking, contained in each of the legs, it is as if walking is part of the steering module's behaviour, as if it controlled every aspect of the walking — the behaviour of the legs being subsumed under that of the steering module.  The subtle difference between subsumption architecture and top-down tree structure (such as a Marxist economy) is the location of specialized control.  In one, it is distributed between specialized parts; in the other, it is centralized at the top which tries to specialize in every component task.

     The specialized parts of a distributed network are tested and debugged in the real world before higher levels of control are added.  This ongoing experimentation brings to mind the rhizomatic principles of cartography and decalcomania.  Subsumption architecture forms a dynamic mapping with reality, whereas the centralized approach makes a model or tracing of the world within its 'mind' and tries to match its behaviour against this tracing.  Resolving and maintaining this internal "world model" incurred high coordination overheads in previous robotic designs, a problem which was solved by abandoning the tracing altogether, in favour of a functional and active form of engagement:

So difficult was the task of coordinating a central world view that Brooks discovered it was far easier to use the real world as its own model: "This is a good idea as the world really is a rather good model of itself." (3-e)

This decentralized control corresponds to the "acentered system" of Rosenstiehl and Petitot, to which Deleuze and Guattari refer:

The principal characteristic of the acentered system is that local initiatives are coordinated independently of a central power, with the calculations made throughout the network (mutiplicity).  "That is why the only place files on people can be kept is right in each person's home, since they alone are capable of filling in the description and keeping it up to date: society itself is the only possible data bank on people.  A naturally acentered society rejects the centralizing automaton as an asocial intrusion" (1987: 519-520; quotation from Rosenstiehl & Petitot, 1974: 62)

     Subsumption architecture is the essence of the Net, although in practice it serves more as a guide than a rule.  To recall a discussion in the previous chapter: the testing and implementation of TCP/IP closely followed the method of mapping or using the world as its own model, whereas OSI formed a tracing of reality on which design was based.  In subsumption architecture, after the primary functions have been established, the next higher level of functioning may be built over it; so after packet-switching has been taken care of, e-mail, FTP and the World Wide Web may follow.  To tamper with the basic layers after they have been established could mean certain disaster.  This explains why the OSI protocol failed to replace TCP/IP despite widespread official support — TCP/IP had already become entrenched in the Internet, and network administrators were unwilling to replace a tried and proven foundation (Hafner & Lyon, 1996: 249-50).  Around 1983 when the smaller arpanet had to make the switch from the older and more limited Network Control Protocol to TCP/IP, it was a painful, albeit necessary, transition (1996: 248-9).

     Apart from being better at building complexity and reliability, a distributed network structure also has the advantage of being more versatile than a tightly controlled centralized system: instead of having a single agent pursuing a singular agenda, multiple, independent and loosely coupled agents are able to seek multiple goals, and owing to a rich repertoire of specialization, they are more tolerant of environmental changes (6-c).  The connections are loose and able to reconfigure themselves to suit various conditions.  In computing, object-oriented programming (OOP) is founded precisely on this philosophy.  Wherever feasible, programme code performing a specific function is contained in a module or subroutine, where it is ready to be invoked by the main function or other subroutines.  Errors are easier to isolate and repair this way, and the programme "grows by incrementally assembling working subunits", increasing in complexity as it does so (11-d).  An important aspect of this modular design is that it easily allows for recursion.  Here, a module may repeatedly invoke itself in an iterative loop until a task is completed; or else two modules may alternately invoke each other, continually keeping tabs on and feeding information back to each other.  This type of connective relationship — the feedback loop — is the foundation of self-organization, according to classical cybernetics.


The connections of a network seem to give it a quality of cohesion, so that it is neither a collection of isolated parts nor a strictly bound unity, but somewhere in between.  In this in-between state, disparate elements begin to take on the ontological status of a singular entity, and the emergent characteristic of this entity is a shade of selfhood: self-regulation, self-governance, self-organization, and even self-perpetuation.  Kelly asserts that the emergence of this selfhood begins with the feedback loop, and cites the early example of an invention by Ktesibios who lived in Alexandria in the third century BC: a water clock which "kept extraordinarily good time (for then) by self-regulating its water supply", accomplished by means of a float and a valve regulating each other — an idea resurrected in the modern toilet flush (7-a).

     There are two distinct forms of feedback loops: positive and negative.  Rothschild's example of the slime mold is an instance of complex self-organization arising spontaneously from simple amoebas engaged in a positive feedback loop.  A slime mold appears to have cohesive unity, but is in fact "just one phase in the remarkable life cycle of certain amoeba species" (1990: 255).  In order to bypass the limits on its ability to find food, an amoeba sends out pulses of a hormone called cyclic AMP to neighbouring amoebas, causing them to approach while returning the same signal.  These signals trigger even more signals.  The escalating loop of information exchange results in a mass of cells coming together and acting as if it were a single-minded organism, offering up a "bulbous tip" for other creatures to eat and deposit elsewhere (1990: 255-6).  Rothschild claims that just as self-organization can spontaneously arise out of the collective behaviour of amoebas exchanging chemical signals, so too can it arise out of the consolidated actions of independent traders exchanging market signals — a market economy is hence a self-organizing system operating on feedback loops using prices as the signalling mechanism (1990: 263).  In the case of amoebas, a positive feedback loop escalates to the point where a new form of order — the slime mold — emerges.20

     At the heart of self-organization, however, lies the negative feedback loop, in which information is used to adjust the system so that it stays within certain boundaries.  The thermostat is an example of a simple negative feedback system, as is James Watt's innovation of the "governor" (a rotating speed regulator) for the steam turbine engine (7-a).  In each case, control is achieved by the flow of information: one module's output becomes the input of another module.  Cause becomes effect and effect become cause, in what the cybernetician Heinz von Foerster calls "circular causality" (7-c).  In artificial feedback loops, where modules are tightly coupled, full control of one module means full control of the entire system (7-c).  Instead of regulating every single action, a higher level of control — what Kelly calls "metacontrol" (7-c) — delegates control to several different feedback loops.

     Chaos theory provides another perspective through which to understand self-organization.  This young interdisciplinary science sees "underlying patterns" in seemingly chaotic systems such as the weather, economies and ecosystems (Rothschild, 1990: 260).  The physicist Doyne Farmer observes the ways in which order mixes with disorder:

Here was one coin with two sides.  Here was order, with randomness emerging, and then one step further away was randomness with its own underlying order.  (Quoted in Gleick, 1993: 252)

In the 1960s, it was discovered that complex systems could be modelled by simple nonlinear equations, and that these systems displayed "sensitive dependence on initial conditions" — "[t]iny differences in input could quickly become overwhelming differences in output" (Gleick, 1993: 8).  Because of this quirk, chaotic systems such as the weather cannot be predicted far into the future with great accuracy.  Yet it is common knowledge that the weather is "locally unpredictable, globally stable" — it tolerates continuous noise, absorbs occasional disturbances and readily returns to its stable irregularity (1993: 48).  These two key characteristics of chaotic systems — unpredictability and stability — are respectively described by the concepts of "nonlinearity" and "attractors".

     The study of nonlinearity is important because while complex feedback loops may be approximated by linear functions, ecologists discovered in the 1950s that they were better modelled by nonlinear equations (Gleick, 1993: 61-4).  This is a realm outside straightforward linear mathematics, or in the words of Rothschild, the "nonlinear world occupies a middle ground between perfect Newtonian predictability and utter randomness" (1990: 260).  Nonlinear equations yield results that vary dramatically with every executed loop or iteration, as every new development feeds back into the system to change the conditions affecting future results.  As Gleick puts it: "Nonlinearity means that the act of playing the game has a way of changing the rules" (1993: 24).  The only way to predict the final outcome is to run the equations through the required number of iterations.  Nonlinearity has cropped up in the networking world.  For instance, "the total unraveled length of a [cable] network can be shortened by adding nodes to it!" — but only up to a certain limit before cable requirement goes up again (2-g).

On the other hand, in 1968 Dietrich Braess, a German operations researcher, discovered that adding routes to an already congested network will only slow it down.  Now called Braess's Paradox, scientists have found many examples of how adding capacity to a crowded network reduces its overall production.  (2-g)

There is therefore no proportional, linear link between connectivity and performance.

     In addition to nonlinearity, many observed chaotic systems are also subject to "attractors" — fixed points or patterns towards which chaotic systems gravitate (Gleick, 1993: 121-53).  So in the face of nonlinear unpredictability, stabilizing tendencies do occur:

Nonlinear feedback regulates motion, making it more robust.  In a linear system, a perturbation has a constant effect.  In the presence of nonlinearity, a perturbation can feed on itself until it dies away and the system returns automatically to a stable state.  (Gleick, 1993: 193-4)

While the exact temperature of a given day could not be predicted long beforehand, "the weather will not go outside the limits defined by its underlying chaotic pattern" (Rothschild, 1990: 260-2).  And while nothing much is known about strange attractors generally, they are an underlying phenomenon that has been mathematically plotted (Gleick, 1993: 140, 143, 151).  Before we speculate on how attractors might come into play on the Net, let us consider the Net's connectivity.

     Kelly's analysis of self-organizing systems pays close attention to the connections between members of a network.  He believes that indiscriminately linking up everything simply won't do; the key to creating self-organization lies in connecting things "in an organized, indirect, and limited way" (20-e).  Most people who have multiple phone numbers, faxes, pagers and e-mail addresses would agree with Kelly that being too connected does nothing to increase one's organizational efficiency — networks which are too highly connected regress into a state of red-taped grid-lock (20-d).  Based on this suggestion that too much connection can be counterproductive, Kelly makes the case for technologies of discriminate disconnection, such as encryption, being a "necessary counterforce to the Net's runaway tendency to link" (12-f).  The need for useful disconnection in the face of too much connection may well be part of a negative feedback loop adjusting for optimum connectivity.  This could be one self-organizing tendency of the Net.

     In chaos theory, attractors are understood to be behind stability and self-organization.  Kelly suggests that when some level of connectivity is reached, forms of self-organization would eventuate:

I mentioned ... the controversial idea that in any society with the proper strength of communication and information connection, democracy becomes inevitable.  Where ideas are free to flow and generate new ideas, the political organization will eventually head toward democracy as an unavoidable self-organizing strong attractor.  (20-d)21

Assuming that Kelly has a point, we need to ask: has the Net reached that "inevitable" point of self-organization?  Certainly, there appears to be no central agency organizing the Net.  The World Wide Web Consortium founded by Tim Berners-Lee, the inventor of the Web, comes close to being a global Net authority, but Berners-Lee insists that the Consortium does not "enforce anything" or "dictate", and is nothing more than a facilitator for consensus between companies which produce specifications for the Net (Schwartz, 1997).  In other words, a form of distributed control or democracy permeates the structure of the Net.22

     Self-organization is to be distinguished from the self-consciousness by means of which intelligent organisms direct themselves.  Strictly speaking, it is not the Net which produces its own specifications; it arguably obtains them through coevolutionary relationships with the companies facilitated by Berners-Lee's Web Consortium.  The Net does not have to be a living organism capable of conscious self-direction and negotiation with other entities in order to engage in such relationships.  Biological viruses are not strictly classified as organisms, since they have no metabolism and can survive crystallization (OERD: 1614-5), but that does not stop them from self-replicating.  According to the evolutionary biologist Tom Ray, it appears to be "a universal property of life that all successful systems attract parasites" (quoted in 15-a).  In technology, the Net has become the breeding ground for scores of computer viruses which are, broadly speaking, functionally identical to their biological counterparts.  Computer viruses have been common enough before the popularity of the Net, but the ease with which information can be exchanged on the Net has made them even more common.  In order for the Net to support computer viruses, it must already have attained a level of success as a system.  This is hardly surprising, considering Dawkins' assertion that all biological life is essentially informational (Schrage, 1995).  Since electronic information flows much faster than biological information, it follows that an electronic informational system would be able to attain self-organized stability much sooner than ecosystems.

     The Net is a system circulating pure information, and if we entertained Kelly's notion of connected information being subject to the forces of some strong attractor, self-organization could well be part of that attractor.  At the heart of the Net is packet-switching, the fundamental mechanism providing automatic traffic control for the flow of data (December, 1996).  Here, machines freely exchange information without human intervention, except for maintenance and repair.  In packet-switching, traffic control is distributed, so that autonomous data packets are able to find their own ways to their destinations — not unlike a nest of ants or the traffic in a city.  But lest we forget that the Net would not have been possible without humans, and would serve no conceivable agenda without them, a more interesting scale of self-organization might be observed by considering a broader system comprising a coupling between humans and the Net — a notion not entirely unfamiliar to Licklider and Engelbart.  That discussion must be properly postponed until the section to follow, "Coevolution".  For now, let us consider a specialized form of self-organization with an agenda.


As mentioned in the previous chapter, Baran was the first to realize that the more redundant links a communications network has, the more robust it is.  (Indeed, he intended redundancy as a means to survive a nuclear disaster.) As Kelly puts it: "More is different" (2-e).  This synergy is derived from the level of connectivity — more connections per node means greater reliability.

     The notion of redundancy is based on the assumption that, at any given time, some portion of it will have failed.  This assumption is crucial, for as long as such failure is an inevitable reality, redundant links provide the alternative routes through which data packets are able to bypass untrafficable ones.  In a sense, failure is incorporated as part of the design, fulfilling the call of Malpas and Wickham for sociological inquiry to recognize the inevitability of failure and incompleteness in any ongoing project, and for the role of governance to use failure and resistance as positive forces serving to constrain and direct, like the gears of a machine (1995).  In the language of chaos theory or cybernetics, failure would be incorporated as informational feedback into the system, and be a crucial element of self-organization.  In natural selection, attrition is employed in the vital role of weeding out the unfit, so that populations remain healthy and competitive.  Evolution has come not only to expect that form of local failure which is attrition, but also to require it in order drive ecosystems to greater global robustness.  Prairie ecosystems, for example, have become so adapted to seasonal fires that they depend on fire to eradicate intruding species and to hatch their prairie seeds (4-a).  Fire is so essential to prairie ecosystems that they cannot be established or maintained without it.

     The notion of continual failure is essential not only to redundancy, but also to the evolution of the Net.  Garfinkel predicts that just as the US power grid needed to go through the "Great Blackout" of 1965 in order to reach the level of service it currently provides, so too will the Net need to experience a series of frequent outages to immunize it against future breakdowns (1996).  Bob Metcalfe, one of the founding fathers of the Net, has been "one of the loudest voices proclaiming the coming doom", but, as Garfinkel notes, Metcalfe has also predicted that "each collapse" will bring the Net closer to "a new industrial-strength Internet" (1996).

     What is commonly put down to experience and growth belies the incremental and indirect process of assembling complexity, a process which, as Kelly points out, requires "multiple tries over time", economies of increasing returns, and the transient establishment of supporting elements (4-d).  The older search technology, Gopher, is one such transient supporting element which has paved the way for the more graphical and user-friendly Web browser, and as the World Wide Web overtakes Gopherspace, the usefulness of browsers increases.

     Another form of redundancy as a strategy for greater adaptability aims for more types of specialization than is economically sensible, in order to adapt to harsh conditions.  For example, an observation made during the Biosphere 2 project was that savannas "have evolved in conditions of periodic disturbance and require a natural kick every now and then" (9-a).  Tony Burgess has also noticed that rich desert ecologies thrive on variable rainfall: "if you have a constant schedule of rainfall with respect to the annual temperature cycle, the beautiful desert ecology will almost always collapse into something simpler" (quoted in 6-a).  Complex systems adapt by being flexible, which means developing a range of repertoires.  In natural selection, the production of varied specialization by mutation is needed to counterbalance the forces of attrition.  This "duet" of "author" and "editor" generates new, successful species (19-d).  Kelly infers that a network must always remain open to a heterogeneity of elements if it is to be robust; it must weigh the overheads of developing and maintaining diversity against the benefits of tolerance to harsh conditions — even if such openness to diversity proves less than useful at times:

In exchange for a swarm system of 17 million computer nodes on the Internet that won't go down (as a whole), we get a field that can sprout nasty computer worms, or erupt inexplicable local outages.  But we gladly trade the wasteful inefficiencies of multiple routing in order to keep the Internet's remarkable flexibility.  (2-f)

     Another way of studying network adaptability focuses on the quality of connections.  Ashby and Gardner wrote a paper in 1970 indicating that increasing the connectivity of a network circuit beyond a certain point "would suddenly decrease the ability of the system to rebound after disturbances" (6-b).  Stuart Kauffman has found that increasing the connectivity between nodes of a network increases the network's adaptability to environmental changes, but "beyond a certain level of linking density, continued connectivity would only decrease the adaptability of the system as a whole" (20-d).  Once the network has attained optimal connectivity, it can change in size while maintaining its adaptability, by keeping to the same "average link rates" (20-d).  Based on Kauffman's Law, Kelly advises as follows:

In our networked society we are pumping up both the total number of people connected (in 1993, the global network of networks was expanding at the rate of 15 percent additional users per month!), and the number of people and places to whom each member is connected.  Faxes, phones, direct junk mail, and large cross-referenced data bases in business and government in effect increase the number of links between each person.  Neither expansion particularly increases the adaptability of our system (society) as a whole.  (20-d)

One consequence of an environment which encourages tight coupling between species in an ecosystem is the development of evolutionary convergence or coevolution between the species.  This convergence or coevolution is evident between the various technologies of the Net, as will be illustrated in the following section.


Coevolution describes general relationships of which symbiosis is a mutually beneficial variety.  It is not a completely new idea — Darwin in his On the Origin of Species witnessed "coadaptions of organic beings to each other" (quoted in 5-b).  Coevolution sees a convergence, as Kelly puts it, "like a tango":

[T]he butterfly was reflected in the plant, and the plant was reflected in the butterfly.  Every step the milkweed took to keep the monarch larvae at bay so the worm wouldn't devour it completely, forced the monarch to "change colors" and devise a way to circumvent the plant's defenses.  ...  In defending itself so thoroughly against the monarch, the milkweed became inseparable from the butterfly.  And vice versa.  Any long-term antagonistic relationship seemed to harbor this kind of codependency.  ...

     ...  The milkweed and monarch, shoulder to shoulder, lock into a single system, an evolution toward and with each other.  Every step of coevolutionary advance winds the two antagonists more inseparably, until each is wholly dependent on the other's antagonism.  The two become one.  (5-b)

This sounds almost like rhizomatic "becoming", except where "becoming" conflates time into a simultaneity (Deleuze, 1993), coevolution moves along a trajectory of temporal progression.  The process is irreversible; the more a relationship coevolves, the more difficult it is to break out of it.  Without the villain, the superhero's career would end.  The US faced an identity crisis when the "Evil Empire" collapsed.  It is not in Microsoft's interest to see Apple sink.  Companies producing anti-virus software count on the continual appearance of new computer viruses.  To use examples from Kelly, prairies need fire and deserts need harsh conditions (4-a; 6-a).  He sees coevolution in just about every form of relationship in which partners adapt to and are codependent on one another: "any organism that adapts to organisms around it will act as an indirect coevolutionary agent to some degree" (5-b).  One begins to suspect that coevolution was broadly defined precisely to bring attention to the web of relationships that exists in every aspect of any network, but a qualitative engagement with the concept yields more usefulness.

     Coevolution is a complex form of feedback loop in which partners not only exchange information and act on it, but also adapt to one another to the extent of developing a dependence on the connections between them.  In a sense, as partners direct themselves and one another, so the system as a whole appears to direct itself — it self-governs and self-organizes.  Where Darwinian evolution stresses the competition between species for resources, coevolution emphasizes the rewards of synergy arising from cooperation and symbiosis, and "each species is locked with its entire environment in a number of feedback loops" (Marshall & Zohar, 1997: 93).  Kelly refers to James Lovelock's Gaia Hypothesis which considers the Earth's biosphere to be one giant, self-organizing, coevolutionary system: "If a thermostat or a steam engine can own self-governance, the idea of a planet evolving such graceful feedback circuits is not so alien" (5-d).  This begins to sound like the connectionist model of neural networks, in which mental faculties are derived from the rich interconnections between individual processors (Marshall & Zohar, 1997: 107-8).

     The level of coevolution is determined by the condition of network connections.  The Spanish ecologist Ramon Margalef has observed a tradeoff between the quality and quantity of connections in a network — a "conservation of connectance":

Margalef noticed ... that systems with many components would have weak relations between them, while systems that had few components would have tightly coupled relationships.  Margalef put it this way: "From empirical evidence it seems that species that interact freely with others do so with a great number of other species.  Conversely, species with strong interactions are often part of a system with a small number of species." (6-b)

Kelly makes the generalization that we "should expect to find a similar law of the conservation of connectance in cultural, economic, and mechanical systems", although such a conjecture has yet to be validated by actual studies (6-b).  In an environment which encourages more coevolution than isolated physiological adaptation, species become so closely linked that "it would be difficult to distinguish where the identity of one began and the other left off", and "we should expect to see many instances of weird symbiotic and parasitic relationships — parasites preying upon parasites, males living inside of females, and creatures mimicking and mirroring other creatures" (6-d).  That is exactly what we find on the Net, where a reasonably small number of distinct 'species' are highly connected to one another.  Mirror sites abound on the World Wide Web.  Client programmes emulate the configurations of other programmes in order to negotiate with network protocols.  Users assume online identities of the opposite gender.  Modems pretend to be telephones and phone-jacks simultaneously.  High levels of recursion occur, such as entire computer networks linking up to the Net at points meant for single computers.  Services straddle or spring up within other services: Gopher insulates the user from the difficult languages of Telnet and FTP with a layer of menus and simple commands (Rheingold, 1994: 107); Web-based chat services appear independently of IRC.  Tom Ray's experiments in artificial evolution of computer viruses have succeeded in producing highly coevolved parasites which prey on viruses, as well as hyperparasites which prey on parasites (15-a).23  Similarly, the Net has not only seen the appearances of computer viruses and anti-viruses, but also anti-anti-viruses and anti-virus-viruses (Hyppönen, 1994).

     Computer viruses represent an interesting case for examination of coevolution.  Viruses — whether computer-based or biological — carry the code for their own replication, but more specifically they "utilize the host cell's own biochemical mechanisms to assemble replica viruses" (OERD: 1614-5).  In terms of reproduction, viruses on their own are as incomplete as drones from a bee hive; but viruses coupled with hosts form an entirely new reproductive unit.  So, the coevolution of viruses with hosts produces fresh possibilities, in this case the self-replication of viruses.

     Borrowing from Lynch's idea of memetic replication and the theories of Deleuze and Guattari, I would like to propose a specific mode of rhizomatic proliferation — a process of technological self-replication achieved through coevolution.  McHoul has recently proposed the notion of "cyberbeing", which constitutes "a new relation between human being and equipment; to the point where the two cease to be distinct ontological categories in the strictest sense" (1997: 17).  By its utility, the printing press enables the duplication of its own enabling technology (1997: 22) — instructions on how to make a printing press and use it, or in other words the code for its replication.  The potential for replication must be realized by its relationship with human users.24  Strictly speaking, the printing press does not carry any replication code within itself.  McHoul's "equipmental formation of equipmentality" (1997: 22) might well be provided by coevolution, as the relationship carries out the actual replication which is only latent in the equipmentality of the printing press.  And as the partners adapt to and exploit one another, the coevolutionary relationship shapes them in specific ways.  The cybernetician W. Ross Ashby wrote in 1952:

[An organism's gene-pattern] does not specify in detail how a kitten shall catch a mouse, but provides a learning mechanism and a tendency to play, so that it is the mouse which teaches the kitten the finer points of how to catch mice.  (Quoted in 5-b)

On a larger scale, it is the coevolutionary relationship the cat has developed with the mouse which 'teaches' the kitten how to catch mice.  Likewise, I could easily retrieve information from the Net on how to implement a Web server or, to be less ambitious, how to set up a Web page, but it would require that I pick up certain navigational skills for instance.  As an individual, I need to establish a coevolutionary relationship with the technology before the technology, and its usefulness to me, can proliferate.  As a species, humanity has always had this relationship with technology — each serves and is shaped by the other, in an alliance which propels both further down the coevolutionary trajectory.  One might argue that technology is a passive partner, that it does not actively adapt, that it requires human agency to modify it, but the notion of coevolution does not see adaptation to be necessarily self-determined.  For instance, the environment can be a coevolutionary partner as well:

Many evolutionary biologists in the last century such as T.  H.  Huxley, Herbert Spencer, and Darwin, too, understood it intuitively — that the physical environment shapes its creatures and the creatures shape their environment, and if considered in the long view, the environment is the organism and the organism is the environment.  Alfred Lotka, an early theoretical biologist, wrote in 1925, "It is not so much the organism or the species that evolves, but the entire system, species plus environment.  The two are inseparable." (5-d)

If we take the Net to be a virtual environment, then coevolution takes place in this environment, the components of which are hardware, software and 'liveware' (humans).  This holistic perspective sees the virtual environment as an evolutionary cybernetic system.


One of the main themes in Kelly's book is that as we gain an adaptable, self-organizing, evolving, neo-biological civilization, we must necessarily give up a degree of control over this creation (2-f).  A number of people have expressed scepticism at the idea that the Net can take care of itself, and have even indicated in no uncertain terms that letting it do so would be a recipe for disaster.  The next chapter will explore these views, and propose that they do not in fact contradict Kelly.

Chapter 3