The Cybernetic Brain

by Andrew Pickering

  Figure 6.9.Cockroach-controlled robot. (Photograph by Garnet Hertz. Used by permission.)

  Figure 6.10.Eduardo Kac, The Eighth Day, 2001. Transgenic artwork with biological robot (biobot), GFP plants, GFP amoebae, GFP fish, GFP mice, audio, video, and Internet (dimensions variable). The photograph shows the biobot in the studio, with its internal amoebae already in place, before it was introduced into the transgenic ecology that constitutes The Eighth Day. Source: www.ekac.org/8thday.html. Used courtesy of Eduardo Kac.

  The Viable System Model

  When Beer's dreams of biological computing came to an end in the early 1960s, this implied not an abandonent of his vision of the cybernetic factory but a transformation of it. Beginning in 1972, a trilogy of books developed his account of what he called the viable system model—the VSM for short: Brain of the Firm (1972; 2nd ed., 1981), The Heart of the Enterprise (1979), and Diagnosing the System for Organizations (1985). The VSM was at the forefront of Beer's thinking and consulting work from the 1960s to the 1990s and attracted a considerable following. A two-day workshop on the VSM held at the Manchester Business School in January 1986 led to the production of an edited volume describing further interpretations and applications of the VSM by a range of academics, consultants, and people in industry and the military (Espejo and Harnden 1989), and variants of the VSM are still practiced and taught today.

  The VSM transformed Beer's earlier vision of the cybernetic factory along two axes. First, the simulation of the cybernetic factory discussed above, where human management filled in for the not-yet-built U-machine, became in effect the thing itself. Beer continued to look forward to as much computerization of information gathering, transmission, and transformation as possible (as in the T- and V-machines). But the ambition to dispense with the human entirely was abandoned. Instead, human managers were to be positioned within purposefully designed information flows at just those points that would have been occupied by adaptive ponds or whatever (e.g., the position they in fact occupied in the earlier simulations).

  Second, Beer extended and elaborated his conception of information flows considerably. In Brain of the Firm, the first of the VSM trilogy, he argued thus: The aim of the firm had, as usual, to be to survive in an environment that was not just fluctuating but also changing—as new technologies appeared in the field of production and consumption for example. How was this to be accomplished? What would a viable firm look like? The place to look for inspiration, according to Beer, was again nature, but now nature as the source of inspiration in the design of viable organizations, rather than nature as the immediate source of adaptive materials. Beer's idea was to read biological organisms as exemplary of the structure of viable systems in general, and to transplant the key features of their organization to the structure of the firm. In particular, he chose the human nervous system as his model. In the VSM, then, Beer's strategy was to transplant the organic into the social, but not as literally as before. The firm would no longer contain trained mice or Daphnia at its heart; instead, information flows and processing would be laid out as a diagram of human bodily flows and transformations.

  A B Figure 6.11.Control systems: A, in the human body; B, in the firm. Source: S. Beer, Brain of the Firm, 2nd ed. (New York: Wiley, 1981), 131, figs. 23, 22. Permission: John Wiley & Sons.

  The spirit of the VSM is strikingly expressed in the juxtaposition of two figures from Brain of the Firm. Figure 6.11A is a schematic of the body; figure 6.11B is a schematic of the firm. Brain goes into considerable detail in rehearsing the then-current understanding of human neurophysiology—the pathways both nervous and biochemical along which information flows, the operations performed upon it at different points—and how it might be transcribed into the organization of the firm. I am not going to attempt an extensive review. But some key features need to be singled out. The VSM divided the "nervous system" of the firm into five subsystems, numbered 1–5 in figure6.11B. Although the VSM was supposed to be applicable to any organization (or any viable system whatsoever), for illustrative purposes figure 6.11B is a diagram of a company having four subsidiaries, labeled 1A–1D, and as indicated in figure6.11A, one can think of these in analogy to systems in the body, controlling the limbs, the heart, the kidneys, and so on. A notion of autonomy arises here, because such systems in the body largely control themselves without reference to the higher levels of the brain. The heart just speeds up or slows down without our ever having to think about it. It adapts to the conditions it finds itself in by reflex action largely mediated somewhere down the spinal column. Beer's contention was that subsidiaries of the firm should be like that. They should act in the world and on one another (supplying materials to one another, say) as indicated by the circles with wavy lines and arrows moving off to the left from them, and their performance would be monitored at appropriate points on the "spinal column"—the square boxes labeled 1A and so on. This monitoring would consist of a comparison of their performance in relation to a plan already given by the higher management of the firm, and deviations could be compensated for by appropriate adjustments to their behavior. The model here would be a simple servo-controlled negative feedback mechanism.

  But even at this level of the body, autonomy is not complete. Figure 6.11B shows direct connections between the control systems, 1A, 1B, and so on. The idea here is that if something unusual is happening in subsidiary 1A, say, which supplies parts to 1B, then 1B should know about it so that it can take steps to allow for that. There must, that is, be some information channel linking these subsidiaries, as there is between the heart and the lungs. And further, Beer observed, in the human body there are usually several different channels linking levels of the nervous system. Figure 6.11A thus distinguishes two further channels—the sympathetic and the parasympathetic systems— and figure 6.11B shows their equivalents—lines of information flow upward, from the controllers on the spinal cord (the squares) and from the operational sites (the circles). The equivalent of the sympathetic system is system 2 of the VSM. Beer understood this as attempting to damp conflicts that could arise at the system 1 level—the various subsidiaries trying to hoard some material in short supply to each other's detriment, for example. This damping, which Beer knew enough not to expect to be necessarily successful, would be accomplished by reference to system 3. Corresponding to the pons and the medulla at the base of the brain, system 3 would be basically an operations research group, running models of the performance of the entire ensemble of subsidiaries, and thus capable, in principle, of resolving conflicts between subsidiaries in the light of a vision available to none of the subsidiaries alone.19

  At this stage, no information has traveled upward beyond system 3 into higher layers of management. The parasympathetic system, however, was envisaged to act somewhat differently. This traveled straight up to system 3 and was intended to transmit an "algedonic" "cry of pain." Less metaphorically, production data would be monitored in terms of a set of dimensionless ratios of potential to actual performance of the kind that Beer had introduced in his 1953, paper discussed earlier. If one of those ratios departed from a predecided range, this information would be automatically passed onward to system 3, which, in the light of its models, would act as a filter, deciding whether to pass it on to levels 4 and possibly 5.20

  Figure 6.12.World War II operations room, near London, during the Battle of Britain. Source: Beer 1968a, 23.

  I am inclined to think that system 4 was Beer's favorite bit of the VSM. The equivalent of the diencephalon and ganglia of the human brain, this had access to all the information on the performance of the firm that was not filtered out by system 3; it was also the level that looked directly outward on the state of the world. If the level 1 systems had access to information directly relating to their own operations, such as rising or falling stockpiles or order books, level 4 had much wider access, to national economic policies and changes therein, say, to the price of money, the results of market research, and what have you. System 4 was, then, the T-U-V
system of Beer's earlier model, with the humans left in.

  Beer envisaged system 4 as a very definite place. It was, in fact, modelled on a World War II operations room, of the kind shown in figure 6.12 (taken from Beer's 1968 book Management Science), as developed further by NASA at "Mission Control in the Space Centre at Houston, Texas, where the realtime command of space operations is conducted" (Beer 1981, 193–94), and updated with all of the decision aids Beer could think of (194–97). All of the information on the state of the firm and of the world was to be presented visually rather than numerically—graphically, as we would now say. Dynamic computer models would enable projections into the future of decisions made by management. Simply by turning knobs (197), managers could explore the effects of, say, investing more money in new plant or of trends in consumption. Feedbacks that had passed the level 3 filters would also arrive at system 4 from the lower levels, "signalled appropriately—that is, if necessary, with flashing red lights and the ringing of bells" (194), alerting management to emerging production problems, perhaps to be passed on again to level 5. In terms of social organization, "I propose a control centre for the corporation which is in continuous activity. This will be the physical embodiment of any System 4. All senior formal meetings would be held there; and the rest of the time, all senior executives would treat it as a kind of club room. PAPER WOULD BE BANNED FROM THIS PLACE. It is what the Greeks called a phrontisterion—a thinking shop" (194).

  System 4, then, differed from system 3 in the range of its vision, a vision which now encompassed the future as well as the present, and Beer imagined system 4 as a primary locus for decision making on change. If the levels below strove to implement given production plans for the firm, level 4 was the level at which such plans were drawn up and modified.

  Finally we arrive at system 5, the equivalent of the human cortex. This was the level where policies were deliberated upon and the most consequential decisions were made (Beer 1981, 201). This was where the human directors of the firm had been imagined to continue to exist in the original blueprint for the cybernetic factory. The firm's viability and continued existence, and even growth and evolution, were maintained by systems 1–4. The job of system 5 was, therefore, to think big at a metalevel superior to questions of mere viability.21

  This outline of the VSM is now almost complete, but two points need to be added. First, the various levels of the viable system were intended to be coupled adaptively to one another. The 3 and 4 systems, for example, would engage in the process of reciprocal vetoing discussed earlier. Level 4 might propose some change in the overall operating plan for the firm; this would be run through the OR models at level 3 and might be rejected there—perhaps it would place excessive strain on one of the subsidiaries. Level 3 could then propose some modified plan back to level 4, which could run it through its models. Perhaps the plan would be vetoed again, once more transformed, and returned to level 3. And so on, back and forth, until some operating plan agreeable to both systems 3 and 4 was discovered.

  Second, we should note a recursive aspect of the VSM. Beer argued that firms were themselves parts of bigger systems—national economies, say. The entire 1–5 structure of the firm would thus appear as a single system 1 on a diagram of the national economy. This in turn should be a viable system with its own levels 2–5 overseeing the ensemble of firms. Proceeding down the scale instead of up it, each subsidiary of the firm should also be a viable system in its own right, meaning that the level 1 systems of figure 6.11 should actually have their own levels 1–5 within them. Figure 6.13 shows what became Beer's standard diagram of the VSM, depicting two levels of such recursion. The two level 1 subsidiaries in square boxes at the lower end of the spinal column (running up the right-hand side) are shown as having their own 1–5 structure projecting downward at an angle of 45 degrees (and each has two subsidiary "operations" within its large circle). Likewise, the 3-4-5 brain at the top of the spinal column is itself enclosed in a square box, indicating that it is part of a level 1 system of some bigger system. Beer felt that such recursivity was a necessary property of viable systems—they had to be nested inside one another "like so many Russian dolls or Chinese boxes" in a chain of embeddings which "descends to cells and molecules and ascends to the planet and its universe" (Beer 1989a, 22, 25).

  Figure 6.13.The viable system model showing recursive embeddings. Source: S. Beer, Diagnosing the System for Organizations (New York: Wiley, 1985), 136, fig. 37.

  The VSM was thus a vision of the firm in the image of man. Especially, functions of management and control were envisaged on the lines of the human brain and nervous system. The brain and nervous system were simulated by a combination of information technologies and real human beings appropriately arranged. One could say that the VSM is one of the most elaborated images of the cyborg in postwar history, though the word "cyborg" is tautologous here, standing as it does for "cybernetic organism." Any viable system was exactly that, according to Beer. We should also note that the VSM was the "circuit diagram" (Beer 1981, 123) of a time machine, an adaptive system accommodating itself to the exigencies of the unknown in real time, ranging from mundane disturbances at the level of production to world-historical changes.

  The VSM as Ontology and Epistemology

  The basic ontological vision that the VSM conjures up is the same as that of the cybernetic factory before it: the world as an ungraspable and unmasterable space of becoming; the organization as open-endedly and performatively adaptable. The VSM, however, also suggests some refinements to that picture. First, my portrayal of the cybernetic factory was centered on the brain of the firm as a unitary entity, the U-machine, in dialogic conversation with the firm's environment. Beer's conception of the VSM, in contrast, was one in which the overall behavior of the firm was the upshot of an interplay of many active but quasi-autonomous elements, the VSM's systems 1–5, themselves also interacting with different aspects of the firm's environment. The recursive aspect of the model adds an indefinite sequence layers of elements to this picture. The VSM thus moves us toward a vision of ontological multiplicity, a multiplicity which is, furthermore, irreducible: the system 3 of a given organization is not reducible to the organization's system 4, say, or to the system 3 of another organization.22

  Second, we can return to the question of goals. Walter's and Ashby's devices had fixed goals that organized their adaptation: the homeostat reconfigured itself so as to keep its essential variables within preset limits. Beer's conception of the VSM, in contrast, specified no goals whatsoever, except adaptation itself. And we could think of Heidegger: adaptation in the VSM was a process of revealing rather than enframing. The process that Beer called reciprocal vetoing between levels of the system, for example, was by no means as negative as the phrase suggests. A veto from one level to another was at the same time an invitation for a novel counterproposal, a way of finding out what the other had to offer.

  The VSM was explicitly about information flows and transformations, so we can return now to a consideration of cybernetic epistemology as well as ontology. In Beer's vision, viable systems do contain knowledge—representations of their own inner workings and of their environment—principally enshrined in the OR models at level 3 of the VSM and the projective models at system 4. What should we make of this? First, we could recall that in his work on truly biological controllers, Beer had sought to avoid this detour through representation. His biological computers did not contain any representational or symbolic elements; they were intended simply to do their adaptive thing. The VSM, then, one might say, was a concession to representation as a response to the failure of biological computation. And it is appropriate to recall that, as I remarked before, Beer did not much trust representational models. He did not think, for example, that one could arrive at a uniquely correct model of the firm and its environment that could function unproblematically at level 4 of the VSM. This is a direct corollary of the idea that both the firm and its environment are exceedingly complex.

 
Beer did not, however, take this to imply that the construction of representational models was a useless endeavor. His idea, instead, was that the models in question should be continually examined and updated in relation to erformance—"continuously adapted" (Beer 1981, 185) or even always "aborting" (Beer 1969 and 1994b, 151). The company should act in the light of the future projections of the model at level 4, but then actual developments in time should be compared with expectations from the model's simulations. These would not, in all probability, match, and the model should be adjusted accordingly.23 The VSM thus stages for us an image of a performative epistemology—a more elaborated version of what we have seen in the preceding chapters. The "knowledge components" of the VSM were not an end in themselves; they were geared directly into performance as part of the mechanism of adaptation, and they were revisable in performance, just like the other components of the VSM; they were not the controlling center of the action.