content

16.1 Beer’s account and its standpoint

In 1981 Beer added five chapters to Brain of the Firm (chs. 16–20, pp. 241–397), and in them he told the story of Chile himself. The first edition had appeared in 1972, before most of those events, and its chapter 13 contained a design for an operations room. The second edition reports what happened when that design, and the model around it, was built at the scale of a national economy. The Chilean project is the episode of Beer’s working life that most readers have heard of, and the one most often told by others. What follows stays with Beer’s telling. Beer opens by declaring where he stands. He rejects the role of detached expert: “the posture of a ‘neutral scientific advisor’ became untenable after the experiences of World War II” (BOTF 1981: 245). That sentence explains the form of what follows: “I can tell this story only in the first person, and in an autobiographical vein”. He says he has “ruthlessly expunged from the story mere gossip and my own opinions”, but he does not claim “a pretence of objective omniscience” (BOTF 1981: 245–246). The reader knows from the start what kind of document this is: a participant’s account, disciplined and openly partisan. Beer adds a statement of responsibility: “I could have pulled out of Chile at any time, and often considered doing so; but I did not, and therefore I hold myself accountable for the part that I played” (BOTF 1981: 246). This standpoint matters for the present book in two ways. What Beer reports about Chile is his own reconstruction, and it is reported here as such; the book’s method excludes secondary literature, so his account is not checked against others. And Beer treats the Chilean chapters as part of the argument of Brain of the Firm. The model of the earlier chapters is applied, and the application is reported in the model’s own vocabulary: recursion, Systems One to Five, algedonic signals, variety. Part Four sets out its own chronology (BOTF 1981: 243). Chapter 16 covers the start in late 1971. Chapters 17 and 18 cover the development up to “the extraordinry [sic] events of October 1972”, which “appeared to mark a watershed”. Chapter 19 covers the end, and chapter 20 is a “prospectus”. The sections below keep that order.

16.2 The starting point: regulation in real time

The project began with a letter dated 13 July 1971. It came from Fernando Flores, Technical General Manager of CORFO (Corporación de Fomento de la Producción), the state agency through which Allende’s nationalization program was being carried out (BOTF 1981: 246–247). The letter spoke of “the complete reorganization of the public sector of the economy” (BOTF 1981: 247). From the outset, then, the object was the growing nationalized sector of a whole country. The preparatory work went quickly. By 12 November 1971 the Chilean team and Beer “had all agreed on a plan for the cybernetic regulation of the social economy of Chile”; the members of the team had read the draft manuscript of the first edition of Brain of the Firm (BOTF 1981: 248). The people who were to build the system had the viable system model in front of them in the form Beer had just written down, operations room included. One idea, Beer says, took the most arguing: “The primary point of which I had to convince my friends was that we should firmly take the wholly innovatory step of seeking to regulate the social economy in real time” (BOTF 1981: 248). His reasoning starts from how economies are normally steered. “Even the most advanced countries in the world suffer from a vast lag in the receipt of economic data, and they suffer too from the bureaucratic time it takes to process these data towards any kind of conclusion” (ibid.). He puts a number on it: “In the world’s most advanced countries, economic data arrive very late, perhaps the average delay is nine months […]. This means that most economic decisions are taken out of phase with economic reality” (ibid.). Because the data are late, governments spend their effort estimating where they are, and have little left for where they are going: “they engage in econometric studies aimed at projecting data — not indeed into the future — but only into the present” (ibid.). Against this Beer sets a plain technical claim: “It is perfectly possible, these days, to capture data at source in real time, and to process them instantly” (ibid.). In his account, what was missing was the machinery and the programs to handle such a flow; the technology itself existed. In the terms of the model, this is a matter of the timing of feedback. A regulator that receives its information nine months late is controlling a system that no longer exists in the state reported. Oscillation and overcorrection are what a cybernetician would expect from such a lagged servomechanism. The concern returns in the second of Beer’s five principles (16.6), and again in the argument that plans must continually abort (see chapter 19).

16.3 The four tools

Beer names the project himself: “The code-name Cybersyn is an abbreviation of ‘cybernetic synergy’” (BOTF 1981: 252). In Chile it was also called “Synco” (BOTF 1981: 278). He speaks of “the whole Cybersyn package: the four tools, and a computer” (BOTF 1981: 275), and he lists the four tools when summarising the second phase: “By now we have seen how the four main tools were created: Cybernet, […] Cyberstride, […] Checo, […] and the Opsroom, a new environment for decision, and dependent for its existence on the existence of the other three” (BOTF 1981: 270). In the same passage he characterises Cyberstride as “the suite of computer programs needed to provide statistical filtration for all homeostatic loops at all levels of recursion”, and he dates the operations room: it was taken over on 1 December 1972, and “everything was finished and in experimental working order on 10th January 1973” (ibid.). The order of the list shows the logic of the design. The operations room comes last because it depends on the other three. Without a network there is no data; without filtering there is only an unmanageable volume of data; without simulation there is no view of the future. Each tool can be read against a part of the viable system model. Cybernet was the communications network: “a system whereby every single factory in the country, contained within the nationalized social economy, could be in communication with a computer” (BOTF 1981: 252). It used what already existed, the telex network and microwave links, and needed no new equipment (ibid.). Beer states its intended use in a sentence that bears on the later charge of technocracy: “Now the intention of Cybernet was to make computer power available to the workers’ committees in every factory” (ibid.). On this statement the network was to serve the operational units as well as the center. Cyberstride was the suite of programs for Systems One to Three. Its purpose “was to monitor information flows […] at all levels of recursion; to provide alerting signals” (BOTF 1981: 261). It measured with the “triple index, which measures productivity, latency, and their product — the overall measure of performance” (BOTF 1981: 253; see 16.5). Beer records a decision taken at speed during the programming. The consultants had just agreed on the mathematical techniques for the statistical filter when Alan Dunsmuir telephoned late at night to ask whether Beer had read a paper by Harrison and Stevens, “‘A Bayesian Approach to Short Term Forecasting’”, in the Operational Research Quarterly for December 1971. Beer writes: “I stayed up all night with this paper, and next day we determined to scrap the agreed mathematical approach in favour of theirs. It was a bold step” (BOTF 1981: 262–263). He also notes that “nothing was known about the performance of the Harrison-Stevens techniques in advance” (BOTF 1981: 263). As the permanent suite developed, “changes were made to deal with the need to generate algedonic signals” (BOTF 1981: 264). Beer places Cyberstride in the longer history of his own work. The underlying technique “was first developed in the years 1949 - 53 for the control of steelworks production” (BOTF 1981: 264). He had no electronic computers then, and “the whole system was operated by hand using nomographs to compute standards, desk calculating machines to compute indices, and visual control charts to provide the probability filters” (ibid.). Cyberstride was thus a computerized version of a method already tried in the steel industry, some twenty years before Chile. Checo was the set of simulation programs, a “prolegomenon to a full-bodied System Four” (BOTF 1981: 267). The word “prolegomenon” marks its status as a first step toward the function that looks outward and forward. Beer’s aim for it was ambitious: Checo “would be updated every day by the output from Systems 1-2-3, and would promptly re-run a ten-year simulation; and this has never been done” (BOTF 1981: 268). The sentence joins the two halves of the metasystem that the model keeps apart and connects: the current state reported from inside (Three) and the long-range model of the future (Four). The Operations Room was where the other three came together. Beer insists that it followed the published design: “In Chapter 13 a basic design was given for an operations room: we built it in Santiago, according to that specification — with the exception of the use of a hybrid computer there advocated for economic simulation” (BOTF 1981: 268). In The Heart of Enterprise he describes the idea in more general terms. The room was a way of discharging System Four through “a mode of integration described as a focus”, in which “the whole model of the viable system, at all levels of recursion, would be displayed. It was to be a club-house” (HOE 1979: 242). He then refers back: “In Brain of the Firm I outlined the design for an Operations Room. This was subsequently built for President Allende in Santiago, Chile” (ibid.). How far did the system reach? Beer reports that by the end of the second phase “something like seventy percent of the socio-industrial economy was operating this system, involving about four hundred enterprises, through Cybernet” (BOTF 1981: 264). Textual note: in the chapter on October 1972 he gives “Roughly 60% of all enterprises” as included in the Cybersyn system (BOTF 1981: 311). The two figures refer to different things (a share of the economy on the one hand, a share of enterprises on the other). They describe roughly the same moment: the first figure the end of Beer’s second epoch, the second figure “the situation as it was at the start of the new epoch” (ibid.), his third. Beer does not reconcile them. They are not necessarily contradictory, but they should not be merged into a single figure.

16.4 Recursion and autonomy in Cybersyn

A system that connects every factory to a central computer invites the suspicion that it is an instrument of central control. Beer answers in the terms of recursion. Take a unit Beta that is one of the System One elements of a larger unit Alpha. Beta does more than report upward; it receives “the whole Cybersyn package: the four tools, and a computer”. The consequence, in Beer’s words: “Its autonomy preserves it from nosey-parkerdom in Alpha System Three” (BOTF 1981: 275). Each level of recursion is itself a viable system with its own metasystem, and so needs its own instruments. Beer states the result: “Moreover, the schema represents maximum decentralization, since any given level of recursion receives directly only its own Cyberstride reports” (ibid.). The higher level does not see the detailed indices of every plant below it. It sees what the lower level’s filtering passes on as exceptional. That is the variety attenuation the model requires on the vertical channels, and it is the structural guarantee of autonomy that the Third of the five principles will state (16.6). In formal terms the same argument reappears as the claim that the metasystem should make “minimal use of variety attenuators” downward (see chapter 17). On this design, an operations room at one level would display only what that level’s own filters had let through. Beer does not claim that practice matched the design. The analysis of each factory was meant to be carried out with the people who worked there: the teams were to explain the flowchart model of a plant, to enlist help in building it from those who worked in it, and to obtain agreement on the performance measures. Beer writes that “this was not always being done in the intended spirit; and it seemed likely that this fact aided detractors who later wished to call Cybersyn ‘technocratic’” (BOTF 1981: 271). Describing the shift in priorities after October 1972, he distinguishes two groups among those who reacted to the changed situation. Some regarded Cybersyn as a political instrument; and “The gratification was experienced by the technocratically minded, who wanted effective management regardless of the political framework” (BOTF 1981: 322). Beer thus separates the design from the attitudes of some who operated it, and he names the technocratic reading as a real tendency inside the project as well as an accusation from outside. He also records proposals made at the end of 1972 to push autonomy further: a worker as “Director of Total Industry”, the formal handing over of the tools, and “a replacement of technocratically oriented OR groups by workers’ groups”. He adds a reservation of his own: “And yet there was something manifestly starry-eyed about this set of ‘expectations’” (BOTF 1981: 328).

16.5 The indices of achievement

The triple index that Cyberstride computed was not invented for Chile. It comes from chapter 11 of Brain of the Firm, where Beer defines three levels of achievement (BOTF 1981: 163):

actuality: “what we are managing to do now, with existing resources, under existing constraints.”
capability: “what we could be doing (still right now) with existing resources, under existing constraints, if we really worked at it.”
potentiality: “what we ought to be doing by developing our resources and removing constraints, although still operating within the bounds of what is already known to be feasible.” The three are ordered: actuality does not exceed capability, and capability does not exceed potentiality. From them Beer derives three ratios: “productivity: is the ratio of actuality and capability; latency: is the ratio of capability and potentiality; performance: is both the ratio of actuality and potentiality, and also the product of latency and productivity” (ibid.). Every index built this way is a dimensionless number between zero and one, whatever is being measured. A copper mine and a textile mill can then be watched with the same statistical machinery, because what is filtered is the movement of a ratio, whether the underlying quantity is tons or meters. That is what allowed a single program suite to monitor hundreds of different enterprises and to raise an alert when an index moved in a statistically unusual way. Beer links each level to a form of planning: “Planning on the basis of actuality I call programming. Planning on the basis of capability I call planning by objectives. Planning on the basis of potentiality I call normative planning” (ibid.). The distinction foreshadows the fuller treatment of planning in The Heart of Enterprise (see chapter 19). The further a plan reaches beyond what is currently being done, the more it involves decisions about removing constraints, and the more it belongs to the metasystem rather than to the operations.

16.6 The People Project and the five principles

Beer describes one wall of the operations room as empty on purpose: “there remained a blank wall in the Operations Room: it was intended to house an algedonic meter” (BOTF 1981: 284). Cybersyn was concerned with the regulation of the industrial economy. In parallel Beer worked on a second project, on “the systems dynamics of the interaction between government and people” (BOTF 1981: 278). Part of it was the algedonic meter (see chapter 12.5), a device by which people could signal satisfaction or dissatisfaction directly. The blank wall shows where it was meant to go: in the same room as the economic displays, as the channel through which the people’s own signals would reach the metasystem. Another part was “The Manual”. Beer formulated five principles, one for each subsystem, “each expressed in a single cybernetic sentence”, and arranged in the order 5, 2, 1, 3, 4 (BOTF 1981: 288–289):

First: “System Five within a people’s government cannot be an élite ruling class: it is somehow the embodiment of the mass of the people themselves.”
Second: “The speed of response in an essentially lagged servomechanism is critical: note especially the anti-oscillatory Systems Two.”
Third: “Variety engineering enables us to design homeostatic subsystems that obey the Law of Requisite Variety, and determine a recursion of metasystems: this preserves System One autonomy.”
Fourth: “Command is neither a matter of wielding authority nor of overloading the central axis with variety-attenuating regulations: System Three delegates authority and accountability to the elemental level, and looks for synergistic advantage in taking a synoptic view of Systems One.”
Fifth: “Special attention is necessary to System Four: otherwise System Five identifies with System Three, and the whole cerebral metastructure collapses; instead of adaptation and self-determination, we are left with crisis management.” The order follows an argument. It starts with identity and legitimacy (Five), moves to the damping of oscillation (Two) and the autonomy of the operating units (One), then to the character of command (Three), and ends with the warning about neglecting the future (Four). Read together, the principles restate the model as political guidance. The first denies that the top of the model is an elite. The fourth denies that System Three is a command center in the ordinary sense. The fifth names the failure in which a government absorbed by day-to-day control loses its capacity to adapt, and its phrase “crisis management” anticipates what actually happened in October 1972 (16.7). Beer meant to translate this “cybernetic substance of ‘the manual’” into simple statements for the population, in booklets, leaflets, posters and, as he hoped, songs (BOTF 1981: 289). The popular version, the booklet “Five Principles for the People, towards Good Government” (early September 1972), was to open with the statement: “The revolution of ownership is two years old. IT IS TIME for the revolution of government to begin” (BOTF 1981: 289). Nationalization had changed who owned the economy; the model, as Beer saw it, was meant to change how it was governed.

16.7 October 1972

In October 1972 the “gremios” went on strike. This is the event Beer calls a watershed, and he insists that the gremios were not trade unions. The operations room was not yet finished, but the communications center and Cybernet were working, and the infrastructure was used for crisis management. Beer describes what was done: “Two of the senior cyberneticians organized a filtration system: some signals were algedonic, requiring instant decisions, while others could be attenuated into elements of the pattern that established the factual situation in real time” (BOTF 1981: 312–313). Two ideas from the model are applied here at once. Algedonic signals bypass normal filtering because they demand immediate action; the rest are attenuated and aggregated into a picture of the situation. In a crisis the distinction lets a small group handle a large flow of messages without ignoring the urgent ones or drowning in the routine ones. Beer draws a general conclusion from the episode: “it had at last been made dramatically clear that properly organized information deployed in real time is a major national resource” (BOTF 1981: 314). This is the claim of 16.2, confirmed in his view by events, and confirmed before the room built to display such information was ready. The success had a cost for the project as first conceived. After October the priorities changed, and Beer spent “only a fraction of my time” on Cybersyn (BOTF 1981: 320). The network grew in a different direction: Cybernet was extended to ministries, the Central Bank and government agencies (BOTF 1981: 327). Beer argues that the changed circumstances obscured a structural problem “in the recursive mapping of the economy” (BOTF 1981: 322). The original modeling of November 1971 had defined the economy “in terms of its assets — the land, minerals, industry — and the ownership of those assets” (ibid.). But, he continues, “the preoccupation with ownership makes sense only insofar as it betokens ruling power”, and other factors supervene: “there seems to be a resurgence in the self-organizing capability of communities to promote themselves as viable systems” (BOTF 1981: 323). The recursions drawn along lines of ownership were therefore not the only ones that mattered. Beer carries the argument through to a redesign. The ownership model “nominated only owned assets (whether public or private) as viable systems, which between them constituted and exhausted System One”, and events “powerfully suggested that two levels of recursion were mixed together in that preliminary model” (BOTF 1981: 323). The economic community and the distributive function had each grown a management independent of ownership, so the social economy and public industry “were not after all co-exstensive [sic]: the latter was a System One of the former” (BOTF 1981: 324). He therefore proposed a new level of recursion in the Ministry of Economics, “with (initally) three Systems One” (ibid.): the industrial recursions regulated through Project Cybersyn and presided over by the workers’ committees; the economic communities, with the People Project, presided over by the neighbourhood councils (JAPs); and a new “commerce” linking them, over which the Comandos de Abastecimiento presided and for which an “Allocation Project” became possible (ibid.; Figure 47, BOTF 1981: 325). The original Cybernet was not adequate at this ministerial level, but Cyberstride could handle its filtration needs, since Beer “could think of no regulatory indicator appropriate to the two new components of System One at the higher level of recursion that could not readily be expressed as a triple index” (BOTF 1981: 326). The fifth principle had warned against a metasystem reduced to crisis management, and the most visible use of the tools came in a crisis.

16.8 The public dimension and the end

Beer also judged the project’s fate in public opinion, and he did so in cybernetic terms: “the battle of Ashby’s Law for the reputation of Project Cybersyn had […] been lost in advance” (BOTF 1981: 336). He applies the Law of Requisite Variety to the public debate itself. The variety of the stories that could be told about the project, in Chile and abroad, exceeded anything the project’s members could generate in reply. On 11 September 1973 Beer was in London (BOTF 1981: 347). In the final chapter of Brain of the Firm he returns to that date in the context of entelechy, the realization of a system’s own potential: “It is not conceivable that ‘Progress’ for the Chilean people could have veered through a hundred and eighty degrees during the morning of September 11th, 1973” (BOTF 1981: 382). The remark is aimed at the idea that progress is a fixed direction. What changed that morning was the regime; the needs of the people stayed what they had been. In The Heart of Enterprise he writes, more briefly, that the experience of using the operations room “did not outlast the date of 11th September, 1973, which saw the overthrow of democracy in Chile, and the death of the President” (HOE 1979: 242). The sentence stands in a chapter about System Four. Beer uses Chile there as the one realized example of the operations room, and reports its end in a single clause. The Chilean experience shapes much of what Beer wrote in the following years. The Massey Lectures of autumn 1973 open the question of freedom as a design problem (see chapter 17). Platform for Change (1975) collects arguments on institutions and metasystems from 1969 to 1973, most of them written before the Chilean work began; its narrative pages refer back to Chile (PFC 1975: 417; see chapters 18 and 18.6). And the critique of planning as a product, developed in The Heart of Enterprise, takes up the lagged information with which the project began (see chapter 19).

16.9 Prospectus: crisis and entelechy

Chapter 20 of Brain of the Firm is where Beer stops narrating. He draws “a veil over subsequent events” and wants “to turn if that is possible to positive conclusions” (BOTF 1981: 349). What he found common to the governments and firms that later invited him was “the perception of impending crisis” (ibid.). He defines the term: “What counts as a crisis is the expectation of loss of control: in other words cybernetic breakdown in the institution” (ibid.). The first lesson of Chile follows from that definition: “ACT FAST” (BOTF 1981: 350). Beer reports that he made speed the essence of his proposals in four other countries, which he does not name, and that in each case the crisis exploded before work could begin (ibid.). The cybernetics of crisis. Beer models three parties, A, B and C, whose interests intersect in a developing crisis (Figure 48, BOTF 1981: 354), and sets out “four major cybernetic requirements” of stability (ibid.). He recapitulates them himself (BOTF 1981: 376):

“(i) The system is obedient to Ashby’s Law of Requisite Variety;”
“(ii) Information channels maintain variety entrusted to them;”
“(iii) Transducers neither attenuate nor amplify variety;”
“(iv) The time cycle is synchronous for all subsystems.” These, he writes, “were drawn respectively from the four Principles of Organization enunciated in this book’s companion volume, The Heart of Enterprise” (BOTF 1981: 377). The core of the argument is that a party’s regulatory model loses variety at each step: in the group onto which the model “condenses” (BOTF 1981: 356), in transmission, and in transduction into the conceptual model. “It follows therefore for Party A that the variety of its own model decides the variety of the crisis” (BOTF 1981: 364). The loss is then made good only in appearance: variety is “ostensibly but not actually restored — essentially by shouting” (BOTF 1981: 365). The feedback loop that ought to damp the system is, on Beer’s analysis, supplied by the mass media, and it works against stability (BOTF 1981: 366). His verdict: “The outlook, by these tokens, for crisis regulation must be adjudged bleak” (BOTF 1981: 376). He applies the model to his own conduct. Before the coup he had set up three coded channels of communication with three groups in Chile (BOTF 1981: 358). To protect the people involved, he made each code specific to one channel and told none of the three that the others existed. He calls this “a mistake of metasystemic proportions”, and one that he “would repeat for the reason given” (BOTF 1981: 362). The cybernetics of progress. The second section starts from the behavioural modes of chapter 15 and argues that the crisis mode has become the normal condition (BOTF 1981: 378). Beer then argues that “the notion of Progress is politically relative” (BOTF 1981: 381). Beyond the intentions that Progress measures lie meta-intentions, “something to be achieved which is politically invariant” (ibid.). For this he revives Aristotle’s word entelechy, defined with the help of the Oxford English Dictionary: the meta-intention “expresses the realization or complete expression of some function; it manifests a condition in which a potentiality has become an actuality” (ibid.). The remark about 11 September 1973 quoted in 16.8 is his illustration (BOTF 1981: 381–382). The measure of progress in this sense is the triple index of 16.5. Beer’s “Tentative and Final Methodology towards Entelechy” requires that the methodology “be able to measure performance — quantitatively defined in Chapter 11 as the ratio of the two estimates”, and it “becomes thereby (Chapter 11, Figures 28 and 29) the tool of normative planning that brings the entelechy into focus” (BOTF 1981: 386). On my reading, this closes a circle in Part Four: the index that Cyberstride computed for each factory returns at the end as the measure of what a system could become. Under “On Technological Method” Beer asks whether the computers and the operations room were “merely frills”, since “the basic success was due more to organizing structural variety than to anything else” (BOTF 1981: 394). He grants the case weight and rejects it: “on balance I disbelieve its humanistic promise”, because “The requirement for speed in the operation of the problem-solution homeostat makes this outstandingly a technological issue” (ibid.). The cybernetics of prospectus. The last section explains why the chapter gives no list of lessons for next time. Not a single question “can even be clearly formulated (much less answered) without specifying an attendant set of circumstances in great detail” (BOTF 1981: 395). The viable system is instead to be interpreted “in terms of the potentiality dormant in its actuality” (ibid.). The chapter ends: “This is the Prospectus of the Entelechy. The futurological prospectus does not have requisite variety” (ibid.).

References

HOE — Beer, Stafford (1979). The Heart of Enterprise. Chichester: John Wiley & Sons.
BOTF — Beer, Stafford (1981). Brain of the Firm. 2nd ed. Chichester: John Wiley & Sons. (1st ed. London: Allen Lane, 1972.)