A Way of Thinking Has No Failure Condition

The field’s official self-description

The ASC Declaration of September 1983, drafted by Ernst von Glasersfeld and prefaced by Stuart Umpleby as past president, is still published as the society’s statement of what cybernetics is. Its first sentence reads: “Cybernetics is a way of thinking, not a collection of facts.” Further on it quotes von Foerster: “Objectivity is a subject’s delusion that observing can be done without him.”^[1]

One caveat decides how this chapter can use that document. The American Society for Cybernetics is not the Viable System Model community. Using its documents to indict Metaphorum would be a category slip, and a reviewer would say so on first reading. The claim here is the narrower one: these documents show what the host discipline rewards, publishes and archives. The VSM community’s own norms are the subject of Chapter 17, and they are argued there from that community’s own material.

The narrower claim matters because the host discipline’s self-description does more than set the background. It is the sentence the tradition reaches for whenever someone raises the question of testing, and it has an institutional home, a named author and a date.

Take the declaration at face value, as it is meant. A way of thinking has no failure condition. No observation could meet it and defeat it, because it makes no claim an observation could contradict. The declaration is not evasive and defends nothing; it gives an accurate description of one of the two things cybernetics is. The difficulty is that the field is also the other thing. The declaration knows this, since it describes “two major orientations” that “have lived side by side in cybernetics from the beginning,” one towards the design of technology and one towards knowledge, but it does not say which of its claims belong to which. Beer had put the other thing in a sentence of his own: what is given in nature is “a collection of principles that systems cannot violate if they are to maintain their identity,” and these principles “are the content of the science of cybernetics.”^[2]

The society’s Definitions page sets out about forty definitions side by side, among them Wiener’s, McCulloch’s, Beer’s, Bateson’s and Pask’s, and declines to adjudicate between them. Larry Richards compiled the list in 1987, when he was the society’s president, and in 1999 stated its intent: “to demonstrate that one of the distinguishing features of cybernetics might be that it could legitimately have multiple definitions without contradicting itself,” and to stimulate dialogue on the motivations of those who proposed them.^[3] A named officer of the society thus sets out, as a goal, that the field need not resolve its own terms.

On the same page sits Beer’s own joke about the impossibility of defining cybernetics: the condemned professor whose last request is to answer the question definitively, and the doctoral student who asks to be executed second.^[4] It is a good joke. Told by Beer in 2001, it is worth more to this book than any outside critic making the same point in earnest, and it should be read the way he told it, as coming from a man who knew exactly what he had and had not pinned down.

What the field rewards

Institutions show their standards in what they honor. The ASC’s stated award criteria are prominence in transdisciplinary research, impact of applications, service to the society and community, excellence in education and pedagogy, and advocacy for the field.^[5]

Empirical test, replication, prediction, having been shown wrong and having revised: none of them is among the criteria. That is a cheap and checkable receipt for the incentive argument in Chapter 9. A field that honors advocacy and application impact, and does not honor refutation, will not produce refutations. Nobody designed it to have that effect. It has it anyway.

The society’s current activity points the same way. As of 29 September 2026 its initiatives page lists a speaker series, the #NewMacyMeetings, a student-run study group and two working groups, one on the society’s archives and one on art, media and cybernetics; the 2026 conference theme was conversational confluences, and the society’s own strapline describes cybernetics as an interdisciplinary conversation.^[6] It has no working group on application, evidence, method or management. Sixty-two years after its founding, the organized capacity of American cybernetics goes into conversation, pedagogy and art. The observation carries a date because it will change, and it should be printed with its date and its URL.

Two absences on the site bear directly on the program in Part Five. The society publishes no methods guidance, no replication or data-sharing norms, and no register of applications and their outcomes; and it keeps no page collecting criticism of the field or of its models.^[7] The live register proposed in Chapter 30 has no counterpart at the discipline’s oldest American society.

The constructivist settlement, and one word

The second-order tradition is not hostile to testing; it relocates it. Umpleby holds that conjectures and refutations are the current consensus and that second-order science should add the observer, the effect of theories on what they describe, and methods as contributions.^[8] Ranulph Glanville read conjecture and refutation as a circular design conversation, and he argued that it is inappropriate to require design to be “scientific,” since scientific research is itself a restricted form of design.^[9] Neither position forbids what this book asks for. Both make it somebody else’s business.

The expected objection from this quarter is that recursion levels depend on the observer, and it is a good objection. Chapter 29 converts it into a test: analyst agreement on the same organization, with and without a declared purpose. An observer-dependence claim that can be operationalized as a measurable agreement rate is a hypothesis; one that cannot is a position. Which of the two ours is, we have never said.

Radical constructivism supplies the sharpest instance, and it is a word rather than an argument. For von Glasersfeld, knowledge is viable as long as it is useful in accomplishing a task or achieving a goal: viability is an epistemic property of a model, held by a knower.^[10] Beer’s viability belongs to the system. He takes the dictionary’s “able to maintain a separate existence” and adds that an organization “is viable if it can survive in a particular sort of environment.”^[11] But on his own account an observer decides which collection counts as the system. “A system is not something given in nature, but something defined by intelligence”; two people “will not necessarily agree on the existence, or nature, or boundaries of any system so detected”; and “it is possible to say only that the treatment of a certain collection of things as a system is helpful,” a point to be tested “by checking on the predictive capability of the insight.”^[12] Which states of the system count depends on “the purposes of the system.”^[13] For Beer, then, viability is a property of a system that an observer has recognized for a purpose. For von Glasersfeld it is a property of the knowledge. Two concepts wear one word, and it is the word this tradition is named for. They are nearer to each other than a quick reading suggests, and part of the distance between them runs through Beer’s own pages.

The consequence runs in one direction only, which makes it dangerous and not merely confusing. Slide from Beer’s sense to von Glasersfeld’s, and “the model is viable” stops meaning that the model is right about what keeps organizations alive and starts meaning that the model has been useful to the people using it. That second sentence is unfalsifiable. It is also, near enough, the sentence the field actually says. Beer’s account makes the slide easy, because it grounds the recognition of a system in its being helpful. It also supplies the brake, a test of predictive capability, and the brake is the part the field does not use. Here the slide gets its name; Chapter 23 will fix the narrow claim in Beer’s sense precisely so that it cannot happen unnoticed.

We have found no source that treats this equivocation, so the argument is ours and must be presented as ours. It may also be wrong, and the way to find out is to put it in front of the people whose word it is.

What this chapter does not establish

It does not establish that the ASC has failed at anything. The society has never claimed to be an evidence-producing body, and holding it to a standard it does not profess would be the error this book accuses others of making.

Nor does it establish that second-order cybernetics is an obstacle. Its central claim, that the observer belongs in the account, is one this book accepts and one Chapter 29 is built to take seriously.

What it establishes is that the discipline’s official self-description offers complete shelter from testing, that its incentive structure does not reward refutation, and that the tradition’s central term has two senses in active circulation, and the field slides between them. None of these is anybody’s fault. All three are ours to work around, because we are the ones making claims about organizations rather than about ways of thinking.

Receipt — say which layer each claim is on, in the teaching material. The target is specific and already public: SCiO’s five-course VSM sequence (VSM1 to VSM5, taught by Patrick Hoverstadt; the first four sold pre-recorded at £200 each, the fifth on enquiry), Metaphorum’s public “Viable System Model” page (Chapter 17), the diagrams reproduced in them, and the two paragraphs on SCiO’s own “Why use the Viable System Model?” page, where “has proven to be a scalable, super-fast, flexible, rigorous, approach” and “a systemic model of what any type of organisation needs in order to be able to survive in a changing environment” sit within a few lines of each other — the first an instrumental claim, the second a constitutive one, and nothing in the page marking the difference.^[14] Label every claim in that set as one of three kinds: a way of thinking (no failure condition, and none claimed), an instrumental claim (assessed by usefulness, with the measure named), or a constitutive claim (assessed by observation, with the disconfirming observation named, per Chapter 28). Publish the labelled set. Price: two working days per course module, done once, by the people who already own the material; a full pass over one certification syllabus is inside a week. The result is falsifiable in a modest but real sense: if the constitutive column comes out empty, then the model is a way of thinking and the neuroanatomical mapping, the necessary-and-sufficient claim and the pathology derivations must leave the teaching, which is a substantial and specific loss. If the column is not empty, every entry in it is now a debt with a named test, and Chapter 30 can price them. What the exercise cannot produce is the present situation, in which the three kinds of claim are taught in one breath and defended by whichever standard the question requires.

Handover

The shelter described here is institutional. The next chapter takes the version that arrives as method, from the neighbour best equipped to give it.