The Corner
Judy’s corner is crowded. Between rounds the people in it lean over the ropes with advice, and she has begun to notice that almost none of it is about winning the fight in front of her. One tells her she has never been beaten, which is true and means nothing: nobody has ever scored her. One has a stopwatch and a clipboard and a method, and has been timing the wrong arm for years. One says that scoring is the wrong frame entirely, that what she does is a dance and a dance cannot be lost. Across the hall, in the seats, sit the people who say she is not a boxer at all. They have never watched a round. What frightens her is not them. It is that nobody has ever bought a ticket.
Parts One and Two brought charges. Part Four will say what survives them, and Part Five what it would cost to put the surviving claim at risk. First comes the question a reader of Chapter 9 is left with: if the defects never cost us anything, who was standing around not charging us?
The answer divides in two, and the two halves of this part follow the division.
The first half is about the communities that appear to be on our side. Some of them are. Checkland’s recoverability, Ormerod’s critical rationalism, Troncale’s and Miller’s call for validation, and the free energy principle’s public split between a principle and its process theories are all positions this book can adopt without amendment. The chapters here are about the others: neighbors whose support is real and whose intentions are not in question, but whose epistemic settlement, once we adopted it, made the model unchargeable. A community that tells us our model is useful rather than true has not attacked us. It has handed us the sentence we use to avoid the question, and we have used it. The first to hand it over was our own founder. “So remember: / a model is neither true nor false: / it is more or less useful,” Beer wrote in 1985, in Diagnosing the System for Organizations.1 In 1966 he had tied the same idea to a test: the usefulness of a model is measured “by straightforward testing of the predictive value of the scientific model. If this predictive value is low, we had better start again.”2 The tradition kept the sentence and dropped the test, and when the neighbors offered their own versions of the sentence, we took those without their tests as well. Beer also made the other kind of claim: the model he taught is “a cybernetic model of the necessary and sufficient conditions of any viable system.”3 The line this part draws between the two kinds of claim runs through the founder’s own books before it runs between us and our neighbors.
These are the false friends of the title. The title accuses nobody of bad faith; it describes what an inherited criterion does when it is inherited or borrowed without the practice that earned it.
The second half is about the adversaries. There are fewer of them than the reception suggests, and they are weaker than the reception suggests, and that is the finding, not a relief. Chapter 16 already took the three objections that land, Werner Ulrich’s, Friedrich Hayek’s and Robert Cooper’s, and this part does not revisit them. What remains is a rival ontology with no test in it; a set of sweeping dismissals of systems thinking in general that were never aimed at this model in particular; a body of formal criticism directed at a theorem we borrowed rather than at anything we built; and a silence. The silence is the serious one, and the last chapter is about it.
Four notes on reading the part.
It is about us, not about them. Every community described here is doing something defensible within its own program. System Dynamics built its validation philosophy and then built the test batteries to go with it. The American Society for Cybernetics has never claimed to be an evidence-producing body and does not pretend otherwise. Radical constructivism has been explicit for forty years about what it means by viability. None of them owes us a failure condition. We took their conclusions and left their obligations, as we had already done with Beer’s prediction test, and that charge is ours to answer.
The standard is the one Chapter 16 set. A position counts as an adversary here when it identifies something the model needs in order to do the job it claims, cannot supply from its own resources, and has not been given by any amendment since. A position counts as a false friend when it supplies a criterion that the model then uses in place of a test, and the criterion cannot in principle settle the claim it is used to settle. The two failures are different, and they are priced differently.
The instrumental/constitutive line does the work. Chapter 23 will fix the narrow claim on this distinction; we need the distinction here first, because it is what makes this part tractable. Usefulness, recoverability, fit to context and boundary critique are adequate criteria for instrumental claims, claims about whether a diagnosis helps. None of them can decide a constitutive claim: that there are five subsystems, that they are necessary, that they are sufficient, that System Five corresponds to anything in a cortex. Wherever this part says a criterion has been misapplied, this is the misapplication it means.
Not everything here has been verified. The quotations, attributions and dates in these six chapters were first drawn from two research notes prepared in September 2026. Some have since been checked against the original sources: the material from SCiO’s own website on 14 September 2026, and on 23 and 25 September the Metaphorum page, the Schwaninger and Scheef sentence, the Hoverstadt podcast transcript, Espinosa’s post and the pages of Pfiffner’s dissertation. That first-hand evidence is concentrated in Chapters 17 and 18, and the footnotes say where it appears. Beer’s own sentences, added on 29 September 2026, do not come from those notes; they are quoted from his books and were checked against the printed pages. Each chapter flags the items still unverified, and they are not few. A book about the difference between a claim and a receipt cannot print a sourced claim on the strength of a summary of a source. Every quotation below must be read against the original before this part goes to press, and the chapters say so where it matters most.
Chapter 17 takes the community closest to us and the warrant it publishes: in fifty years nobody has found a case in which the model failed. Chapter 18 takes the two most recent quantitative tests of the model and shows that neither was built so that it could lose. Chapter 19 takes the host discipline’s official self-description, cybernetics as a way of thinking rather than a body of facts, and asks what a way of thinking can be charged with. Chapter 20 takes usefulness instead of truth, the criterion that has done the most damage by being the most respectable, and begins with Beer’s own version of it and the test he attached. Chapter 21 takes the last shelter, the claim that testing the model is a category error, sets the status Beer declared for his model in 1966 beside the one Karl Friston declared for the free energy principle, and asks what each community did with the exposure its founder accepted. Chapter 22 turns to the adversaries and finds that the one that costs us most is not a person.
Notes
- Stafford Beer, Diagnosing the System for Organizations (John Wiley & Sons, 1985), 2, where the three lines are set apart from the text, after “you are not determining absolute facts: you are establishing a set of conventions.” Chapter 20 gives Beer’s other versions of the sentence. ↩
- Stafford Beer, Decision and Control (John Wiley & Sons, 1966), 113. Chapter 21 quotes the passage in full. ↩
- Stafford Beer, The Heart of Enterprise (John Wiley & Sons, 1979), 235. ↩