---
title: "Ch27 what actually replaces us content"
description: "For mathematics, Aubin, Powers and Friston. For evidence about multi-level governance, Ostrom. For transformation and phase, panarchy and process theorists…"
url: "https://wiki.cybernetic-futures.org/punch-and-judy/part-04-what-we-lose/ch27-what-actually-replaces-us/ch27-what-actually-replaces-us-content"
modified: "2026-11-12T15:26:25.954Z"
language: "en"
site: "Cybernetic Futures"
---

# What Might Replace the VSM

## The handovers

For mathematics, Aubin, Powers and Friston. For evidence about multi-level governance, Ostrom. For transformation and phase, panarchy and process theorists such as Tsoukas and Chia, whom Chapter 7 cites. For meaning, Bateson, Vickers and Luhmann, whom Chapter 6 introduces. For asking whose boundary is in force, Ulrich, as Chapter 16, *The Objections That Land, and What They Cost*, established and the Coda, *Judy*, will put into practice.[^ch27-handovers] That is what actually replaces this tradition, and a reader who needs one of those jobs done should reach past this shelf. In this book, the success of those handovers is nearly an argument against the tradition.

Nearly, because each handover is one-to-one, and the claim we inherited was not. Each neighboring literature takes a job the Viable System Model also claims and does it with instruments we do not have: a formalism, a comparative dataset, a phase model, a theory of meaning, a procedure for exposing whose interests a boundary serves. A literature does a job better, in the sense used here, when it does it with an instrument someone outside it can check: a formalism that computes, a coded comparative dataset, or a procedure others can run. If the tradition claimed only that these jobs matter, the handovers would settle the matter against us. The jobs would be better housed elsewhere, and the model would be a redundant index of other people's work. The argument stays open because the tradition also claimed something no single handover addresses: that a particular closed set of functions must be present together, that the losses between them can be located, and that one measure travels across the whole. The handovers failing would not establish that claim, and their succeeding does not refute it. It has to be tested on its own, in the rest of this chapter and in Part Five.

Reaching past the shelf has a price, and we are asking readers to pay it. It usually means learning a formalism rather than a vocabulary. Viability theory asks for the mathematics before it gives anything back. Active inference wants a generative model of the system before it will compute. Ostrom's design principles ask for a case corpus and the patience to code it, and Ulrich's boundary questions ask for a room in which the affected parties can actually be brought into the discussion. Each time, the practitioner gives up what the diagram supplies at once, a shared picture in an afternoon, for something the diagram cannot supply at all. For the jobs listed above we think the trade is usually worth making. It is not free, and its cost is no defense of the model.

## What the handovers do not carry

None of the handovers is a total victory. Active inference has the mathematics and, in our reading, no demonstrated route to the scale of a polity. The formalism is well posed for an organism or an agent; the step from there to an institution with many semi-autonomous parts has not been made in a way anyone has been able to check. Ostrom has the evidence base, and Chapter 14, *The Evidence: Ostrom*, found its support weakest at exactly the multi-level claim we need. The nested-enterprise principle carries the least weight in the corpus,[^ch27-nested] which is unfortunate, since nesting is the part of her work that speaks most directly to us. Panarchy names transformation and phase, and gives the cycle a vocabulary that most management thinking lacks entirely. Identifying the present phase without hindsight remains difficult, which limits what a practitioner can do with it while the system is still in motion.[^ch27-panarchy] Bateson, Vickers and Luhmann make meaning unavoidable: after them, no one can treat an organization's account of itself as a neutral report on its environment. In our reading, they do not supply a shared operational test for which meaning should govern. Ulrich makes the boundary political and gives a procedure for questioning it that anyone can run. He does not thereby diagnose the organization inside the boundary, and does not claim to.

Each of them does the job handed over better than we do, and none quietly becomes the whole instrument. The pattern holds across all five. The neighboring literatures are strong where they are specific, and they are not offering, and have not offered, the general claim the tradition made. We do not hold that against them. It marks out the space that remains.

## No mathematics, and nothing to run

The most serious of the eight defects catalogued in Part One is also the simplest to state. Variety is not measurable for any real organization, and no one has made it so. Beer said so himself. Having shown how fast variety proliferates, he asked who could use such a measure in a real situation: "Could we ever say: *this* is the variety of the chemical industry, say? Of course we could not." The use of the measure, he concluded, was "not the naive use of counting states, but of matching state generators."[^ch27-beer-count] That moved the problem without closing it. The model's rules are balances between quantities its author says cannot be counted, and he gave no procedure for checking a balance of that kind. It is a fifty-year absence at the center of a model whose central term is a measure.

Beer's own instruments were thin. The actuality, capability and potentiality triple is a ratio scheme that needs a prior decision about what counts as capability. The statistical filters of Cyberstride were built to detect departures from expectation in a small set of production indices, under severe constraints of time and hardware.[^ch27-instruments] Beer described the suite more broadly, as "statistical filtration for all homeostatic loops at all levels of recursion." Neither was ever a variety metric, and neither claimed to be. Most first instruments fall short of the ambition. What matters here is that nothing since has replaced them. The tradition has produced diagrams, case narratives and diagnostic conversations. It has not produced an instrument.

Set beside the neighbors, this bites. System dynamics gives you a model you can execute and be wrong with in public, and that is the property that matters most: a stock-and-flow model makes a commitment the world can contradict. Active inference gives you one you can compute. Aubin's viability theory gives you viability kernels, with the same word in the title and none of Beer's ancestry, an independent formalization of the question of which states permit continued existence under constraint. Powers's perceptual control theory gives you nested control loops that produce numbers, and a hierarchy arrived at from experimental work on behavior rather than from a mapping of the nervous system that Beer claimed was homomorphic[^ch27-beer-model] and never tested.[^ch27-runnable] The tradition after Beer kept the diagram and dropped the models Beer attached to System Four.[^ch27-checo]

What the diagram wins in return is real, and it should be claimed openly rather than smuggled in. It can be used in a room, on a wall, by people who will never run a model, and half of what it does is done there. A model that requires a modeler requires one who is present, funded and trusted. A diagram that a management team can argue over on a Thursday afternoon reaches situations no executable model ever will. That advantage deserves to be stated directly. It is no evidence for claims about necessity or measurement. The trouble is not claiming the advantage; it starts when we claim it in one breath and the mathematics in the next.

## The residue

What remains after the handovers is short. Five items survive. Part Five is meant to turn each into a claim against which evidence can be registered. Chapter 28, *What Would Count Against It?*, does so for the first and records that the other four are so far exposed only indirectly. Chapter 29, *The Two Cheapest Decisive Experiments*, prices the first tests, on recursion levels and on prediction, and a re-analysis of existing data that bears on the first item.

1. **The closed-set necessity claim.** Chapter 24, *Six Things Nothing Else Does*, argued that five functions are necessary for viability rather than merely common. Beer's form of the claim does not predict that an organization lacking one will fail. He holds that a persisting organization always performs the five in some form, System Two because it "is always present, or the organization would shake itself to pieces," and that a weak function produces pathology, such as the institution without a working System Four that "lives on" as a decerebrate cat.[^ch27-beer-present] Stated that way, no surviving organization can count against it, which is why Part Five tests it with the coding in Chapter 28 and with viability, not bare survival, as the outcome. No neighboring literature makes the closure claim in this form. The claim remains scarcely tested. We have cases where the functions were present and the organization survived, which is compatible with necessity and with several other explanations.
2. **The transducer.** Chapter 25, *The Transducer*, located regulatory loss at the translation across a boundary, rather than in the channel or at either end. That is a specific and unusual place to put it, and it suggests where to look when the signal that arrived is not the signal that was sent. The principle is Beer's Third Principle of Organization, which Chapter 25 cites from *The Heart of Enterprise*, and it is unmeasured. We have not established a case in which the loss was located there by measurement rather than by attribution after the fact.
3. **The algedonic architecture, with its justification.** Chapter 24 kept an explicit bypass for pain, and for a reason stronger than design preference: hierarchy attenuates precisely the signals that indict it, so a channel that reports upward through the levels being reported on cannot be relied upon. Chapter 4, *Channels Without Owners*, showed that Beer specifies how a signal qualifies, by a statistical filter on points of stability whose criteria the metasystem sets, and leaves the rest open. Who chooses those points, who may raise or contest a signal, and who audits the filter are questions the model poses and does not answer.
4. **One currency across scales.** Variety lets the same diagnosis travel between a person and a planet, and no other framework on the shelf offers a term intended to hold at both ends. Chapter 24 credits the property and concedes in the same section that it may still be metaphor dressed as measure. Until the measurement problem above is addressed, the travel across scales is a property of the vocabulary, and not demonstrably a property of the world.
5. **Freedom as a quantity commensurable with the diagnosis.** Chapter 26, *Freedom as a Quantity*, followed Beer in putting the room left to act after institutional attenuation into the same units as the diagnosis, instead of treating it as a separate normative concern. In *The Heart of Enterprise* he declared it "in principle a computable function of systemic purpose as perceived"; nobody has computed it.[^ch27-beer-freedom] That is what makes the tradition's political claim continuous with its technical one. *Residual variety* is our name for the quantity, in the sense defined in Chapter 26 and not in Beer's Third Axiom sense.

Part Two identified no other home for this residue. That does not prove no home exists anywhere on the shelf. It is the remainder our comparative search must now put at risk.

Suppose the residue is empty: the closed set only a checklist, transduction only another name for boundary work, the algedonic bypass no more than escalation, variety no common currency, and freedom no quantity. Then the honest conclusion is that the tradition should be wound up. We would rather find that out than preserve it by lowering the test.

[^ch27-instruments]: Stafford Beer, *Brain of the Firm*, 2nd ed. (John Wiley & Sons, 1981), 163–64, for actuality, capability and potentiality, and 261–64 and 270 for the statistical filters of Cyberstride (Beer's description at 270, checked 29 September 2026 in the calibre PDF); Eden Medina, *Cybernetic Revolutionaries: Technology and Politics in Allende's Chile* (MIT Press, 2011), 70–71, 78–80, 103 and 183–84, for what Cyberstride was, what it ran on and what it produced.

[^ch27-handovers]: Sources for each handover are in the owning chapters: Bateson, Vickers and Luhmann, Chapter 6; Ostrom, Chapter 14; panarchy, Chapters 5 and 11; process theory (Tsoukas and Chia), Chapter 7; Ulrich, Chapter 16. For the mathematics, see the note below.

[^ch27-beer-model]: "Models are more than analogies: they are meant to disclose the key structure of the system under study" (Beer, *Brain of the Firm*, 2nd ed., 75); "There can be no interest in analogy at this point" (87). In *Decision and Control* (John Wiley & Sons, 1966), 113, "A scientific model is a homomorphism on to which two different situations are mapped." Checked 29 September 2026 in the calibre PDFs. That the homomorphism was never tested is our reading.

[^ch27-checo]: Beer's Checo and the quantified System Four model (*Brain of the Firm*, ch. 13) are a lost part of the program, not an absence. The Checo team's experiments were undertaken "in the spirit of a prolegomenon to a full-bodied System Four," and 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" (Beer, *Brain of the Firm*, 2nd ed., 267–68). The System Four model of chapter 13 has five control parameters, A, B, C, X and Y, which a management group would set with knobs in the control center (187–97). Checked 29 September 2026 in the calibre PDF.

[^ch27-beer-count]: Stafford Beer, *The Heart of Enterprise* (John Wiley & Sons, 1979), 86–87.

[^ch27-beer-present]: Stafford Beer, *Diagnosing the System for Organizations* (John Wiley & Sons, 1985), 66; Beer, *Heart of Enterprise*, 143 ("System Four exists — it has to: but it is disseminated") and 265.

[^ch27-beer-freedom]: Beer, *Heart of Enterprise*, 158.

[^ch27-runnable]: Jay W. Forrester, *Industrial Dynamics* (MIT Press, 1961), and John D. Sterman, *Business Dynamics: Systems Thinking and Modeling for a Complex World* (Irwin/McGraw-Hill, 2000), for system dynamics; Jean-Pierre Aubin, *Viability Theory* (Birkhäuser, 1991); William T. Powers, *Behavior: The Control of Perception* (Aldine, 1973). For active inference, Karl Friston, Thomas FitzGerald, Francesco Rigoli, Philipp Schwartenbeck, and Giovanni Pezzulo, "Active Inference: A Process Theory," *Neural Computation* 29, no. 1 (2017): 1–49, https://doi.org/10.1162/NECO_a_00912, and Thomas Parr, Giovanni Pezzulo, and Karl J. Friston, *Active Inference: The Free Energy Principle in Mind, Brain, and Behavior* (MIT Press, 2022), as cited in Chapter 12. The claim is about the existence of an executable model, not about the quality of any particular fit.

[^ch27-nested]: Michael Cox, Gwen Arnold, and Sergio Villamayor-Tomás, "A Review of Design Principles for Community-Based Natural Resource Management," *Ecology and Society* 15, no. 4 (2010): art. 38, https://doi.org/10.5751/ES-03704-150438, table 3: nested enterprises has the lowest support ratio of the eleven principles and is the only one whose association with success does not reach significance at the five percent level. Chapter 14 gives the figures and the limits of reading them as evidence about recursion.

[^ch27-panarchy]: Chapter 5, *The Ladder Without a Ruler*, on phases that are easier to identify after the fact than in the present tense; Chapter 11, *The Nesting: Simon, Koestler, Jaques*, on the 2022 systematic review by Rocha and colleagues ("a lack of methods to measure or observe it in real social-ecological systems") and on zu Castell and Schrenk's 2020 method for identifying the phases, which depends on the analyst's choice of parameters.