A Machine That Never Became One

In 1963 Magoroh Maruyama pointed out that cybernetics had studied deviation-counteracting processes predominantly and paid less attention to deviation-amplifying ones, and he proposed a second cybernetics for the latter.^[1] Beer’s model was still forming when Maruyama wrote. Writing in 1979, Beer said that the neurocybernetic model “took over twenty years to develop.”^[2]

The VSM that came out of those twenty years describes the organization as a machine already assembled. Its five systems explain how an existing identity is maintained under disturbance: variety attenuated and amplified across an established boundary, operations coordinated without central congestion, and a purpose that converges between what is declared and what the system delivers.^[3] They do not explain how a system becomes viable in the first place, how it passes through collapse as a phase rather than as a failure, or how it changes its kind. We have used the model in all three situations anyway, and we have rarely said what it cannot see there.

What the five systems are built to explain

Take the ordinary diagnostic occasion. An organization exists, its participants can name its identity, and it is handling some disturbance badly. The model gives us leverage because identity is fixed for the duration of the analysis. System Five’s normative closure, System Three’s resource bargain, System Two’s damping and System One’s autonomy within constraint are each defined relative to something the organization already is. Nothing about this is hidden; it is the plain content of the construction. The trouble lies in the range of cases to which we have applied it without qualification.

The model is homeostatic

Maruyama’s date matters to how we tell the story, though nothing in the argument depends on it. Contemporaneous or not, the VSM as we teach and apply it is a first-cybernetics construction, though Beer’s own texts are not first-order in this sense: they make the system an observer’s act and closure a self-referential process.^[4] It explains how a system holds together under perturbation.

It has no account of morphogenesis: of how a system becomes something it was not, of collapse as a phase rather than a failure, of the generative disorder from which new forms arrive. Panarchy has that account.^[5] The adaptive cycle’s release and reorganization phases have no counterpart among the five systems. No subsystem has the function of letting structure come apart, and nothing represents the interval in which a form is between kinds. System Four is where one might expect the capacity to sit. Beer gives it more than we teach: it contains environmental variety and also generates it, holds a model of the whole that includes itself, and is asked how it will “invent the future.”^[6] It is still not given the task of changing the system’s kind. Beer’s viable system undergoes “every kind of change. But there is no alteration.”^[7] Chapter 2, System Four Is Not a Pair of Eyes, establishes that boundary in detail, and we do not re-argue it here. Related asymmetries appear elsewhere in Part One. Chapter 4, Channels Without Owners, deals with revolt and remember and the couplings panarchy specifies and we never did; Chapter 5, The Ladder Without a Ruler, deals with temporal nesting. Across those chapters the pattern is consistent enough to name as one defect rather than three.

An anatomy applied to things that are becoming

Process theory and panarchy handle dissolution and transformation better than the VSM handles either.^[8] The sentence is easy to overread. An anatomy of maintenance is not defective for being an anatomy of maintenance. A model that explains how identity is held under disturbance has done something difficult and useful, and it is not answerable for questions it never set out to answer. Frameworks built for other phases do not weaken Beer’s construction.

The defect is ours. We have taken an anatomy of the assembled organism and applied it, without qualification, to organizations that are forming, dissolving or transforming: startups before they have an identity to protect, institutions in the middle of dismantling themselves, bodies that came out of an intervention as a different kind of thing than they went in. In those engagements the model still produces a diagnosis. It reports missing channels, a weak System Two, System Three overreach. It does not report that the object under examination was not the kind of object the method addresses. We have not written that qualification into our practice, our teaching or our reports, and Chapter 9, Why None of This Ever Cost Us Anything, explains why the omission has been comfortable to leave in place.

Absorbing the vocabulary is not the same as making the connection

Beer took autopoiesis into The Heart of Enterprise himself and made it a diagnostic criterion. The time a system spends on its own self-production should be “exactly the time required to maintain System One,” and Systems Two to Five “should exhibit no internal autopoietic behaviour whatever”; past that lies the pathology he calls the Establishment.^[9] That is a use of the concept, more than a borrowing of the word. It concerns how much a system may spend producing itself, and says nothing about how a system becomes another kind of system. The wider field later took von Foerster and Maturana into its theoretical vocabulary.^[10] Second-order formulations and the language of autopoiesis are now unremarkable in our discourse, and they do useful work in situating the tradition. A vocabulary, though, does not turn a diagnostic procedure that records none of the analyst’s own choices into an account of autopoiesis,^[11] and the presence of the terms cannot stand in for the demonstration.

A rigorous connection would have to say two things. The first is how the diagnostic observer takes part in producing the distinctions being diagnosed: how the consultant’s choice of system-in-focus, recursion level and identity statement enters into what is then found. Chapter 6, The Brain in the Vat, sets out why that participation is not a marginal correction. The second is how a change of organizational kind appears in the method: what the diagnostician would see, in the model’s own terms, when the object stops being one sort of system and becomes another. On the first, Beer went part of the way. He stated the observer’s role in 1966, before the second-order vocabulary arrived, and his workbook compares members’ perceptions: check the First Axiom “as System Three sees it; as System One sees it … There are ALWAYS surprises.”^[12] He gives no procedure that exposes the consultant’s own choices of focus and level to a second reader. Neither that procedure nor a representation of kind-change has been shown in the material we have examined here. Nor have we run and reported a dated search of the literature for work that might already have shown it. A positive result would need more than a paper citing Maturana alongside Beer. It would need a procedure that makes the observer’s contribution explicit, or a representation of kind-change in the five systems, applied to a case and legible to a reader who did not write it.

The consequence, stated exactly

For any diagnosis of an organization that is forming, dissolving or transforming, the limitation is exposed: a model of survival cannot, by itself, say what is worth becoming. The claim is narrower than it may sound. It does not say that viability is the wrong thing to care about, and it is not an argument about values; Part Four takes up the related question of what survival is for, and Chapter 23, The Narrow Claim, sets the boundary of what we are prepared to assert. The charge here concerns representation. An organization can be in states the model has no way to depict, and we have issued diagnoses in those states without saying so. Chapter 27, What Might Replace the VSM, considers what a client is entitled to expect in them.

Beer’s own record contains one such state. After October 1972 in Chile, he wrote, “there seems to be a resurgence in the self-organizing capability of communities to promote themselves as viable systems.” He handled it by redrawing the recursions after the event, with a new level and three new Systems One.^[13] The method caught up with the change; it did not represent the change while it was happening. Beer also put the identity question himself: after adaptation, will the system “still be able to recognize itself?”^[14]

Receipt — specify kind-change. The way to settle this is to attempt the specification and see what it costs. Write one paper stating what System Four would have to observe, represent and alter in order to anticipate a change in the system’s kind rather than a change in its environment. Price: one conceptual paper tested against at least one documented transformation, beginning with Beer’s own redrawing of the Chilean recursions after October 1972. If no operational distinction can be made, the negative result is that the extension adds vocabulary but no diagnostic power — which is worth knowing, and worth publishing, because it would tell us that the limitation is structural rather than an unfinished piece of work. Either outcome ends the present situation, in which we neither have the capacity nor concede that we lack it. The Coda, Judy, returns to what we owe the organizations that were becoming something while we were describing them as though they already were.