The Enemies We Actually Have

The claim

Count the people actually attacking the Viable System Model, and the list is short. This tradition believes itself embattled. It is unattended, which is worse.

The standard is Chapter 16’s, unchanged. An objection lands 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. By that standard three objections land, Ulrich’s, Hayek’s and Cooper’s, and Chapter 16 has already taken them. Nothing here adds to that list. What remains is everything else commonly treated as opposition, sorted by what it costs us.

The rival ontology

Dave Snowden is the most visible living critic of this model in the market where most of its practitioners work. His position, stated in December 2025, distinguishes Bateson-style cybernetics from its use for the VSM and its relatives, and holds that had Beer known complexity science he would have created something other than the VSM.^[1]

Taken seriously, the claim is an ontological objection: much of organizational life is complex in Cynefin’s sense, where “cause and effect are only coherent in retrospect and do not repeat,” and a model that maps functions and channels is aimed at a kind of object that is not there.

What does it cost us? Commercially, a great deal. Cynefin is embedded in the same DACH consulting market this tradition sells into, and a practitioner who loses that argument in a room loses the engagement. Epistemically, nothing yet. No test is offered. Cynefin’s authors say so themselves: its “value is not so much in logical arguments or empirical verifications as in its effect on the sense-making and decision-making capabilities of those who use it.”^[2] There is no observation that would settle whether organizations are the kind of object the VSM diagnoses or the kind Cynefin describes, and neither side has proposed one. Two ontologies facing each other without a discriminating experiment: that is the condition this book charges the VSM with, now held by both parties.

By the standard, it does not land. It is still the most interesting place to look for an adversary who could be made to land, and the way to make that happen is to propose the experiment ourselves.

The sweeping dismissals

An older literature rejects systems thinking wholesale. Ida Hoos holds that systems analysis is impervious to criticism; David Berlinski’s judgment is that the subject is a sham; Robert Lilienfeld treats systems theory as ideology.^[3]

Our field usually cites these as evidence of hostility. They are evidence of something else. None of them is about the Viable System Model. They address systems analysis and general systems theory in the 1970s, they predate most of what this tradition has produced, and their charge, that the subject is untestable and immune to criticism, is one this book makes in more detail and with more evidence than any of them assembled.

They cost us a reputation more than an argument, and in the specific respect they name, the reputation is deserved. The right response is to note that the general charge was made fifty years ago, that this tradition never answered it, and that the answer was always going to have to take the form of a test, never of a reply.

They do not land, for the pedestrian reason that they do not engage the model. Nor should they serve as a foil: a tradition that defends itself by pointing at its weakest critics is doing what Chapter 17 describes.

Friendly fire on the foundation

The most substantial recent criticism of anything this tradition rests on comes from people who have never heard of Beer.

The Conant–Ashby theorem, that every good regulator of a system must be a model of that system, supplies the formal authority VSM teaching borrows when it wants to be more than a heuristic. Chapter 3 deals with what the theorem says and what we quote instead. Since then the theorem’s own standing has come under attack. Iqbal Thobani argues in Synthese that the notion of an internal model in these principles is far too permissive, and that they do not provide evidence, let alone a proof, that internal models are necessary. Nathaniel Virgo and colleagues, in a preprint, hold that the argument does not strictly succeed even in its own terms. John Baez has argued that the proof is too easy to be of practical help, and a further critique holds that the notion of model in it is trivial and the proof assumes full state information.^[4]

The evidential weight of these differs sharply and should not be flattened. Thobani is peer-reviewed and citable. The rest are a preprint, a blog post and a forum post: legitimate to read, not legitimate to cite as established.

By Chapter 16’s standard this lands, and it lands where the tradition has no defense in place. The model leans on the theorem and cannot repair it, since it is not ours to repair, and no amendment has been made. The cost is specific: wherever VSM teaching says that requisite variety or the good regulator theorem proves something about organizations, that claim is now contested at the source, and we have not noticed.

It is also the one piece of good news in this part of the inventory. A tradition whose foundational borrowings are being examined by analytic philosophers has, for the first time in years, an interlocutor doing rigorous work on something we care about.

The silence

The adversary that costs most is not a person.

Mainstream organization science does not reject the VSM. It engages Ashby, including through Boisot and McKelvey in the Academy of Management Review, and barely engages Beer at all. The VSM literature is concentrated in Kybernetes, Systems Research and Behavioral Science, Systemic Practice and Action Research and the Journal of the Operational Research Society; a bibliometric review finds limited diffusion in the managerial literature.^[5]

Hence the absence of any outside pressure to test. There is no gatekeeper. Nobody’s funding, license, accreditation or reimbursement depends on whether the model is right, so nobody has ever been required to find out.

The contrast comes from a constructivist community that was in our position and no longer is. German systemic therapy holds a constructivist self-understanding at least as strong as second-order cybernetics, and it now has an evidence base: two professional associations commissioned a review of more than eighty randomized trials, scientific recognition followed in 2008, an IQWiG report appeared in 2017, and the Federal Joint Committee confirmed benefit for adults in 2018, with statutory coverage from 2020 and for children and adolescents in 2024.^[6] The community did not change its epistemology to produce that evidence. An institution that controlled payment demanded it.

Organizational consulting on the same ground has only recently begun commissioning evidence at all, and its external critics describe systemic coaching as an immunization strategy against rational critique, a charge that will sound familiar by this point in the book.^[7]

Beer had seen part of this. His proposal for a standing team of “organizational pathologists,” to perform post-mortems on organizations “viable no more” (Chapter 17), describes the work an evidence register would collect.^[8] It does not say who would require the post-mortems or pay for them, and that is the piece still missing.

The inference is ours and should be marked as ours: an evidence register without an institutional stake behind it will stay empty. This part hands that finding to Chapter 30, and it changes what the bill is for. The expensive item is not the experiments. It is arranging for somebody to require them.

What this chapter does not establish

It does not establish that the VSM has no serious critics. It has three, and they are in Chapter 16.

Nor does it establish that Snowden is wrong about organizations, or that the complexity critique can be dismissed. It establishes that no test currently separates the two positions, an embarrassment both sides share.

It does not establish that mainstream organization science is right to ignore this model. Its silence delivers no verdict. It shows an absence of attention, and absences of attention are not evidence about anything except attention.

What it establishes is that this tradition has been shadow-boxing. The opposition it imagines (hostile, numerous, ideological) is thin and mostly aimed elsewhere. What has actually kept the model unexamined is that nothing outside it has ever needed the answer.

Receipt — buy a gatekeeper. Approach three bodies that could make a declared failure condition a requirement rather than a courtesy, and secure one commitment in writing: (a) an indexed journal — Systems Research and Behavioral Science or Kybernetes — to adopt a submission rule that any article reporting a VSM application must state, in a structured abstract field, what result would have counted against the model in that case; (b) a certifying body — Metaphorum, or SCiO, whose Competency Framework route already assesses candidates through a two-hour interview structured as Situation, Objectives, Actions, Results and Aftermath/Learning — to require that the Aftermath/Learning component include one engagement that did not work and what the practitioner concluded from it, and that a submitted diagnosis name its own disconfirming observation. The hook exists; nothing currently hangs on it. Read as published, SCiO’s accreditation criteria are a list of positive capabilities — able to use, can use, has developed — with no requirement anywhere to report an unsuccessful engagement, state a method’s limits, or say what would have counted against a diagnosis;^[9] © one public-sector or foundation funder in the DACH region that commissions organisational design work, to require the same in its tender conditions. Price: three conversations and one drafted rule per body, perhaps ten days of someone’s time in total, most of it spent finding the right person. This is the cheapest item in the entire programme and the one that determines whether any of the others produce data. If all three decline, the register proposed in Chapter 30 should be built anyway and should be expected to stay empty, and the book should say so rather than pretend otherwise: a programme whose findings nobody is obliged to read is a programme that will be read by nobody.

Handover

This part has said who was standing around the ring. Part Five says what to do in it, and Chapter 30 now has one more line on the bill.