The Bill: A Book Plus a Live Register
A bounded audit here, an analyst-week there, a review paper, a facilitated session, an archive fee. Each receipt in this book, read on its own, is small enough to seem costless. Read together they are a program, and a program has to be paid for by people with names, calendars and budgets. Until the items stand beside one another, we can agree with every one of them and commission none. Adding them up is meant to make that harder.
The receipts come from the whole book. Part One set out the defects we inherited and never tested, and closed on the charge that nothing in the model’s core, the necessity of the five functions and the recursion claim, has been stated in a form that could be disconfirmed. Part Two credited the neighboring literatures that arrived at the same components without us. Part Four narrowed the claim to what we can defend and named the residue that has no other home. Part Five has been turning those charges into work: experiments with a price on them, each with a statement of what a negative result would establish. What remains is the total.
The bill has three parts, and only the first is money. The first is the work: what would have to be done, and what it would cost at minimum. The second is the people: which competences the work requires, most of which are not ours, and some of whose holders have every reason to decline. The third is the standing record: the obligation to keep the account open after the book is printed, so that items close on evidence rather than on fatigue.
Why the complete list is not printed here
The complete research program is not here. These pages print the bill earned by Punch and Judy; the wider program belongs in a maintained, dated and versioned companion register, the live register of this chapter’s title.
The accounting made that decision, not convenience. One of this book’s sharpest findings, in Chapters 19 and 29, is that the field never built and maintained a record of what it did. Beer asked for one: a small team of “organizational pathologists” to study organizations that are “viable no more,” and thresholds for his identity exercise “to exceed which will predict collapse.”[1] A numbered list, fixed on the day of typesetting and never touched again, would repeat in the answer the failure named in the question. A register is not merely a publishing format. It is the thing the accounting demands, and building it is the first item on it.
There are three further reasons. The accounting concerns more than fifty years that have already happened, and it does not decay. An agenda is a claim about what is not yet known, and it decays the moment somebody does the work, so the two do not belong in one object. If this book succeeds, problems close; a printed chapter cannot record its own success, while a register can show a problem closed, by whom, and with what result. And a register is the only part of this book a reader can act on rather than assent to. A problem can be claimed, a case added, a verdict contested, and none of that is available on a page.
One consequence, stated once: this book never refers to a research problem by number. Numbers belong to the register, and they will change as items are added, closed, split and merged. The printed bill names the claim and the work. The register is kept by Krishan Mathis and revised quarterly, and each version is deposited on Zenodo with its own DOI, so that a citation of the register names a fixed state of it.
What would make the register worthless
Splitting the bill from the register helps only if the register meets a standard the tradition has not held itself to before. As already said in Chapter 29, the worthless version is a case-study collection, which we already have, written by people with an interest in the outcome. The discriminators are three, and strict: success criteria registered before the intervention;[2] failures included; and the register open, so that a hostile reader can recount.
None of the three is expensive. Each is uncomfortable, and that is what makes them work. Registration removes the option of discovering the criterion after the result. Including failures removes the selection that made the existing literature persuasive. Openness removes our privilege of being the only party able to check the arithmetic. Anything short of that reproduces the existing literature at greater length. These conditions apply to every item below. The conditions specific to individual experiments were set out in Chapter 29 and are not repeated here.
The bill, ordered by minimum viable cost
The costs are bands rather than quotations: desk means weeks of documentary work; small study means one organization or a bounded corpus; program means multiple cases, specialist collaboration or longitudinal access. Ordering by cost makes refusal visible. The cheapest items do not wait for the expensive ones.
| Claim at risk | Experiment | Evidence required | Minimum viable cost | Result that counts against us | Threatened assumption | Who could do it | Planned companion dossier |
|---|---|---|---|---|---|---|---|
| The mapping taught is not the mapping Beer wrote | Audit Beer’s System Two–Four assignments against dated teaching sources, recording Beer’s own conflicting assignments as conflicts[3] | Paginated Beer text; captured primers; independent extraction | Desk | The assignments materially match | The tradition drifted from its source | History-of-cybernetics researcher plus neuroanatomist | Neurophysiological remapping |
| The good-regulator theorem is cited beyond its proof | Blind-code a reproducible citation sample | Search protocol, corpus, coding rules, second coder; Beer’s own restatements as a separate stratum[4] | Desk | Citations consistently preserve the criterion and concessions | Citation practice substitutes authority for reading | Information theorist and research assistant | Foundations and citation audit |
| System Four has been reduced to scanning | Search current teaching material against registered functional criteria; the Beer side is answered[5] | Editions, search surface, quotations and dated captures | Desk | Teaching retains the wider function | A central function disappeared in transmission | VSM scholar and independent textual coder | System Four dependency analysis |
| The six components have independent lineages, and the related entries do not stand in for recursion | Completeness and priority audit: earliest independent statement of each component, citations checked in both directions, and for each entry whether it reached Beer’s object or a related one (Chapter 10) | Six bounded searches; preregistered evidence table with a related-object column; external spot-checks | Desk | One or more components has no independent lineage | The model’s components are non-arbitrary | History-of-science researcher and an external spot-checker | New — added 29 September 2026 |
| The method literature dropped Beer’s freedom quantity | Systematic survey of the VSM method literature (Chapter 26): stated inclusion rule; coding of freedom and management vocabulary and of where authority over the model is placed; pilot from Espejo and Reyes (2011), then Hoverstadt and Martin Pfiffner (2020, 2022) | Inclusion rule, corpus, coding scheme, reported counts, second coder | Desk | The method literature keeps freedom as a quantity, or asks whose residual variety an intervention should increase | The method descends from the handbook’s narrowing | VSM scholar plus a second coder | New — added 29 September 2026 |
| An algedonic channel performs a distinct safety function | Red-team the mechanism against modern incident and escalation practice | Failure cases, escalation traces, comparator mechanisms | Desk / small study | No discriminable function or advantage remains | The channel is structurally necessary rather than metaphorical | Safety researcher and incident-response team | Algedonic red team |
| Recursion levels are discovered rather than drawn | Compare blind VSM diagnosis with rate separation[6] and time-span measures | Interaction matrix, prior diagnosis, registered protocol | Small study | Levels fail to track either independent measure | The recursion ladder has an observer-independent rung | Organizational network analyst, VSM analyst, requisite-organization practitioner | Recursion measurement protocol |
| VSM diagnosis predicts something alternatives miss | Reconstruct pre-outcome predictions for ten interventions and compare them with a baseline and a neighbor | Original engagement documents, outcomes, failures, independent coding | Small study | VSM adds no predictive or diagnostic value | The framework is more than a redescription language | Evidence-based-management team and independent coders | Intervention register pilot |
| The embodied observer can be specified without smuggling in a view from nowhere | Write the 4E reconstruction and test its operational distinctions on one diagnosis | Sensorimotor boundary decisions, observer accounts, rival formulation | Small study | The reconstruction changes no diagnosis or prediction | Embodiment repairs the observer problem | 4E cognition and organizational-cybernetics collaborators | Embodied observer paper |
| The model handles change of kind, not only change in environment | Specify a discriminating morphogenesis case against the current architecture, starting with Beer’s Chilean redesign of the recursions[7] | Longitudinal transformation record and rival accounts | Small study / program | Existing System Four and recursion machinery explains the kind-change without residue | Morphogenesis is missing from the architecture | Developmental-systems and organizational-change researchers | Kind-change study |
| The six Part Four residues are unique to the VSM | Adversarial comparison with named neighboring theories | Matched concepts, operational consequences, specialists from both sides | Program | A neighbor supplies the same work with equal or greater precision | The claimed remainder belongs uniquely to this tradition | Cross-theory panel with adversarial review | Six-residue comparison |
| Each of the five functions carries weight in viability that the others cannot supply | Formative re-analysis of Michael D. Pfiffner’s 135-case dataset: re-specify the measurement model, remove each function block in turn, compare against a rule published beforehand | Existing dataset (2017, N=135); the author’s participation; a registered decision rule, including how variance shared between blocks is read | Desk / small study | A function block can be removed without loss of discrimination | The five functions are each necessary (tested in Beer’s sense only where absence is coded as in Chapter 28) | Structural-equation specialist with the original author | New — added 14 September 2026 |
| Viability requires the five functions | Seek deviant cases using an open coding protocol and configural analysis | Multi-case register; the coding in Chapter 28 (official, disseminated or absent, by registered indicators); viability outcomes independent of that coding, not bare persistence | Program | Viability kept with one function coded absent, or lost from an unseen cause with all five coded present | The configuration states a necessity of viable organization | Comparative-organization and QCA researchers | Structural disconfirmation register |
| A single price per commodity lacks requisite variety for some regulatory tasks | Requisite-variety analysis of Hayek’s price mechanism: does one scalar per commodity have requisite variety, and a short enough settling time, for each regulatory task (Chapter 16) | Classes of disturbance, a model of the price signal, settling-time evidence | Not yet priced | One scalar per commodity suffices in variety and settling time for every task examined | Requisite variety answers Hayek’s objection | Economist with a control theorist | New — added 29 September 2026 |
The two necessity rows overlap. The formative re-analysis of Pfiffner’s dataset is attempted first; the deviant-case search is what remains if the dataset cannot be obtained.
The table prices bands, not quotations. A band tells a reader which items can be refused on grounds of cost and which cannot, and it is fit for nothing more. It follows that no total has been struck. The receipts in the earlier chapters are stated in mixed units, and the bands here are a third unit again, so the components cannot be summed without inventing an exchange rate between them. We will not invent one. The bill is a list of components whose sum has not been computed, and printing a figure we cannot defend would be a worse fault than printing none.
Whether any of this happens will be decided in the column that names who could do it. Most of the competences listed there are not held inside this tradition. We cannot commission a neuroanatomist, a safety researcher, an information theorist or a comparative-methods team. We can only ask, and we should expect some to decline and others to accept and then produce results we would rather not have. That is the arrangement we are proposing. The cheap items at the top of the table are the ones we can start without anyone’s permission, and starting there is the only evidence available that the offer is made in good faith.
The state of the count
The same discipline applies to what we currently know about our own exposure. Below are the figures we have, with their status stated before they are read.
:::: {.as-of} As of August 2026
Both searches reported here are dated, and both will be wrong before this book is old. The arXiv full-corpus search returned three all-time hits for “viable system model” and none for either “management cybernetics” or “organizational cybernetics”. Kybernetes had indexed four hundred and eighty-three papers since January 2024, of which none addressed neuroscience, predictive processing, active inference or allostasis.
A count of this kind is a photograph. What gives it value is the date on it, and the register is where the next photograph goes. ::::
What we owe
The obligations that follow from the bill are narrower than the bill. We owe the work at the cheap end, because cost gives us no excuse for leaving it undone. We owe the three discriminators to anyone we ask to take an item, since without them the register becomes another literature nobody outside the tradition has reason to read. We owe the register an owner and a cadence, and it has both. And we owe the results that go against us the same treatment as those that do not. Of the four, that obligation is the easiest to meet in principle and the easiest to avoid in practice.
Every item in the bill is offered on those terms. Take one, try to break it, and report what broke, including to the people who would prefer it had held. The components are listed above; the total has not been struck; the register is where the account continues after this page.
Stafford Beer, Think Before You Think, ed. David Whittaker (Wavestone Press, 2009), 150; Stafford Beer, The Heart of Enterprise (John Wiley & Sons, 1979), 432. ↩︎
On registration before the result, see the note on Nosek et al. (2018) in Chapter 28. ↩︎
Stafford Beer, Brain of the Firm, 2nd ed. (John Wiley & Sons, 1981), 132, where the sympathetic loop sits within System Three, against 176, “the sympathetic (System Two) circuits”; see also 129. ↩︎
Beer, Heart of Enterprise, 235 and 352, as quoted in Chapter 28. ↩︎
System Four must “contain a model of the viable system of which it is the System Four” (Beer, Heart of Enterprise, 235), raises “the issue of initiative” (237), provides “self-awareness” (Stafford Beer, Diagnosing the System for Organizations (John Wiley & Sons, 1985), 115), and runs a loop “anchored in THE UNKNOWN FUTURE” (119). ↩︎
Rate separation as a criterion of level is Herbert A. Simon’s, in “The Architecture of Complexity,” Proceedings of the American Philosophical Society 106, no. 6 (1962): 467–82; Chapter 11 discusses it and Chapter 29 sets out the design. ↩︎
Beer, Brain of the Firm, 322–26; on 323 he concludes that “two levels of recursion were mixed together in that preliminary model.” ↩︎