---
title: "Ch20 usefulness instead of truth content"
description: "On the second page of Diagnosing the System for Organizations (1985), Beer sets three short lines apart from his text: “So remember: / a model is neither true…"
url: "https://wiki.cybernetic-futures.org/punch-and-judy/part-03-false-friends-and-enemies/ch20-usefulness-instead-of-truth/ch20-usefulness-instead-of-truth-content"
modified: "2026-11-12T15:26:25.926Z"
language: "en"
site: "Cybernetic Futures"
---

# Usefulness Instead of Truth

## The claim

On the second page of *Diagnosing the System for Organizations* (1985), Beer sets three short lines apart from his text: "So remember: / a model is neither true nor false: / it is more or less useful."[^ch20-beer-dts] It is the most respectable sentence in the systems literature, and it is not wrong. It is also the most damaging sentence in our inheritance, and we did not borrow it. It is the founder's, and it was not new in 1985. In *Decision and Control* (1966) a methodology is "neither right nor wrong, but only more or less useful," and later in *Diagnosing the System* the model has no "correct" interpretation, only "more or less useful interpretations."[^ch20-beer-variants]

In 1966 Beer also said how the usefulness of a model is to be measured: "by straightforward testing of the predictive value of the scientific model. If this predictive value is low, we had better start again."[^ch20-beer-dc] The sentence came with a way of losing. The tradition kept the sentence and dropped the test.

Our neighbors reached the same sentence by their own routes, and the most careful of them built the machinery that makes it safe. The machinery is the subject here, and so is what the sentence becomes when it travels without it.

## The neighbour who earned it

System Dynamics has the most developed validation philosophy of any tradition adjacent to ours, and it arrived at the usefulness position by argument, not by convenience.

Forrester and Senge state that there is no method for proving a model correct, and that confidence rather than proof is the proper criterion.[^ch20-forrester] Barlas and Carpenter engage Popper directly, including the Duhem problem (a failed prediction never tells you which of your assumptions to give up), and conclude that models lie on a continuum of usefulness.[^ch20-barlas-carpenter] Barlas gives the operational form: right behavior for the right reasons.[^ch20-barlas] Sterman rejects validation in the sense of establishing truthfulness.[^ch20-sterman]

So far this is our sentence, said better. The next part is the one we did not take.

Having given up proof, System Dynamics built structured test batteries in its place. Barlas separates structure tests from behavior tests. The first ask whether the model's mechanisms correspond to the mechanisms in the system, independently of whether the output looks right; the second ask whether the output matches, judged against patterns and not point predictions. Forrester and Senge had already named them in 1979: structure-verification, parameter-verification, extreme-conditions, boundary-adequacy and dimensional-consistency tests of structure, and behavior-reproduction, behavior-anomaly and behavior-sensitivity tests of behavior, among others, each with a stated procedure.[^ch20-forrester-tests] A model that fails them is in trouble, however useful anyone found it.

"More or less useful" in that tradition is the conclusion of a program with named ways of losing. In ours it began with one, Beer's prediction test. Beer stated the test; neither he nor anyone after him turned it into a battery for the Viable System Model, with named procedures and a threshold for "low." We kept the epistemology, skipped the battery, let the founder's test lapse, and were left with a sentence that sounds like methodological sophistication and works as an exemption.

The right response is to pay for the position, and the structure/behavior split is directly portable. A VSM diagnosis makes structural claims (this function is present, this channel carries this variety, this unit is a System One at this level) and behavioral claims (this pathology will produce this oscillation). Barlas's distinction says the two must be tested separately, and that a diagnosis passing the second while failing the first is not corroborated. The node, edge and subgraph typing in the companion working paper, *A Stress Test for the Viable System Model*, connects to this directly,[^ch20-stress-test] and Schwaninger works in both traditions, so the bridge is a short one to build.

## The same move, three more times

Replacing the falsification criterion is the field's standard maneuver, well beyond System Dynamics, and almost nobody in the neighborhood rejects testing outright.

The Hull school judges a methodology by fit to the problem context, by pluralism, and by practical and emancipatory results, not as a hypothesis. Michael Jackson's position is that systems thinking should learn which approaches bring about beneficial change in a particular context; Gerald Midgley holds that theories are more or less useful depending on the purposes of intervention.[^ch20-hull] Within that frame the VSM is the functionalist lens, and the metamethodology does not ask whether it is *right* about organizations. Jackson himself does, from the functionalist side: an empirical research agenda around the VSM is, he writes, "certainly essential, and it is disappointing that so little has been done."[^ch20-jackson-agenda]

Werner Ulrich replaces the criterion with boundary critique, deliberately and with an argument: his charge against critical rationalism is that it has not really overcome the empirical foundationalism of earlier positivist research practice.[^ch20-ulrich] Chapter 16 takes his objection at full strength and concedes what it establishes. The concern here is narrower. Boundary critique is a criterion for the legitimacy of an intervention, and it cannot settle whether there are five subsystems. An empirical program does not answer Ulrich, Ulrich does not answer the empirical question, and this book should say both.

Peter Checkland is the one who ought to be named an ally rather than a false friend. He and Holwell accept that testing hypotheses to destruction carries the strongest truth claim in natural science, and for social inquiry they substitute recoverability: a research process based on a prior declaration of the epistemology, such that another person can retrace what was done and why.[^ch20-checkland] Prior declaration is exactly what the receipt protocol asks for. Recoverability is a weaker obligation than a test, but a real one. This book can adopt it without amendment, and it is the model for what the other replacements should look like.

Across all four the pattern is the same. Each substitute criterion is adequate for instrumental claims: whether a diagnosis helps, whether an intervention is legitimate, whether an account can be retraced. None can decide a constitutive claim: that the five functions are necessary, that they are sufficient, that System Five corresponds to a cortex. Those claims are in our teaching and in our books. Usefulness cannot assess them, because a false model can be useful, and frequently is.

Beer's 1966 account shows how a constitutive claim escapes. A scientific model, he wrote, "actually defines the extent to which [two situations] are structurally identical," and if it has ignored as irrelevant something that is in fact relevant, "the model will lose in utility, but it cannot lose in validity. This is its peculiar strength."[^ch20-beer-dc] The claim of structural identity is secured by construction, and prediction is left as its only exposure. Drop the prediction test, and nothing is exposed.

## The asymmetry that makes this a friendship

A complaint that the neighbors refuse to test anything would be easy to write. It would also be false, and it would miss what is actually happening.

System Dynamics runs test batteries. The Hull school evaluates interventions with stakeholders. Researchers on problem-structuring methods have built evaluation frameworks and are candid that claims for these methods usually rest on single case studies of intervention, and that such modeling yields insights rather than testable results.[^ch20-psm] That candor is more than this tradition has managed. Ormerod maintains a critical-rationalist strand inside operational research and is probably the closest ally the program has in that field.[^ch20-ormerod]

Every one of these communities tests its own work more than we test ours, while holding an epistemology at least as permissive. The philosophy is shared. They treated it as an obligation to build something else; we treated it as permission to build nothing.

That is why they are false friends and not adversaries, and why the word "false" does narrow work. Nothing they hold threatens this model. What they hold has been used, by us, as a reason not to find out whether the model is any good. Their versions confirmed a sentence we already had from Beer, and a position confirmed by respected neighbors does not feel like a choice that needs defending.

## What this chapter does not establish

It does not establish that usefulness is the wrong criterion for instrumental claims. It is the right one, and Chapter 24 will assess the model's instrumental claims by exactly that standard.

Nor does it establish that System Dynamics, the Hull school or critical systems heuristics owe this tradition anything. They owe us nothing. The sentence came from our own founder, and so did the test we dropped.

What it establishes is that the tradition took a criterion suited to one class of claim and applied it to another, that it dropped the test its founder attached to the criterion, and that System Dynamics shows what the criterion costs when taken seriously.

**Receipt — run one diagnosis through the structure/behaviour battery.** Take a single completed and published VSM diagnosis, for example the M∗Clean case in Espinosa's 2023 book,[^ch20-mclean] and subject it to the System Dynamics test set in Barlas's form, adapted minimally: boundary adequacy (does the recursion cut include what drives the behaviour at issue?), structure verification (is each claimed function and channel independently confirmable by someone who did not produce the diagnosis?), parameter and structure sensitivity (does the diagnosis change if a defensible alternative recursion cut is taken?), extreme-condition behaviour (does the model's predicted pathology appear when the named variety imbalance is driven to its limit?), and behaviour-pattern reproduction (did the predicted oscillation occur?). On a published case the extreme-condition item runs on the case record: find the period of the largest named variety imbalance and check whether the predicted pathology appeared then. Publish the battery and the result together. Price: one analyst for two to three weeks on an existing written-up case, plus a second analyst for three days on the independent structure verification; the sensitivity item is desk work. If a published diagnosis cannot survive structure verification by an independent analyst, the field learns that its diagnoses are not reproducible, which is Chapter 5's charge made concrete and would be the most consequential negative result available at this price. If it survives, the field has its first diagnosis with a battery attached, and a template the next one can be held to.

## Handover

Three chapters have taken shelters made of criteria. The last of them is made of a distinction, and it is the strongest.

[^ch20-forrester]: Jay W. Forrester and Peter M. Senge, "Tests for Building Confidence in System Dynamics Models," in *System Dynamics*, ed. Augusto A. Legasto Jr., Jay W. Forrester and James M. Lyneis, TIMS Studies in the Management Sciences 14 (North-Holland, 1980), 209–28. Read in the MIT System Dynamics Group memo D-2926-7 of 8 June 1979, 8: "We believe confidence is the proper criterion because there can be no proof of the absolute correctness with which a model represents reality. There is no method for proving a model to be correct." The page in the published version has not been checked.

[^ch20-barlas-carpenter]: Yaman Barlas and Stanley Carpenter, "Philosophical Roots of Model Validation: Two Paradigms," *System Dynamics Review* 6, no. 2 (1990): 148–66. **Unverified.**

[^ch20-barlas]: Yaman Barlas, "Formal Aspects of Model Validity and Validation in System Dynamics," *System Dynamics Review* 12, no. 3 (1996): 183–210. **Unverified:** the article has not been read; its separation of structure tests from behavior tests is carried from the research note. The list of named tests in the text is taken from Forrester and Senge (see the note to that sentence), not from Barlas.

[^ch20-forrester-tests]: Forrester and Senge, memo D-2926-7, 2 (table of contents): section III, "Tests of Model Structure" (structure-verification, parameter-verification, extreme-conditions, boundary-adequacy and dimensional-consistency tests); section IV, "Tests of Model Behavior" (behavior-reproduction, behavior-prediction, behavior-anomaly, family-member, surprise-behavior, extreme-policy, boundary-adequacy and behavior-sensitivity tests); section V, tests of policy implications. Each test is described in its own subsection.

[^ch20-stress-test]: Krishan Mathis and Margeret Heath, "A Stress Test for the Viable System Model — A Cumulative Research Program," open preprint (Zenodo, 2026), https://doi.org/10.5281/zenodo.22306416, §2.5, where claims are typed as node, edge and subgraph claims.

[^ch20-sterman]: John D. Sterman, "All Models Are Wrong: Reflections on Becoming a Systems Scientist," *System Dynamics Review* 18, no. 4 (2002): 501–31. **Unverified.**

[^ch20-hull]: Michael C. Jackson, "Pragmatism and Critical Systems Thinking: Back to the Future of Systems Thinking," *Integration and Implementation Insights* (i2Insights), 4 April 2023, https://i2insights.org/2023/04/04/pragmatism-and-critical-systems-thinking/; Gerald Midgley, "Theoretical Pluralism in Systemic Action Research," *Systemic Practice and Action Research* 24, no. 1 (2011): 1–15, https://doi.org/10.1007/s11213-010-9176-2, where theories "are more or less useful depending on the purposes of intervention" (checked against the abstract). Jackson's critique of the VSM is in *Critical Systems Thinking and the Management of Complexity* (Wiley, 2019), §13.4, pp. 325–37 (checked 23 September 2026 against the calibre copy of the 2019 Wiley edition). He classes the VSM as "functionalist in orientation," though structuralist rather than positivist (p. 329), and places it in the complex-unitary cell of his system of systems methodologies (pp. 334–36). Chapter 16 places Jackson's argument.

[^ch20-ulrich]: Werner Ulrich, "Rethinking Critically Reflective Research Practice: Beyond Popper's Critical Rationalism," *Journal of Research Practice* 2, no. 2 (2006): P1. The charge that critical rationalism has "not really overcome the empirical foundationalism" of earlier positivist research practice is checked against the abstract; the full text is open (ERIC EJ805693) but has not been read.

[^ch20-checkland]: Peter Checkland and Sue Holwell, "Action Research: Its Nature and Validity," *Systemic Practice and Action Research* 11, no. 1 (1998): 9–21. **Unverified.**

[^ch20-psm]: Gerald Midgley et al., "Towards a New Framework for Evaluating Systemic Problem Structuring Methods," *European Journal of Operational Research* 229, no. 1 (2013): 143–54, https://doi.org/10.1016/j.ejor.2013.01.047, for the reliance on "single case studies of intervention"; Fran Ackermann, "Problem Structuring Methods 'in the Dock': Arguing the Case for Soft OR," *European Journal of Operational Research* 219, no. 3 (2012): 652–58, https://doi.org/10.1016/j.ejor.2011.11.014, for "insights rather than testable results." Both checked against the abstracts only.

[^ch20-ormerod]: Richard J. Ormerod, "The History and Ideas of Critical Rationalism: The Philosophy of Karl Popper and Its Implications for OR," *Journal of the Operational Research Society* 60, no. 4 (2009): 441–60, https://doi.org/10.1057/palgrave.jors.2602573; Richard J. Ormerod, "Critical Rationalism for Practice and Its Relationship to Critical Systems Thinking," *Systems Research and Behavioral Science* 33, no. 1 (2016): 4–23 (online 2014), https://doi.org/10.1002/sres.2326. Identified from the metadata and abstracts; the full texts are paywalled and have not been read.

[^ch20-beer-dts]: Stafford Beer, *Diagnosing the System for Organizations* (John Wiley & Sons, 1985), 2. The lines follow "you are not determining absolute facts: you are establishing a set of conventions," in the passage on fixing the organization to be modelled and its boundaries.

[^ch20-beer-variants]: Stafford Beer, *Decision and Control* (John Wiley & Sons, 1966), 96, where "a cohesive collection of fundamental methods, which may be called a methodology, provides a framework for research which is neither right nor wrong, but only more or less useful"; Beer, *Diagnosing the System*, 144: "There is no 'correct' interpretation of the VSM. We have spoken instead of more or less useful interpretations. Even so, there may well be incorrect interpretations, in the sense that the model's power to account for viability may become denatured by their use."

[^ch20-beer-dc]: Beer, *Decision and Control*, 113. The passage describes the scientific model in operational research in general, not the Viable System Model in particular; Chapter 21 quotes it at more length. Jackson reads Beer's 1966 account the same way: the cybernetic concepts were to turn management's conceptual models into a homomorphic model "formulated rigorously enough to be falsifiable, and so scientific in Popper's terms," and a valid model "can be tested for its predictive value"; he adds that in Beer's view "the model is yet to be falsified" (Michael C. Jackson, *Critical Systems Thinking and the Management of Complexity* (Wiley, 2019), 299–300, checked 29 September 2026 in the calibre PDF).

[^ch20-jackson-agenda]: Jackson, *Critical Systems Thinking and the Management of Complexity*, 330, checked 29 September 2026 in the calibre PDF. The agenda he has in mind concerns "what structures are beneficial for viability and which impair it"; he raises it while discussing Achterbergh and Vriens (2011).

[^ch20-mclean]: Angela Espinosa, *Sustainable Self-Governance in Businesses and Society: The Viable System Model in Action* (Routledge, 2023), 116–17, "Case Study – M∗Clean (with A.C. Martinez)": a six-month project in a Mexican cleaning-services corporation with six regional offices, with a recursive analysis (fig. 4.4) and a VSM diagnosis of a small service unit (fig. 4.5). Checked 29 September 2026 in the calibre copy of the first edition. Whether the published account records enough of the case to run every item of the battery has not been checked.