A stress test for the Viable System Model, published one chapter at a time
Today we are publishing the first chapter of Punch and Judy: A Stress Test for the Viable System Model. It is Chapter 3, “The Law We Quote Is Not the Law Ashby Wrote”, and you can read it here
This post explains what the book is for, how it is built, and why we are asking for help.
Why this book
2026 is the centenary of Stafford Beer’s birth. Centenaries usually produce retrospectives and collected volumes. That work is worth doing, and others are doing it well. We chose a different use for the year. Beer built the Viable System Model by taking things apart, and we think the more useful tribute is to take his model apart in the same way.
The book makes one charge, and it is narrow. The VSM has never been put at risk. It has not been shown to be false, and it has not been shown to be right, because the community around it, ourselves included, has never said which observation would make the model give something up. When an intervention worked, we credited the diagnosis. When it failed, we said the model had been applied badly. A framework that can absorb any outcome describes well, but it cannot tell us where we are wrong.
The book is not an obituary, a defence or a demolition. It is addressed to the people who teach the model, apply it and argue about it. We are among them, and where the record is embarrassing, it is our record.
One device runs through the book. Each charge ends with a receipt: the smallest piece of work that would settle the question, what it would cost, and what a negative result would show. A criticism that names no such test is a complaint, and complaints have never cost this model anything.
How the book is built
The book has thirty chapters in five parts, framed by a prologue, an introduction and a coda.
Part One — Inside: the model we did not read (Chapters 1–9). Nine things we inherited and did not check: the neuroanatomical mapping we teach, System Four reduced to scanning, the law of requisite variety as we quote it, the channels without owners, the criterion for recursion levels, the missing people, the missing account of how a viable system comes into being, and the participation loop that stayed an intention in the founding case. Chapter 9 asks why none of this ever cost us anything.
Part Two — Outside: the people who found it without us (Chapters 10–16). Other fields, among them sociotechnical design, organization theory, complexity science and Ostrom’s work on the commons, reached several of the model’s components on their own, and often measured them. The part ends with the three published objections that we have never answered.
Part Three — False Friends and Enemies (Chapters 17–22). Who has been standing around not charging us? Criteria that have kept the model out of reach of a test (usefulness instead of truth, first stated by Beer himself; “a way of thinking”; “testing is a category error”), held by neighbouring communities and by us, and the few real adversaries.
Part Four — What we lose (Chapters 23–27). What remains that is ours: a maybe narrower claim, and six things we know of nothing else that does, each stated with what it costs. The part is written so that it could come out empty.
Part Five — What’s at Risk in the VSM (Chapters 28–30). What would count against the model, the two cheapest decisive experiments, and the bill for the whole programme.
Coda — Judy. A viable system is viable for someone. The model helps a cartel survive as well as a hospital. The coda asks who has standing in the design.
Chapter 3 sits in Part One. It separates three things we have allowed to blur: Ashby’s law, which holds; Beer’s compression of it into a design maxim, which is useful but is not the law; and the good regulator theorem, which does not prove what its title claims, as its own authors wrote in 1970. The correction it proposes costs nothing: a change in what we put in our footnotes. We applied it to our own published material first.
A milestone, and a beginning
A first public chapter is a real milestone, and we are treating it as one. It is also only the start. The book is a stress test. What we want in the end is an initiative for a better foundation of the VSM: stated claims, stated failure conditions, experiments that could go against the model, and a public record of what changed and why. The working paper A Stress Test for the Viable System Model: A Cumulative Research Programme sets out that programme: https://doi.org/10.5281/zenodo.22894039
The remaining chapters will follow one at a time. Every correction goes into the text with its date, and the record stays visible.
We are looking for help
We think the potential of this project is large. So is the work: thirty chapters, primary sources to check line by line, experiments to design and run, and a research programme that no two people can carry.
We could use help from people who:
- know the primary literature, or hold copies of the texts we cite
- have tried to measure variety, draw recursion levels or run a VSM diagnosis, and kept records
- work in the neighbouring fields of Parts Two and Three and can tell us where we have them wrong
- simply think we are wrong, and can say where
Bring one thing: a source, a correction or an objection.
We have only just started, and the way we work together is not fixed. If you are looking for a more inclusive way to shape this, beyond reviewing chapters we have already written, please come forward. We would rather build the process with the people who join than hand them one.
How to take part
- Read Chapter 3 and comment on the page or the Slack channel linked from the page
- Join the Community
- Follow new chapters through the newsletter.
- Or write to us directly: Margeret; Krishan
Krishan Mathis and Margeret Heath