POSIWID. It stands for “The purpose of a system is what it does,” and it sounds — let’s be honest — a bit goofy. Like a detergent brand or a forgotten 1970s children’s show. POSIWID. And yet the phrase keeps turning up in certain corners of the Internet. You’ll find it quoted on Hacker News, Reddit, Medium, Substack, and, inevitably, LinkedIn posts written by management consultants. Increasingly, it’s showing up in conspiratorial YouTube videos and weird AI-generated TikTok explainers about how the “system” is rigged.
POSIWID isn’t quite mainstream – more like a niche cybernetics meme. But where did it come from, and what does it actually mean? Well, my PhD research involves several historical actors associated with cybernetics in the 1960s and 70s, including some who drew on the work of Stafford Beer, the man most commonly linked to POSIWID. So, let’s talk about what POSIWID originally meant, what it has come to mean, and why this shifting meaning presents us with a challenge when it comes to the question of digital memory.
Recently, POSIWID came under fire in a blog post by Scott Alexander — a San Francisco Bay Area psychiatrist, professional blogger, and a central figure in the online “Rationalist” community. Writing on his Astral Codex Ten Substack, Alexander argued that the phrase is often used to make sweeping claims that are (under analytic scrutiny) obviously untrue — a kind of rhetorical shortcut that substitutes unthinking mantra for rational analysis.
Alexander offers some “obviously false” examples to demonstrate the absurdity of POSIWID:
- The purpose of a cancer hospital is to cure two-thirds of cancer patients.
- The purpose of the Ukrainian military is to get stuck in a years-long stalemate with Russia.
- The purpose of the British government is to propose a controversial new sentencing policy, stand firm in the face of protests for a while, then cave in after slightly larger protests and agree not to pass the policy after all.
- The purpose of the New York bus system is to emit four billion pounds of carbon dioxide.
However, I suspect that both Alexander’s criticism and the debate it provoked rely almost entirely on the definition of POSIWID from its Wikipedia page. The historian Ronald Kline writes about how the “rich discourse of cybernetics and information theory” has been “flattened” since the 1980s, and I think the way POSIWID has become a Wikipedia meme, detached from its original context of 1970s British management cybernetics, might be a good example of this.
POSIWID (called a “systems thinking heuristic” on the Wikipedia page – and I agree it is a heuristic) is associated with the British cybernetician Stafford Beer – a man who wore three-piece suits, drove luxury Rolls-Royce motorcars, and once tried to build a socialist utopia in Chile using telex machines and molded 1970s fiberglass swivel chairs. However, the phrase does not appear in Andrew Pickering’s The Cybernetic Brain or Eden Medina’s Cybernetic Revolutionaries, both of which cover Beer. Nor is there anything obvious listed in the Stafford Beer papers.
Nor can I find the acronym POSIWID in Beer’s published writings (it might be there, but I wasn’t able to find it after a quick hunt). He does use the full phrase “the purpose of a system is what it does” several times. The earliest specific mention of the acronym POSIWID that I can find is from a 1998 paper by another author (who does not mention Stafford Beer as the source).
Yet while Beer doesn’t use the exact acronym POSIWID, he does use the full phrase “the purpose of a system is what it does” multiple times – notably in his 1979 book The Heart of Enterprise, which is possibly his first use of the heuristic. I think that book is a good place to start to understand both the original richness and complexity of “the purpose of a system is what it does” and its later “flattening” into Wikipedia’s POSIWID page.
Beer introduces “the purpose of a system is what it does” in the very first chapter of The Heart of Enterprise, where he invites us to consider a familiar system: a tiger. Probably something you’ve seen at a zoo. “We are fairly sure about the boundaries of the Tiger System,” Beer says. “Those boundaries were imposed on the Tiger System by the curator of the zoo, and we accepted his say-so on the matter.”
Beer then lists several plausible purposes of a tiger:
- to be itself
- to be its own part of the Jungle System
- to be a link in animal evolution
- to eat whatever it eats, for Ecology’s sake
- to provide tiger-skins
- to perpetuate the genes of which it is the host
At first glance, this list might suggest a sort of pluralism – multiple possible purposes, all coexisting, and we can pick our favourite arbitrarily. But I think Beer is instead unsettling the very idea that ‘tiger’ is a fixed system at all. Because the next thing Beer does is to destabilise the category of tiger into absurdity. He asks us to imagine encountering a tiger in the Arctic. Or one the size of a mouse. Would it still be a tiger? What defines the Tiger System? Do climate and size count as part of the Tiger System? What about a tiger that was made of voxels? Or a tiger that didn’t have orange and black stripes? Or a cross between a tiger and a lion?
Pushing things further, Beer begins to sketch out what Andrew Pickering calls a black box ontology – a view of the world as composed of systems within systems, each defined by what it does, not what it is. When Beer says the tiger may be “its own part of the Jungle System,” he’s invoking precisely this nested structure – where systems are defined relationally and functionally.
Yet Beer makes still one more move – beyond what has been called first-order cybernetics (systems defined by input/output behaviour) – and into what is sometimes called second-order cybernetics, where the observer becomes part of the system. I should point out the first-order/second-order division is not uncontested historically. Nevertheless, the act of bringing the observer into the system under study is often considered a second-order cybernetic gesture.
“It is you
the observer of the System
who recognizes its purpose.
Come to think of it, then, is it not just YOU — the observer — who recognizes that there is a System in the first place?”
Beer’s point – expressed in his typically gnomic (some would say “guru-like”) style – is that both the identity and the purpose of a system are grounded in our perception of what it does. And that perception is always contingent, situated, and subjective. As a heuristic, we very often define a system’s boundaries by the effects it produces in our experience – that is, by what it does to us.
“The recognition of a system and of its purpose is a highly subjective affair.”
In other words, the system – and its purpose – always emerge in relation to a point of observation, a perspective. Pragmatically, Beer is arguing that when managing a complex system, it’s not just helpful to observe that system from different perspectives – it’s necessary, because each perspective actively helps constitute what the system is. The boundaries, the behaviour, even the purpose of the system shift depending on the position, experience and assumptions of the observer. Any model of the system that wants to regulate that system should account for multiple perspectives.
Whether you agree or disagree with Beer’s argument is besides the broader point I’m making here. We can hopefully at least acknowledge this is already a much richer set of concepts and metaphors than appears on the POSIWID Wikipedia page.
After the tiger, Beer turns to other examples: a car with faulty spare parts provisioning; the convoluted logic of the British railway system. In each case, he asks: what does the system do, not what does it claim to do?
- A car meant to offer luxury and reliability but that constantly breaks down? Its purpose, from the customer’s perspective, is to generate inconvenience.
- A national railway that makes travel so difficult that it discourages passengers? Its purpose, as inferred from observed behaviour, might be to dissuade anyone from traveling by rail. “It would be more sensible to travel by car”.
This all sounds uncannily similar to the “obviously false” list of purposes that Scott Alexander included in his Substack post. But here, in what is quite possibly the origin of the heuristic “the purpose of a system is what it does” in a management book from 1979, Stafford Beer uses several “obviously false” purposes as the jumping off point for his argument. So, what is Scott Alexander missing? And this is not necessarily Alexander’s fault, because none of this is mentioned in the Wikipedia article on POSIWID.
Beer’s argument is this:
“The facts about the system are in the eye of the beholder…
Systems are nested within other systems; the observer who imputes purpose to a system belongs to several of them, and is excluded from others. Thus his perception of purpose in any one system is not only subjective, not only conditioned by his acceptance or rejection of boundaries, but is also a function of the relationships that he maintains across the complicated webs of interactive systems, one with another.”
This is a far cry from the version of POSIWID that circulates online. It’s not about proving bad faith (“the system wants to fail!”), nor about pretending intention doesn’t matter. It’s about situated observation – recognising that every model of a system embeds assumptions about purpose and system boundaries.
Beer was writing a management book, and this historical context is also important. He wanted to give advice to management teams on how to regulate and steer their firms more effectively. Therefore, he also suggests a solution. He invites his reader to consider the perspective of the system itself — or more precisely, the internal logic of its regulatory phenomena as distinct from the perspective of the system designer.
This might sound odd. After all, firms don’t “have” perspectives in any conscious sense. But it’s central to POSIWID (and see Aashish Reddy’s Substack that digs into this idea a bit more). Beer’s point is that systems regulate themselves by responding reflexively to external disturbances in order to maintain internal stability. In cybernetic terms, these regulatory processes determine both the boundaries of the system and what counts as relevant within those boundaries – they constitute the system’s world. And when we analyse a system, we can choose to model its purpose not from the outside, but from the standpoint of its internal regulation. The purpose of a system, according to this view, is revealed by what it regulates and what it works to reproduce – the patterns and structures (and perspectives and experiences) it maintains over time. As Beer might ask: is the purpose of the Tiger System to maintain the Zoo Visitor System?
In a 2002 lecture (EDIT: Sorry, the lecture was in October 2001, the paper that the lecture was published in was from 2002), Beer put it like this (and it’s a line notably absent from the POSIWID Wikipedia page, which quotes the rest of the paragraph that follows but omits the leading sentence):
“With the complete range of cybernetic discoveries to hand, it is always possible to analyze a situation from the point of view of its regulatory phenomena.”
This is a crucial shift. Instead of imagining ourselves standing outside the system, measuring it against external goals, we ask: what is this system actually trying to stabilise? What does it treat as noise and what as signal? This reframes our understanding of “purpose” – not as a stated aim, but as the effect of internal regulation. The purpose of a system is what it does. POSIWID.
You can disagree with Beer’s argument (and, personally, I don’t agree with everything above, though I find it rich and fascinating). Nevertheless, hopefully you can see how what Ronald Kline calls the “flattening” of information discourse has taken place from the publication of The Heart of Enterprise in 1979 to the creation of the POSIWID Wikipedia page in 2008.
If we treat POSIWID like it’s presented on its Wikipedia page – a self-contained unit of knowledge summed up in a pithy one-paragraph quote – we’re impoverishing our reading of Beer and of cybernetics.