April 21, 2025

What is the purpose of POSIWID?

Or, a short piece in response to The Discourse, and an experiment with being somewhat timely

---

A somewhat-famous internet blogger recently authored a post about "POSIWID" ("The purpose of a system is what it does"). It's a lazy and disappointing post so I won't bother linking it, but it did spark some interesting discussion about what people actually mean by this catchphrase.

Reflecting on this conversation -- and on my own relationship to the "POSIWID" meme, especially as it sometimes pops up on Twitter, Mastodon and their ilk -- I notice (at least) two distinct sentiments that the POSIWID meme can denote, and that these dual meanings contribute to people talking past each other, and cause (some) of the confusion and/or ill will in these fights.

This note is an attempt to name and articulate both of these ideas, with the hope of maybe helping someone, somewhere, to stop talking past their interlocutor.

One interpretation

The first meaning I see goes something like so:

When trying to reason about a system, especially a complex system containing many interacting subsystems, you must understand it, and its subsystems, in terms of their actual behavior, and not the behavior someone intends or believes them to have.

This sentiment is the one I personally most-associated with the "POSIWID" idea, and is the one I'm probably trying to invoke if and when I reference it¹.

To me, this sentiment is almost a truism, but one that serves as a useful reminder to stay close to reality, and to be skeptical of our and others' preconceptions about a system, especially when dealing with complex or unexpected behavior. "When the message enters subsystem Y, does Y actually log it, or is it just supposed to log it?" is a flavor of question that is often very useful when debugging thorny software issues, and I expect similar logic applies in any engineering domain.

In my mind, this sentiment is related to the injunction against speaking of "root causes" of many systems thinkers; both point at the fundamentally multi-causal and complex nature of systems interactions and behaviors, and warn (in different ways) away from trying to condense reality into overly-simplistic linear chains of cause and effect.

I have not extensively read Stafford Beer's work, but I have the vague sense that this meaning is roughly compatible with how he used the idea. For instance, Wikipedia quotes a speech in which Beer explains:

According to the cybernetician, the purpose of a system is what it does. This is a basic dictum. It stands for bald fact, which makes a better starting point in seeking understanding than the familiar attributions of good intention, prejudices about expectations, moral judgment, or sheer ignorance of circumstances.

To me, "It stands for bald fact, which makes a better starting point in seeking understanding …" sounds like a very similar sentiment to the one I articulated above.

Using similar words to "POSIWID," but attempting to be a bit more explicit, I might summarize this interpretation as "Don't worry about a system's (alleged) purposes; analyze what it actually does."

A second meaning

There's a second interpretation of POSIWID, though, that I see (implicitly) bandied about in online discourse. This one, I think, is closer to the meaning behind this other response post which I've seen going around.

I'd summarize this reading as something like:

When you see a system persistently exhibiting some harmful behavior, consider whether the designers or operators of the system actually intend for it to behave that way, no matter what they claim. If the pattern is persistent enough, you may consider the pattern to be prima facie evidence of hidden intent.

In this frame, "POSIWID" is less a tool for understanding behaviors as for inferring motive.

To my mind, this reading is a useful heuristic, but definitely just a heuristic, which must be balanced against other evidence and against the details of any given situation. It's useful, but can be overused.

Importantly, as well, while the first reading aims to be relatively value-neutral -- the Beer quote above sets aside or minimizes questions of "good intent" or "moral judgment" -- this reading is extremely value-laden, and steers directly for moral territory. It is a tool for assigning blame or inferring motive, and is thus largely a tool of social or political analysis, more so than engineering or technical analysis. That's not intended as a criticism or a dismissal, to be clear -- social and political analysis and argument are worthy goals -- but it's just worth being clear which lens we're using and which argument we're making.

I might summarize this interpretation as "If a system [persistently] does something, assume that its creators [or operators] intend that to be its purpose."

In closing

As I said, I think my first interpretation is close to a truism, but a very useful reminder, especially to engineers and designers of complex systems. The second meaning, meanwhile, has value as an interpretative heuristic, but is much more or conditional or context-dependent. Importantly, they also speak to somewhat different domains or levels of analysis; the first meaning is (relatively speaking) "just technical advice," but the second is a tool for social or political analysis and inferring intent.

They do share a common core, however: both interpretations advise you to set aside claimed or documented purposes, and to focus primarily on observed behaviors, instead.

However, they take you in different directions, or highlight different aspects of what you do with those observed behaviors, and. In particular, the second (sometimes implicitly) jumps more-or-less straight to assigning blame, while the other mostly stops there, and lets the user decide whether or how to assign meaning to those behaviors.

It seems clear to me that some amount of the disagreement and fighting over the usefulness or the role of "POSIWID" comes from readers and users intending different interpretations, and thus talking past and misunderstanding each other.

Of course, some of it also comes from people being smug assholes on Twitter; sometimes being a dick is, indeed, the purpose.