the purpose of a system is you can't always get what you want

giving up on an old friend

I am, I think, finally realising that the “POSIWID” principle is more trouble than it’s worth.  It’s a great slogan, and I don’t think I’m going to stop using it occasionally.  But it seems to cause lots and lots of confusion, and if that’s what it does, then … well, it’s not really available to argue that this is unfair, or the fact that it’s systematically achieving a result other than the one intended can be ignored.

So I’m going to have one more go at explaining what POSIWID really means, then give up.

OK, to start with (and one of my main reasons for giving up on it), “The Purpose Of A System Is What It Does”, although it’s a fun thing to say, isn’t a fundamental principle of cybernetics at all.  It is not even an axiom of Stafford Beer’s own model, the Viable Systems Model.  What it is (as I started writing in a previous post) is a heuristic to try to explain to people how to draw the black boxes and go about applying the model.

It's important that this is a heuristic that’s only to be used in the context of something that’s at least potentially a viable system.  So, for example, if you have an air conditioning unit that’s emitting black smoke, you can’t say “the purpose of this air conditioning unit is to emit black smoke”, because it’s not an ongoing system that’s going to be allowed to keep doing that.  One of the big misconceptions that the POSIWID slogan fosters is that it should always be (but rarely is) mentally expanded to “the purpose of a system is what it systematically does, on an ongoing basis, with the permission of the other systems which form its environment”.

And that last clause about the other systems is really important.  Immediately after coining the POSIWID phrase in Diagnosing the System, Beer emphasises that the “purpose” of the system you are analysing is always going to be a compromise with other systems in which it is embedded.

So, for example, a (kind and clever) reviewer of my book applied the principle to his train journey and concluded that “the purpose of the [British] rail network is to disincentivise people from making train journeys”.  Is he right?

Well, to start with, we can definitely say that the purpose of the British railway system isn’t “to provide comfortable and affordable rail travel to everyone who wants to use it”.  As Beer says, it’s just not sustainable to claim that the purpose of something is an outcome that it almost never achieves.

(Of course, this is the other big source of confusion – the use of “purpose” language, which sounds like it’s attributing mental states, to systems which don’t have any mental states.  As I’ve said before, I think this is defensible because sentences like “the purpose of a screwdriver is to drive screws” seem to make sense. But it causes a bit of a lurching sensation in a lot of people.  One thing that’s very important to remember here is that POSIWID is not meant to be a claim about the universe – it’s a heuristic for deciding what information needs to be taken into account.  The whole business of cybernetic, and any other kind of, modelling is to decide what features of the world you’re going to pay attention to and ignore.  The step logically prior to POSIWID is “a variable needs to be taken into account if a change in that variable has the potential to affect the system’s purpose”, which raises the question “how do you define that purpose”. I’ll come back to this in a couple of paragraphs).

So, “the purpose of the rail network is to disincentivise people from making train journeys”, true or not?  Well, in my view more true than not true, but you have to be super clear what system you are analysing.  Lots of individual people working on the trains have the purpose of “driving it to Manchester on time”, “selling crisps”, “giving information to the passengers” and so on.  But each individual train is embedded in a system which has the purpose of “run the timetabled journeys safely”, which in turn is embedded in a system with the purpose “set a timetable consistent with the available resources”, which is embedded in a system to “balance the demand for rail travel with the politically sustainable subsidy”.

You could do a lot more work on identifying those systems and being more clear and rigorous about what they do, but the point is that the part of the rail network you interact with has to compromise its own priorities with those of the higher levels.  Given past decisions about investment and current decisions about subsidy, its does indeed, to a large extent have to disincentivise people from making the train journeys they might prefer to make if the constraints were different.  And that is a steady state for the rail network, so I think it’s right to say that the purpose of the consumer-facing part of the system is to act as a rationing system for rail journeys. 

Which affects the kind of information you might consider relevant to collect in analysing that system (this is the bit that I’m coming back to from two paragraphs ago).  For one thing, if you understand that the British rail system is a rationing system, you’ll be less inclined to compare the cost of a journey from Cardiff to London with a journey from Brest to Paris.  You wouldn’t post a photo on social media saying “this train in Denmark is nice and empty and the journey only cost £4, why is my commuter train from Surbiton three times the price and horribly overcrowded?”.  And conversely, you would start regarding a whole load of other information about electoral geography and politics as much more relevant to the higher levels of the British system than similar considerations might be to the SBB.

So anyway – POSIWID.  As you can see, I think it can be defended and I think it’s important.  But it’s complicated and confusing, and management cybernetics has enough problems getting off the ground.  So I think it’s adieu from me – I’ll try not to use it so much in future.