budgeting makes hypocrites of us all

we pretend to make a forecast and they pretend to hold us to it

[PROGRAMMING NOTE: Lots of people have pointed out this story, about how Air Canada set up a chatbot on its website, then the chatbot “hallucinated” a bereavement fare policy, and then (unaccountably, and yet somehow also predictably), Air Canada tried to wriggle out of being obliged to honour the promises made on by the chatbot.  I’ve got surprisingly little to say about this, I find – I’m sorry to be so promotional, but if you read my book when it comes out you won’t be at all surprised at this sort of thing happening, and I don’t think I can do justice to what went on here at much shorter length than the whole book.  If this had happened a few years earlier it would have made a great case study.  One clue I’ll give while we’re waiting for publication is that although ChatGPT output isn’t very like creative thought, ChatGPT hallucinations are quite like some types of organisational decision making system – they’re crud thrown up by something that doesn’t really know what it’s doing, but is trying to produce output that’s superficially consistent with what it’s done in the past.  Consequently, corporations like Air Canada can, and often do, have hallucinations].

Anyway, on to this week’s post:

I am haunted by a question raised by “Policy Sketchbook”, which is the Twitter name of a friend who made some useful comments on the book draft, and who has quite deep knowledge of NHS budgeting practices.  This practice involves managers putting together an annual plan, and then being held to account for divergences from it.  That’s quite a normal and useful thing to do; as PS says, in Stafford Beer’s model of the “viable system”, one of the fundamental tools used to stop the information flow becoming overwhelming is that “resource bargain”.  That’s shorthand for the principle that a sub-unit ought to be allowed to do what it wants as long as its plans do not fail to achieve goals necessary for the functioning of the rest of the system, or drain resources needed for the rest of the system. 

In information theory terms, it means that most of the time, the system’s demands on bandwidth are only the single bit needed to transmit “everything’s ok”.  Then from time to time, when everything ceases to be OK (because the unit has encountered something in its external environment that it can’t handle with the resources it has), much more management bandwidth needs to be allocated for a period of time, until the subsystem has “relaxed” again.  Since hopefully not everything will go wrong at once, this is a tool of “variety amplification” for the higher level management – it’s an organisational form which allows them to manage a much bigger system than they could handle with their own information-processing capacity.

A lot of management theory – explicitly in Beer’s cybernetics, implicitly in most of the rest – is just about making sure the information balance sheet adds up and things do not become, literally, unmanageable.  But what can you say about a system like the NHS budget as described by PS, where there is a clear pathology; the budget is never fulfilled and everyone involved knows that it’s a fiction which will have to be revised a few months in.

I think you have to start the analysis by saying “POSIWID” – the Purpose of a System Is What It Does.  Although it’s got the form of a resource bargain, it’s not sustainable to say that the NHS budgeting system is one – that would be equivalent to saying that its purpose is to do the thing that it consistently and predictable fails to do.

So its purpose is more likely to be the thing that it does do – it puts NHS hospitals, trusts and other delivery units in a constant state akin to being a defaulted debtor.  It’s the opposite of a resource bargain – it’s a system which exists to undermine the independence-subject-to-accountability that was built into the system-as-planned, and to put them in a position where they are always subject to control from the centre, as a condition of getting their revised budgets approved.

This is, of course, a very inefficient way of getting this outcome.  If you want centralised control, then just control things centrally.  Of course, one reason why you might want to preserve the forms of independence is that if it’s known you’re in control of something, then the location of the accountability switches from them to you.  Like so much else in the public sector, it’s an accountability sink.

I note, uneasily, that this practice is growing.  Local authorities in the UK have a rule that they aren’t allowed to declare a deficit budget, not even for one year.  (This leads to news stories in which they’re called “bankrupt”, even though that’s very much not what bankruptcy means).  This too is used as a means of control – at present, Birmingham is being forced to sell off a load of assets, and being blamed for the fact that its central government subsidy was cut and statutory obligations increased until it inevitably reached this point.  It’s a scarily effective tactic for people with a different moral sense to mine.

There’s a lot of internal links in this post, aren’t there?  I hope that’s reflective of this ‘Stack being a coherent weekly piece of work outlining an overall system, rather than of me going further and further down a self-referential rabbit hole.