Now that it's finally over, let's have a go at writing about the PhD. Here's a github page with the full PDF and live model runs if anyone's interested in a glance-through. The model output pics in the results chapters are probably the nice bits to look at.

I'll break it down into blog chunks rather than attempt to deal with it all at once. Unfortunately I think there's no way around starting with a cathartic "once upon a time" ramble back through how it ended up the way it did - as always, likely of more value to me than anyone else.

The original idea was all that Hayekian stuff: is there really a problem with planning in complex systems? Can something like agent-based modelling (ABM) be used to dig into the question? (Some people think Hayek would have been an agent modeller had he had access to the technology.)

That question lost its appeal when it became apparent (due to the work of Elinor Ostrom and others) that people can generally muddle through to solutions that have little to do with either Hayek's "plan to resist all planning" (Oakshotte) or the totalitarian demon he so vividly summoned (Cartoon version here!)

More than that, diverse ways to manage resources had evolved several times in the past, falsifying Hayek's claim that the price system was galactically unique freak chance. It may well be an entirely normal outcome of you take a bunch of humans, shake them up in a beaker and leave them for a time.

That's not to say there aren't a bunch of fascinating issues there - in fact they're increasing in importance as we continue to give planetary boundaries a good kicking (this was the topic of my recent talk, PDF available here, at Modelling on the Move). But - as often happens - I got waylaid as we were knocking around ideas for case studies to think about planning vs spontaneous order. There's more to write about that time (including going back to the "adaptive landscape" stuff) but let's try a short summary to get to where the PhD actually ended up.

Case study? Food! Why not? A specific, interesting sector and also a hot topic in transition circles - both for practitioners and in academia, where a whole literature around alternative food networks was developing (that link's rather out of date now but gives a flavour). There's a simple underlying rationale: if you're going to have to radically relocalise your economy, deal with essentials first. Food networks are also intrinsically geographical - how does food get to where it's needed? (With the questions `who feeds Paris' and `can Britain feed itself' nicely summarising two very differing views of the role of markets and planning in getting from field to fork).

Two things happened next. The transition movement claims that a radical relocalisation of the economy is both inescapable and desirable as we adapt to climate change and peak oil. I'm not a peak-oiler; climate change will get us long before oil depletion does. Bill McKibben gives a great summary of this argument; this issue of the Oxford Review of Economic Policy is an excellent resource for the deeper economic forces underpinning that. The transition movement localisation `hypothesis' (if we can call it that) is very clear. The problem was, it didn't appear to be based on anything other than `common-sense' assumptions about the role of oil, plus a generous dollop of "and this is the right way to live anyway".

So I started digging around for modelling ideas that might shed light on that fundamental question: are the transitionistas right about a radically relocalised future? Are they right about the unique role oil plays? For myself, the more important question was: if we're going to have to phase out carbon-based fuels at some point, do we actually know what the impact is going to be? Can we can perform an equivalent of the tablecloth trick with our entire infrastructure, where perhaps nothing changes except perhaps the source of fuel itself? If not, quite what manner of changes can we expect and what impact can we have on shaping them? A vicious problem: how to observe a set of systems to learn their innate function while imagining ways to steer or even remake them?

I should manage expectations at this point: I don't know the answer. I'm a bit closer to working out why we don't have answers and I've met some other people starting to ask these questions in a variety of ways. But yes - there's no great punchline...

The second thing that happened: it became apparent you can't model food separately from the rest of the economy. Having spent some time trying to build some sort of "food plus rest-of-economy" model, I then discovered that economists had this incredible abstract framework for describing the way components of the economy interact: micro-economics.

The PhD was about an agent-based modelling approach. But in parallel, it's been about discovering a rich seam of economics and some very self-aware modellers who have thought deeply about what they're doing. That shouldn't really be a surprise, though from a background in political studies at Sheffield University, it contrasts with the caricature of economics one gets through osmosis, where economics reduces to a figleaf for power and transmogrifies into "neoliberalism". Again, something to write about another time.

As someone struggling to actually build a model, these ideas were gold dust. They were also, in various ways, quite at odds with the stories that ABM tells - and I'm hoping (though we'll see) this is where the value of the PhD comes from.

I did, in fact, make an agent-based model, but it's a bit of a weird one. By hooking myself to something like the `transition question' I was forced to deal with a whole set of problems that, in the normal course of agent modelling, just don't arise. Why not? I think because ABM has a research agenda that's shaped by the method itself - or as Maslow puts it, `method-centring' rather than `problem-centring' (Maslow 1966 p.15). He also says on the same page, more famously: "it is tempting, if the only tool you have is a hammer, to treat everything as if it were a nail" - but `method-centring' actually captures better what's happened with ABM.

I'll leave it there for now and probably have a go at explaining "the problem with ABM" in the next post.