Tuesday, November 22, 2011

Bayesianism and politics

I went to Westminster Skeptics last night to hear Dr Graeme Archer's very entertaining talk about evidence-based policy and the problems that there are with using statistical evidence to inform policymaking decisions.

Through necessity, Graeme skipped though quite a lot of it and I didn't fully follow everything, but broadly speaking, his talk appeared to give a concrete underpinning to a lot of views that I hold already.

Obviously, this is a very good thing. I appeared to have reached the same position by a bayesian process that he says that he has reached as a practicing statistician in the pharmaceuticals industry.

This may have come up after I left (I couldn't stick around for the questions after) but there was one thing that jarred. He seemed to think that an ability to treat statistical evidence to tweak bayesian priors rather than use it as a device by which we wipe previous assumptions out (and replace them with a beleif in whatever the 'evidence' tells us to beleive in) is a trait that is widely found in the political left but not the right (Graeme is a Tory).

A few quibbles:

Firstly, the pre-Thatcher Conservative Party was a good deal more Burkean than they have been since. Paraphrasing Burke very swiftly, he was clearly of the 'if it ain't broke, don't fix it' view, and that tradition (the Bayesianism of the most powerful sections of society?) needs an overwhelming case to be made before reform is acceptable. This appears to be the political manifestation of small-c conservatism.

Secondly, I think that he would have had a point if he'd argued that New Labour were particularly guilty of gathering evidence that appeared to support some radical-ish managerial approach, and using it to force through Year-Zero type policies, but there's a political context behind that which I will come to shortly.

Thirdly, if you're a big fan of market processes as a way of making decisions, then Graeme's (and my) views are comforting ones. I'm quite happy to sign up to this statement by way of a general creed:
The distributed wisdom of lots of small decisions will usually be a great deal better than less frequent big decisions made as a result of a formalised process. The main brake upon this means of making decisions should be a counterweight from elected bodies that apply distributed moral wisdom 
 (Apologies again for self-linking).

I accept that this view is held by socialists who aren't dismissive of the markets and by Tory wets, but probably not by Democratic Centralists and their fellow travellers on the left or the Tory right. As such, if you were to map it on a simple (fictional) linear left-right axis, the big bump would probably be on the centre-right.

My criticism of most Conservatives is that they're far to relaxed about the distorting power of monoplies on the economic side of this issue, and of commercial pressure-groups on the political elements.

And surely the phrase 'there is no alternative' rings a bell with any Conservative?

I'd also argue that politicians are behaving rationally (in that very particular definition of the word) when they embrace certaintly - particularly politicians who don't generally get an easy ride from the press. It's one of the reasons that the governing style of the current government is a good deal more superficially attractive than the the white-knuckled hyperactivity of the previous lot.

This may read like an excuse from a political grouping that is sick of constantly losing elections because of media hostility, but I can understand where it comes from.

Finally, to start another hare running, I think I'd be able to argue Graeme into a position where he'd oppose all future uses of referendums based on his views on this, but then I regard almost everything as an argument against referendums.

(This is another of my posts that is too long because I don't have time to boil it down and tidy it up - sorry)

No comments: