Bayesian arguments about climate sensitivity

… seem unconvincing to me. I’m just not convinced by the reasoning behind the choice of prior. Which isn’t to say I’ve thought about it enough for a coherent critique to emerge, otherwise you would be reading it. The argument just doesn’t compell me. What I’d like to see to find this stuff more compelling is perhaps an iterative process where the effect of each new study/piece of evidence is shown.

Seems to be there is a tendency to go from ‘we should use Bayesian probability for decision making’ to ‘we should use Bayes theorem for all our beliefs’. Which is quite a big philosophical jump. I’m still rather partial to the Popperian view that a belief is rational provided it has not been discredited, and you are open to (indeed, seek) evidence to its contrary.

Like many other dynamical modellers I’m suspicious of statistical models. The underlying premise of statistical models is that the future will be like the past, which reeks of all swans are white style inductive fallacy. I’m using Bayes theorem in some ideas I’m pursuing at the moment, and trying to be quite open to it, but there’s something about it I find a little uncomfortable. Maybe when I get around to reading Jaynes properly I’ll be converted. The idea that all rational agents must believe the same things I find disturbing, and perhaps a bit dangerous.

Not going to link to any of Annan’s stuff from here as it is mostly drafts (look up his website if you are interested), and this critique is hardly coherent. I just needed to get my reservations down on paper.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: