Cross-posted from http://mleseminar.wordpress.com/
...
You can find the handout for this week here.
I thought this was a really good paper, and we didn’t find all that
much to criticise in it. It was a bit frustrating not to hear more
about Fitelson’s positive story, in particular about the bridge
principle that he would endorse instead of the various versions of RTE
that he criticises. He’s clearly saving the juicy stuff for his book.
In particular, I find it hard to see how he plans to steer a middle
ground between the Carnap/Williamson-style ‘a priori priors’ version of
objective bayesianism, and the subjective bayesian approach. My naive
take on the matter is that you either think that there’s a unique
correct set of priors or you don’t. Maybe these priors aren’t a priori
knowable (contra the Carnap/Williamson approach), although it
seems that a position like this would be committed to complete
epistemic rationality being in principle unattainable.
I wasn’t sure how strongly Fitelson meant to criticise the
subjective Bayesian’s RTE’. Although as it stands the principle is
useless, presumbly the subjective Bayesian wants to find a principle
which is extensionally equivalent to RTE’ but which is not useless,
because it picks out K’ in a different and more illuminating way.
Fitelson gives no argument that this will prove difficult.
Gonzalo pointed out an interesting consequence of Hempel’s confirmation theory – all propositions are equally confirmed simpliciter.
Of course, what we are interested in is confirmation of propositions by
particular other propositions; this just underlines that Hempel’s
confirmation relation is a logical and not an epistemic relation.
Another observation Gonzalo made is that M is trivially false if we
allow for properties like ‘being such that this grass is green’. ‘This
grass is green’ confirms ‘all grass is green’; but obviously statements
like ‘this grass is green and this grass is such that that grass is
blue’ do not confirm ‘all grass is green’, as M says it should.