In the last four weeks I’ve been to two epistemology conferences on (or about) the west coast – the INPC conference _Knowledge and Scepticism_ at Washington State/Idaho, and _Formal Epistemology_, or FEW, at Berkeley. First a couple of quick sociological notes.

The male/female ratio at each conference was about 5:1. I might have mentioned this before, but epistemology, formal or informal, seems to be the least gender-balanced of all subfields of philosophy.

Last fall there were two philosophy of language conferences on the east coast with minimal overlap between the attendees. And the same thing has happened with the two west coast(ish) epistemology conferences. I was the only person at both INPC and FEW. Now this was in part coincidental. Some of the people at INPC (such as John Pollock or Jonathan Schaffer) could easily have fitted in at FEW. And plenty of people at FEW do stuff with a tight enough connection to traditional epistemology that they could have fitted in nicely at INPC. Peter Vranas and Sherri Roush are the most obvious candidates for this, but many of the philosophers of science and probabilists there would have had useful things to say to the traditional epistemologists.

In fact, it was too bad that there wasn’t more mixing between the groups, because they could learn a lot from each other. The advances in formal methods in recent years are really quite stunning.

Some of the work on Bayes Nets by Stephan Hartmann and Luc Bovens suggested ways to question the thesis that more varied evidence has greater evidential force.

Hartmann and Bovens, among others, have also been trying to find a way to state the coherence theory of justification in precise probabilistic terms, which seems like it should be interesting to lots of epistemologists.

And Clark Glymour’s work, also using Bayes Nets to do draw conclusions from messy data, is leading to remarkable successes in both practice and, I suspect, theory. It’s one thing to have a theory about when we know a lottery ticket will lose, it’s another thing altogether to have a theory about when we can _know_ that, say, exposure to lead causes reduction in children’s IQ when our only way to analyse the data uses probabilistic techniques. (One might even suspect the latter is a more important question.) And I think the work done by people using formal methods is much more likely to lead to philosophical breakthrough here than the work done by people like, say, me. Obviously philosophers who aim to make their work practically relevant will have a better chance of being practically relevant than those of us who are comfy in comfy armchairs. But there’s a philosophical payoff to practical relevance too. In a lot of cases it’s impossible to really understand a method, including understanding its philosophical significance, without mucking about with the method and some actual data. Doing practical (and important) work on inferring causation from correlation in real-world cases of practical importance could be a necessary first step to philosophical understanding in these cases.

And these weren’t the only highlights of the workshop, which was really very good. The talks are mostly available “here”:http://socrates.berkeley.edu/%7Efitelson/few/schedule.html, though in a few cases that will only give you the slides which might not make much sense without the accompanying presentations. But there’s a few things to read over.

So far I’ve been mainly stressing what traditional epistemologists could learn from the formalists. But I think there’s also room for useful instruction in the other direction. (And for that matter from a third party. The most notable absence at each conference was anyone defending a broadly Williamsonian approach to epistemology. The presence of Tim, or a Timophile, would have been incredibly valuable at either conference I think.) In particular I think the probabilists could use thinking about contemporary work on evidence. Since we’re now in traditional epistemology mode, I’ll illustrate this with an utterly absurd example.

Jack and Jill both rent Ford Prefects, a new and not altogether troublefree breed of car, from Seattle airport. Prefects occasionally have a bug in them – 1 car in 100 has a gas gauge that returns more or less random signals whatever the state of the gas tank. Jack has one of the duds, Jill has one of the functioning cars. There’s no obvious way to tell the good cars from the duds, except that of course the duds have faulty gas gauges. But Jack and Jill only check the state of their gas tanks via the gauges. In particular, they each look at their gas gauge on leaving the airport and each see that it reads {3/4}, rather than slightly over F as it should. This is a rather pressing matter, so they each turn around to try and get the situation remedied.

We could go on with what may happen to Jack and Jill next, but let’s turn to three epistemological questions.

1) Do Jack and Jill see the same things when they look gas-gaugeward?
2) Do they get the same evidence when they look gas-gaugeward?
3) Is the evidence (if any) they get of equal quality?

As far as I can tell, everyone working on evidence using formal or probabilistic methods at FEW was committed to answering “Yes” to all three questions. But many traditional epistemologists will answer “No” to some or all of these, and they may have a point.

Now admittedly I used to think the answers here were all obviously “Yes” as well, and I still think they _might_ all be “Yes” even though I no longer think that’s obvious. But in each case the argument for “No” has some force. Here are some such arguments.

1) Jill sees that her gas tank is not full, Jack does not see this, so they do not see the same thing. (Obviously Jack does not see that _Jill’s_ gas tank is not full, but nor does he see that his own tank is not full. If it isn’t obvious I meant (1) to have the answer obviously “Yes” if they both have working gauges. On reflection that’s assuming quite a lot about the role of indexicals in perception, but I don’t think I’m on too shaky ground here.)

2) Jill gets some evidence about her gas tank, namely that it’s not full. Jack is not connected to the state of his gas tank by any reliable mechanism, so he gets no evidence whatsoever about the state of the tank. Hence they don’t get the same evidence.

3) Just like (2), except we allow that seeing something that _could_ be reliably connected to the state of the gas tank counts as evidence, it is just in this case very bad evidence, while Jill’s evidence is very good.

Maybe all these arguments are very bad, but I think they all deserve to be addressed. And I think that more interaction between the kind of people who go to scepticism conferences (or the Rutgers epistemology conference, which is a pretty similar crowd) and the formal epistemologists would lead to them being more fruitfully addressed.

Possibly more to come on developments concerning formal epistemology. For now, watch this space!

On a more personal note, the conference was lots of fun. (Except for the being as nervous as a rookie leftie facing Randy Johnson because I was delivering a wacky logic talk to a room full of real logicians. That was informative and helpful, though not exactly _fun_.) There were some surprising claims about the future of philosophy made, several references to this august website, a relatively moderate amount of drinking (especially before said logic talk) and of course lots of great philosophy and philosophers. All in all a very nice way to wrap up the semester. For those inclined towards formal epistemology, highly recommended for when it goes on the road to Austin next year.