Massimo and Julia visit Indianapolis for a heated debate, in this live episode of Rationally Speaking. At a symposium organized by the Center for Inquiry (CFI), they join up with John Shook, Director of Education and Senior Research fellow at the CFI, and the author of more than a dozen books on philosophy and religion.
Sparks fly as the three debate questions like: Should science-promoting organizations, like the National Center for Science Education, claim publicly that science is compatible with religion? And is philosophy incapable of telling us anything about the world?
John's pick: "Meaning and Value in a Secular Age: Why Eupraxsophy Matters—The Writings of Paul Kurtz."
Excellent debate!
ReplyDeleteMassimo, I was slightly surprised to hear you say that Ockham's razor is some sort of an aesthetic preference, then mention as an argument the fact that sometimes the non-Ockhamian hypothesis turns out to be true. Isn't Ockham's razor just another statement of the conjunction rule of probability that P(A&B) ≤ P(A)? This would explain both its usefulness and the fact that it is not exceptionless; i.e., A&B sometimes do both happen.
Ian,
Deletewell, I don't think that the likelihood of different scientific theories of different complexity being true is something to which we can easily apply Bayesian principles. There is a well documented history of more complicated theories turning out to be correct, so that I think Occam provides only a generic heuristic (better to start simple), and generally reflects a psychological preference on the part of most scientists for simple and elegant answers.
>well, I don't think that the likelihood of different scientific theories of different complexity being true is something to which we can easily apply Bayesian principles.
DeleteReally? I would argue that that is the canonical application for Bayesian methods. Probably most of the appeal of Bayes in epistemology comes from its attraction as a model of scientific progress; if it failed in that area it would not be very well motivated anymore.
>There is a well documented history of more complicated theories turning out to be correct, so that I think Occam provides only a generic heuristic (better to start simple), and generally reflects a psychological preference on the part of most scientists for simple and elegant answers.
Well, Ockham says nothing about elegance! (Which I agree is a merely aesthetic criterion.) But as I mentioned above, the fact that more complicated theories sometimes turn out to be correct is *completely expected* if you cash out Ockham as equivalent to the conjunction rule; i.e., a constraint on priors for scientific hypotheses.
As you know, for any set of empirical data there is an infinity of empirically identical hypotheses that predict it - but those hypotheses are not all equally likely! That is exactly where priors, and Ockham, come into play - more complex hypotheses are automatically less probable by the axioms of probability (ceteris paribus, unless background information makes them more likely).
So scientists' 'psychological preference' for simple hypotheses is no mere arbitrary preference, it is forced on them by the laws of probability & is exactly as well-grounded as those laws.
Maybe an example would help us to clarify things? I bet as a philosopher of science you'll be able to come up with a better example than I would.
Ian,
DeleteThough I suspect there is one, I think you overplay the connection between simplicity and probability. In any case, if you have not already, seeing as you have interests along these lines, you may want to review Kevin Kelly's work on simplicity (on this, see: http://www.andrew.cmu.edu/user/kk3n/homepage/kelly.html), and in particular his paper on simplicity and probability (which can be read here: http://repository.cmu.edu/philosophy/369/).
That said, I disagree with Massimo that simplicity is at base an aesthetic preference. For instance, there are practical reasons why we might want to keep our hypotheses parsimonious. For any hypothesis, H, we can deduce observational consequences, O, against which we can test H. So, we may reason thus: 'If H, O; not-O; Ergo, not-H'. However, with H there are many auxiliary hypotheses, A1, A2, ... An, (e.g., about experimental design, instrument reliability, observational conditions, etc.) that could have accounted for not-O.
In brief, we cannot infer not-H from 'If H, O and not-O' since the antecedent is properly a conjunction such that the conditional is more accurately represented as 'If {H, A1, A2,... An}, then O'. If not-O, it may be that H is false or some subset of auxiliary hypotheses may be false or both. We will have to eliminate properly Holmesian fashion the likely culprits before we can infer not-H given not-O, and the simpler our hypothesis and the auxiliary hypotheses, the easier that will be.
I commented on the Podcast post, but might as well re-ask here. What type of empirical data are you thinking of when you say there is no empirical data about the utility of accommodating religious belief? I may be phrasing this completely wrong, but I think you should get the gist of my question.
ReplyDeleteMark,
ReplyDeleteI don't think that's my position, that sounds more like John. I'm sure there is social utility in religious belief, which is why I'm often cast with the infamous "accommodationists" by the New Atheists.
Ian,
Bayes theorem is correct in terms of a theorem in probability theory. But that's not enough to make it an epistemological theory. Indeed, there has been a vigorous discussion in philosophy of science about the validity or limits of "Bayesianism" in scientific epistemology. You may get a sense of it from the chapter on Bayesianism in Alan Chalmers' What is This Thing Called Science?
> as I mentioned above, the fact that more complicated theories sometimes turn out to be correct is *completely expected* if you cash out Ockham as equivalent to the conjunction rule <
Maybe, but the principle is of no use in any specific case. Do we expect this particular scientific theory to be simpler than the alternatives? And which alternatives, precisely?
> Maybe an example would help us to clarify things? <
Lee Smolin (in The Trouble with Physics) has an entire chapter detailing several less simple theories in modern physics that turned out to be (more) correct than their simpler rivals.
I had to go back and listen again. You did say there is some empirical evidence about how people shed beliefs - slowly and piecemeal. Shook said the last 400 years of religion's failure is empirical evidence. I thought you and Julie said at the same time "We don't have empirical evidence about that" just before Shook's statement. But I didn't find it. Whoops.
ReplyDeleteThis guy is a pure scientism hack. His dismissal of math is that he doesn't understand how numbers are defined as counting objects. Similarly he is not the word police but John berates Massimo for confusion caused over his own arbitrary definition of words.
ReplyDeleteThe points about mathematical platonism went right over his head. He should have been forced to explain how his beloved engineering which uses calculus and linear algebra is able to correctly predict physical reactions of things when those maths are based on proofs, not any kind of empirical data.
Where did you get the concept of objects? or counting for that matter?
DeleteMassimo,
ReplyDeleteI wonder if your original disagreement with Eugenie Scott is echoed in your argument with John. Back in those days, though, I think you would have agreed with John that science should brook no quarter, even for pedagogic reasons.
Personally, I think he is wrong on political grounds, but I find his rejection of mathematical Platonism interesting.
OneDay,
Deletewell, concerning Genie, I ended up thinking that she is right, there is a principle distinction btw methodological and philosophical naturalism. As for mathematical Platonism, as you might have surmised from this blog, the more I think about it the more it makes sense to me...