Sorry if I'm late to the party.
If the argument is that no evidence of any form could ever change the Austrian's probability estimate of a certain theory, I charge the Austrian is either being overconfident or violating the conservation of expected evidence.
Not true; such a position can be perfectly Bayesian: You would just believe that P(E|~H) = P(E|H). In other words, "On average, learning E tells you nothing about H, and vice versa." Such an H wouldn't be very useful, but it's Bayesian. (I had to point this out on a post by Bob Murphy critical of Bryan Caplan, but I can't find it ATM.) All Bayes prevents you from doing is taking both E and ~E as evidence for H. That is, believing both that P(E|H) > P(E|~H) and P(~E|H) > P(~E|~H). That would violate conservation of probability.
By saying "it's not a controlled experiment so it doesn't count," the Austrian misses out on a key insight about how people can be milked
Please refer to the OB discussion on the Allais paradox, where the participants' actions were justified quite well by poster "gray area" early on, and never answered by Eliezer_Yudkowsky. Long story short: the choice one is making fundamentally changes depending on whether it's one-shot or indefinitely repeated. The money pump doesn't manifest until you actually offer the choice multiple times at which point people not surprisingly pick up on it. But that money pump is just not happening in the one-shot version; ergo, no money is being pumped and nobody is being cheated, nor liking it.
Also, I'd like to comment on the price control example. While I certainly don't want to defend the Austrians, I also don't see how this is a point against them: rather, it's a confirmatory example.
he proclaims that price controls will lead to rationing by non-price means. But this is only true if the provider of the good in question is attempting to maximize profit; if the producer is willing to take a hit in the wallet out of the goodness of his heart for his customers' well-being, as Mises' tautological definition of self-interest allows, a small price ceiling could conceivably have no effect.
If the producer is continuing to sell at the ceiling rate, that is still non-price rationing because he is rationing the good based on who gets there first or who he likes most, etc, exactly the thing Austrians are warning about, and exactly the kind of non-trivial insight many people still don't appreciate.
So I agree with the general thrust of this post, but this is a bad example.
I agree money pumps don't work forever. Experimental economics bears out, to some degree, that rationality is often a product of incentives. But if people can be milked in the very short run (the Allais Paradox is of course not the only example), this might have some bearing on economic theory.
I disagree that a producer who produces at a ceiling rate rations based on who gets there first. The producer could increase supply in response to the price control. The only thing limiting him from doing so is the size of his bank account.
Response to: The uniquely awful example of theism
Why is theism such an ever-present example of irrationality in this community? I think ciphergoth overstates the case. Even theism is not completely immune to evidence, as the acceptance of, say, evolution by so many denominations over time will testify. Theism is a useful whipping boy because it needs no introduction.
But I think the case is overstated for another reason. There are terrible epistemologies out there that are just as bad as theism's. Allow me to tell you a tale, of how I gave up my religion and my association with a school of economics at the same time.
I grew up in a southern Presbyterian church in the U.S. While I was taught standard pseudo-evidential defenses for belief, such as "creation science" and standard critiques of evolution, my church was stringently anti-evidentialist. Their preferred apologetic was something called presuppositionalism. It's certainly a minority apologetic among major defenders of Christianity today, especially compared to the cosmological or morality arguments. But it's a particularly rigorous attempt to defend beliefs against evidence nonetheless.
Presuppositionalism (in some forms) hangs on the problem of induction. We cannot ultimately justify any of our beliefs without first making some assumptions, otherwise we end in solipsism. Christianity, then, justifies itself not on evidence, but on internal consistency. It is ok for an argument to be ultimately circular, because all arguments are ultimately circular. Christianity alone maintains perfect worldview consistency when examined through this lens, and is therefore correct.
Since I've spent a lot of time thinking about this--it can take a considerable effort to change one's mind, after all--I can imagine innumerable things wrong with it, but they're not the focus of this entry. First, I just want to note how close it is to a kind of intro-level Bayesian understanding. Bayesians admit that we must have priors, that it's indeed nonsense to think we can even have an argument with one who doesn't. We must ultimately admit that certain justifications are going to be either recursive or based on priors. We believe that we should update our priors based on evidence, but there's nothing in the math that tells us we can't start with a prior for some position of 0% or 100%. (There is something in the math that tells us such probability assignments are very bad ideas, and we have more than enough cognitive bias literature that tells us we shouldn't be so damn overconfident. But then, what if you have a prior that keeps you from accepting such evidence?) It doesn't have any of the mathematical rigor, but it comes very close on a few major points.
This is why Bayesianism appealed to me. It seemed similar to the supposedly deep argument I understood for God's existence, like something I could actually work with. (This is why, I think, anti-religion Overcoming Bias posts didn't throw me into defense mode.) This is also why I used to find Austrian economics so compelling.
For those who aren't familiar, Austrian economics is a radical free-market school, the intellectual product of Ludwig von Mises, Friedrich von Hayek, and Murray Rothbard. Before I continue, in hopes of taking any Austrian economists reading this out of defense mode: I still find many Austrian insights useful, I admire Hayek for his work on knowledge and institutions, and Mises for the economic calculation argument. But the first section on epistemology in Mises' magnum opus, Human Action, is probably the best example of Dark Side Epistemology I have yet seen outside of religious apologetics or standard woo-woo. What does economics (or in Mises' case, praxeology, an expanded science of all human action that seeks to understand more than resource allocation) investigate? After excluding psychology, Mises tells us,
Well, ok. So how does economics tell us anything at all?
In other words, the assumptions built into economics (which is a subset of praxeology)--people have preferences, are selfish (in the tautological sense--even altruist acts are self-serving to Mises), and they take rational action to satisfy those preferences--are unquestionable, ultimate givens. No evidence could ever confirm or disconfirm the predictions of economics, because it is an a priori science, just like math or logic. It is deductive--it starts from some assumptions, and its case rests on those assumptions alone, not on any evidence. (And he has a word for those of us seeking instances of human irrationality. On page 103, he claims out that any sign of preference reversals can never be considered irrationality, because preferences cannot be considered stable, even across spans of a few seconds. If your by-the-second preference changing leads you to be pumped for money, so be it. You're still by assumption a rational actor, satisfying his desires.)
You can understand why I think this sounds so similar to presuppositionalism. And, if you've been following Overcoming Bias, you can see how a Bayesian would differ from these views.
I saw the same problems with presuppositionalism as I did Mises' epistemology. So what if it's deductive? What if your deductive logic doesn't conform to the real world? This could be true of math just as well as economics. What if 2 + 2 didn't really equal 4 in our world? Could there be any way to convince you? If the answer is no, then aren't you just starting from the bottom line? If your deductively valid economic argument makes a prediction that is observed to never be true in the real world, would this not affect your rating of your deductions' usefulness? If your deductions are non-disprovable, why do you make so many claims regarding their predictive value? What does your logic not predict?
To really solidify the feeling that Mises' predictions about economics are comparable to the Bible's predictions about how the world works, consider the following. As I mentioned, Mises defines self-interest tautologically:
However, Mises specifically predicts economic outcomes based on self-interest as, well, actual self-interest. For instance, on page 763, he proclaims that price controls will lead to rationing by non-price means. But this is only true if the provider of the good in question is attempting to maximize profit; if the producer is willing to take a hit in the wallet out of the goodness of his heart for his customers' well-being, as Mises' tautological definition of self-interest allows, a small price ceiling could conceivably have no effect.
So when are we to believe Mises? When he says economics is a deductive logic that can never be tested in the real world, or when he makes predictions that can be tested in the real world? When should we believe presuppositional apologists? When they claim that "the Bible is the word of God" is an ultimate given, or when they tell us all about miracles (evidence for God) that we can test in the real world (by finding evidence for a global flood)?
The insistence on placing assumptions further and further away from our real ultimate givens, our real recursions, our real mystical priors, is a dark side epistemology. If we can devise a test for one of our assumptions, by golly, as rationalists we're called to test it. If that assumption fails, we have to perform a proper Bayesian update. We have to use all of our evidence available to us.
So to answer what other forms of irrationality we can regularly cite, I'd like to nominate Austrian economics, or at least those of its followers who still eschew the introduction of statistics, behavorial economics, or experimental economics into the discipline. It certainly isn't as pervasive as religion. It's a very minor branch of a specific discipline. Not all of its conclusions are wrong, but I think there's at least a little evidence of dial-cranking in Austrianism. And I think its epistemology is quite awful, as awful as the most evidence-defying justification for theism.
Reference: Ludwig von Mises. Human Action. San Francisco: Fox & Wilkes, 1996.