Your conception of "theism" -- a tremendously broad concept -- is laughably caricatured and narrow, and it pollutes whatever argument you're trying to make: absolutely none of the logic in the above post follows in the way you think it does.

I don't believe this is a valid thought in this form, or maybe you failed to formalize your intuition enough to communicate it. You list a few specific failure modes, which I don't believe can cover enough of the theistic people to reduce the probability of a theistic person producing valid singularity thinking down to nothingness. Also, some of these failure modes overlap with related failure modes of non-theistic people, thus not figuring into the likelihood ratio as much as they would otherwise.

What is the rational significance (significance in the context of rationality) of a belief being rational or irrational if it has no correlation with rational or irrational outcomes? [I really should qualify here that I am referring to some kind of theism which is consistent with, though not supported by, all empirical evidence.]

I'm not sure what you mean by rational or irrational outcomes; there are only outcomes that a belief predicts. If a belief predicts results identical to its own negation (that is, does not correlate with results at all), it is useless and can be safely culled by Occam's Razor.

Also, the only theistic hypotheses thus far that have not produced poorer results than some non-theistic hypothesis are ones positing supernatural entities that either do not interact measurably with the universe or do so only in ways indistinguishable from random chance, neither of which is a particularly common idea among actual theists, so it's something of a distraction to limit ourselves only to theistic beliefs that are not reflective of a typical theist.

It's because ("Theism is evidence that a person is irrational.") theism is evidence that is very easily screened off by other easily noted characteristics. Suppose for the sake of argument that given no other knowledge about a person than their label "theist" rather than "atheist" it is more likely they will be wrong about some other subject than if their label was "atheist".

Fine and dandy, but trace the flow of evidence through the causal diagram: "theist" is less likely given that "they're rational", so now it's more likely that "they're irrational". In particular, by irrational here I mean they have some set of cognitive algorithms not shared by all humans which makes them wrong about many subjects. This then directly propagates evidence that they will be wrong about some other specific subject. But it is screened off by evidence that these counterfactual cognitive algorithms do not in fact make them be wrong about that other specific subject. And that evidence is readily gathered by reading a post or two of theirs on the subject in question.

Everyone above a certain baseline level of mental functioning believes at least some irrational and/or false things.... including you. Some of the things you believe are at least as irrational as theism, and some subset of them have at least as much influence on your behavior as theism does on the average theist.

This is useless blurring, Fallacy of Gray.

"...what is some evidence that theists are less [rational] than atheists are?" is an incomplete question.

(tl;dr By talking about theists as a group, we are organizing people around their belief in something false that they generally should not believe in that both causes and is correlated with more general irrationality. Other than the criteria we used to organize the group, we shouldn't expect to find many other universals, just significant patterns with exceptions.)

Are all theists less rational than all atheists are? Obviously not, under any important definition, for the same reason each person who eats 4000 calories and less than 50g protein daily is not less healthy than each person who eats fewer calories and more protein, and each person looking at a Japanese newspaper does not speak better Japanese than each person not looking at a Japanese newspaper speaks Japanese.

We can still say important things about the basis by which we organized people into these groups. They can have both direct causal effects and statistical significance from indirect links to other measurable things. For example, looking at a Japanese newspaper can cause one to get better at speaking Japanese, and looking at a Japanese newspaper is correlated with having Japanese relatives who help one learn Japanese.

Finally: all else equal, looking at a Japanese newspaper is better than nothing for learning Japanese, and it's also better than what most people are doing now for learning Japanese.

If we're organizing people into "theist" group and a second group made of everyone else, "theists are less [rational] than atheists are," represents some different notions that have different proper responses among them.

If the assertion is "being a theist is correlated with being irrational (and/or playing the banjo, etc.)," then that claim needs to defer to science and new evidence, as I think you are saying.

I say "defer to" because there is an appropriate confidence someone with my amount of evidence should have in the claim. I feel very comfortable claiming that the top contributors to lesswrong are almost certainly not also the top contributors to the magazine Seventeen, despite a dearth of scientific studies on the subject.

It may be worthwhile to discuss the amount of confidence someone with a certain amount of evidence should have in a specific claim. My first response to a claim like "People who believe the soul influences some human speech (or theists, or whoever) are, on average, as rational as those who believe speech is not influenced by a soul," or "The (first? I'm not sure what conspiracies are popular) moon landing was faked," is to ask about what evidence the claimant currently possesses and how they process it. This is often more important than determining the truth of the original claim, which often will be best determined by gathering new evidence. In such a case, what's really being discussed is not the truth of the original proposition, but the reasonableness of the original statement, so no evidence on its truth is relevant.

If the assertion is "being a theist causes irrationality," truths are entangled. That's not a dogma I cling to, and each individual has other influences in his or her life that may make them an exception, but I'd like to hear some kind of response to those arguments or I won't feel obliged to go looking for evidence (unless something important hinges on my being right).

This depends on theism being irrational, which I think it is for most people - not having conducted studies, of course. For many people, theism is rational, particularly the very young, who should notice a pattern forming in which their parents are eventually right about things the child does not understand because they are too complex.

This does not depend on acts (such as thoughts) designed to induce a belief in theism being irrational.

If the assertion is "all else equal, a person with a given set of beliefs is more rational without the additional belief of theism," then yes, on average...if we have organized all human minds by their belief in a proposition that most should think false, then those who are inappropriately theist are many, those who are inappropriately atheist are few, those who are appropriately theist are few, those who are appropriately atheist are comparatively many.

If the assertion is "based on the knowledge held by the reader of this sentence, he or she is almost certainly being irrational if he or she is theistic," that is true with a good deal of help from selection bias, but one could say a similar thing about American adults.

Now that Roko's actually explained Roko::smart, we don't need to keep arguing about what counts as "smart" vs. Roko::smart. It's enough to note that when Roko uses the term, the above nonstandard definition is what is meant. Let's not argue over semantics if we don't have to.

It would seem to follow from this that you don't think the following people can be "smart":

• children
• people who have not been educated on the subjects you mention
• people who are good at partitioning, possibly because of a considered belief in fideism
• people who have conflicting desires and have not worked out which they prefer to endorse on a second-order level
• people who self-deceive on some level, or avoid thinking about the subject of religion very hard, in order to achieve a good quality of life, after having established that they consider the quality of life their top priority

Am I reading you incorrectly?

Well, the religious claim is incompatible with our observations - wildly so.

Roko, could you give more detail in your reasoning here? Is any religious claim incompatible with observations, or are you thinking of a specific one?

You know, before Darwin, the Argument from Design really was a good reason to accept some form of theism, although most versions of Christianity should still have been considered stupid. (To be blunt, the world looks more like it was designed by a group of assholes like the Greek gods than it would if it really was made by the single "loving God" of the New Testament.) People were only aware of one kind of optimization process - human intelligence - that was capable of creating complicated artifacts, so it was reasonable to believe that the optimized artifacts present in the natural world were also the product of an intelligent designer. Hence, the Deism of Thomas Jefferson and his contemporaries.

Then humans discovered the alien god and theism became much, much less rational.