IMO there may be one Bayesian who falls for this, and there may be one frequentist who is struck by lightning twice.
depending on your prior that some unknown person is mailing you tomorrow‘s stock prices (probably very low) there is some number of correct predictions which should convince you this is true. If that prior weight is low enough (and it should be tiny), it might be that the number of correct predictions is so high that the entire world’s population of Bayesians is not enough for one Bayesian who falls for it. But if there IS one bayesian who falls for it, that’s bad luck for her, and seems like no argument against bayesianism.
I think your intuition that this is a problem comes from the idea that “oh now all the frequentists can start running this scam and money pump all the bayesians.” but if you actually do the calculation, it won’t work out that way, because the scams interfere with each other. When you receive the 101st (first) letter from a new scammer, you don’t bother to open it because your prior that it’s a scam has increased a lot; all the previous cold letters have turned out to be scams.
When you receive the 101st (first) letter from a new scammer, you don’t bother to open it because your prior that it’s a scam has increased a lot; all the previous cold letters have turned out to be scams.
Already there. The advice I always give people asking about a dodgy-looking message they’ve received is “IT’S A SCAM, IT’S ALWAYS A SCAM.”
One you accept that IT’S A SCAM, IT’S ALWAYS A SCAM, you can then at your leisure speculate on how the scam works, i.e. what was the generating process that created the message, but it doesn’t really matter if you can’t work out how the scam works, because IT’S A SCAM, IT’S ALWAYS A SCAM.
I’m not actually that interested in the scam, more how Bayesianism handles the problem. If we assume the Bayesian has reasonable priors and isn’t naive, then your answer makes sense. But when we are talking about science and the frontier of knowledge, we don’t have that luxury.
I think I can demonstrate this with a much more abstract problem: how can I use Bayesianism to evaluate multiple schools of epistemology to find the best one? Can Bayesianism come to a definitive answer or will there always be a non-zero probability that any of the schools are the correct one? (assuming any of them are)
While I’m asking this to demonstrate that bad priors and naivety are the default state, I’m also genuinely interested in the answer.
how can I use Bayesianism to evaluate multiple schools of epistemology to find the best one?
The same way you would use any school of epistemology to evaluate multiple schools of epistemology (including itself). Start from where you are, for there is no other way to begin, and see where it gets you.
Okay, so what events or things do you pay attention to for updates relating to Bayesianism? Like say I’m a Bayesian with random priors, except for P(Bayesianism) which is say 0.51 currently. What kind of things do you observe that cause updates and why/how?
Sorry for the delayed replay. My comments are being held for review.
I don’t think P(my epistemology) is a practically useful concept.
Various people have suggested paradoxes of Bayesian reasoning. The few I’ve looked in detail at, I don’t think hold water. But that’s the sort of thing to study.
Does critical fallibilism have anything to say about itself? If it gives itself a full bill of health, that’s no more useful than the statement “This sentence is true”, which can consistently be assigned any truth value.
I don’t think P(my epistemology) is a practically useful concept.
This is surprising to me—I’ve always thought that whatever our understanding of knowledge was should apply to itself. If it doesn’t, how do you know that your epistemology is right? An epistemology is the thing that’s meant to give you the answer to the question. So it seems like a problem if an epistemology can’t apply to itself.
One response might be ‘Bayesianism is about science, not philosophy’, but then on what basis are you adopting bayesianism? if that decision is supported by some other epistemology, then why not use that instead? (it seems like it might be more powerful because it can do something bayesianism can’t)
Does critical fallibilism have anything to say about itself? If it gives itself a full bill of health, that’s no more useful than the statement “This sentence is true”, which can consistently be assigned any truth value.
Yeah it does, though a ‘full bill of health’ might mean something different. CF says that we should use the best ideas we have access to. “Best” means that it works to achieve a goal and is unrefuted (no unanswered decisive criticisms). CF claims it has no decisive criticisms (at least insofar as the epistemology is concerned), but does not claim such criticisms are impossible (it’s fallibilist after all). CF is Elliot Temple’s school and he is very willing to debate (which means a forum discussion).
Debate is important because it enables errors to be corrected. If an intellectual is wrong, and a critic has an important correction, then refusing to debate is one of the common ways the intellectual can stay wrong.
One of goals of CF-ists is to find and debate people who disagree with you because that’s the best way to get criticisms, which are an integral part of error correction and improvement.
But also the world isn’t very good for debate, at least not at a high level. It’s hard to find people and forums who will continue a discussion to it’s conclusion. So maybe CF isn’t as good as it could be if there was more debate.
If someone knew of a criticism of CF, then Elliot would want to know. If he disagreed, he’d want a public discussion about it.
Arguably, using Bayesianism to evaluate schools of epistemology (which may not be Bayesian) is a type error. However, I think I still endorse assigning tentative probabilities to beliefs about epistemology and updating them by Bayes rule in practice, so I think I don’t have a full answer for you (only a bunch of disorganized thoughts; it’s a deep question).
While I’m asking this to demonstrate that bad priors and naivety are the default state, I’m also genuinely interested in the answer.
It’s not quite clear to me what you’re getting at.
Arguably, using Bayesianism to evaluate schools of epistemology (which may not be Bayesian) is a type error.
If it’s a type error is that because bayesianism’s domain does not extend to all knowledge?
If bayesianism doesn’t cover all knowledge that would make it an incomplete epistemology. This is problematic for a few reasons (so I’m not sure this is what you mean); one is that it becomes unreliable for any knowledge because it cannot tell you what you are missing or how that missing part is involved in meaningful judgements.
If bayesianism does cover all knowledge than either epistemology (and bayesianism) seemingly doesn’t count as knowledge or we have a contradiction.
Or maybe I’m misunderstanding what you mean?
Personally, I’ve long thought that an epistemology should be able to evaluate itself (and other epistemic schools/ideas). This belief was quite formative for me and is one reason that I gravitated to fallibilism (broadly) since it seemed like any ‘true’ theory of epistemology would have this quality.
While I’m asking this to demonstrate that bad priors and naivety are the default state, I’m also genuinely interested in the answer.
It’s not quite clear to me what you’re getting at.
All bayesian answers to the original question (that I’ve seen in this thread) assume the bayesian already has good priors and isn’t naive. But that isn’t how we find ourselves in the world. We start from bad priors (random or worse) and naivety. So, while I agree real-world bayesians can avoid being scammed, I am not sure that bayesianism avoids it since the way of avoiding it is to have already obtained the knowledge you need to avoid it. I’m trying to see if there’s something more fundamental. That real-world bayesians don’t get scammed isn’t really a surprise, one doesn’t need a correct understanding of epistemology to reason about the world—e.g., we were able to do science and make progress before we understood science—but if we are concerned about whether bayesianism is correct or not, then it does matter.
I’m interested in how these kinds of issues (novel unintuitive systems) are dealt with when one doesn’t have obvious crutches like a known good state to start from.
The reason for asking about using Bayesianism to evaluate schools of epistemology is because epistemology is notoriously hard to make progress in. I was hoping that the idea we start with bad priors and naivety was more obvious in this context.
Hope that helps explain what I’m getting at.
Sorry for the delayed replay. My comments are being held for review.
epistemology is not complete as a project! I think meta-epistemology in particular is still pretty confusing. i suspect that the pragmatically right approach still involves Bayesian updating, because Bayesian updating is just correct in many situations.
I think you’re using the word “Bayesian” to mean “someone who believes all knowledge can be derived from Bayesian updating” and this is not really standard. There’s a lot of types of Bayesian, e.g. most Bayesians just favor Bayesian statistical methods. The claim that we ought to use Bayesian updating is probably consistent with many epistemologies.
Now you’re arguing that you have to learn to not trust mail scams. But this is true whether or not youre Bayesian, for standard no free lunch reasons. Bayesianism tells you how to work with the knowledge that you have. I feel like you’re being uncharitable here.
IMO there may be one Bayesian who falls for this, and there may be one frequentist who is struck by lightning twice.
depending on your prior that some unknown person is mailing you tomorrow‘s stock prices (probably very low) there is some number of correct predictions which should convince you this is true. If that prior weight is low enough (and it should be tiny), it might be that the number of correct predictions is so high that the entire world’s population of Bayesians is not enough for one Bayesian who falls for it. But if there IS one bayesian who falls for it, that’s bad luck for her, and seems like no argument against bayesianism.
I think your intuition that this is a problem comes from the idea that “oh now all the frequentists can start running this scam and money pump all the bayesians.” but if you actually do the calculation, it won’t work out that way, because the scams interfere with each other. When you receive the 101st (first) letter from a new scammer, you don’t bother to open it because your prior that it’s a scam has increased a lot; all the previous cold letters have turned out to be scams.
Already there. The advice I always give people asking about a dodgy-looking message they’ve received is “IT’S A SCAM, IT’S ALWAYS A SCAM.”
One you accept that IT’S A SCAM, IT’S ALWAYS A SCAM, you can then at your leisure speculate on how the scam works, i.e. what was the generating process that created the message, but it doesn’t really matter if you can’t work out how the scam works, because IT’S A SCAM, IT’S ALWAYS A SCAM.
I’m not actually that interested in the scam, more how Bayesianism handles the problem. If we assume the Bayesian has reasonable priors and isn’t naive, then your answer makes sense. But when we are talking about science and the frontier of knowledge, we don’t have that luxury.
I think I can demonstrate this with a much more abstract problem: how can I use Bayesianism to evaluate multiple schools of epistemology to find the best one? Can Bayesianism come to a definitive answer or will there always be a non-zero probability that any of the schools are the correct one? (assuming any of them are)
While I’m asking this to demonstrate that bad priors and naivety are the default state, I’m also genuinely interested in the answer.
The same way you would use any school of epistemology to evaluate multiple schools of epistemology (including itself). Start from where you are, for there is no other way to begin, and see where it gets you.
Okay, so what events or things do you pay attention to for updates relating to Bayesianism? Like say I’m a Bayesian with random priors, except for
P(Bayesianism)which is say 0.51 currently. What kind of things do you observe that cause updates and why/how?Sorry for the delayed replay. My comments are being held for review.
I don’t think P(my epistemology) is a practically useful concept.
Various people have suggested paradoxes of Bayesian reasoning. The few I’ve looked in detail at, I don’t think hold water. But that’s the sort of thing to study.
Does critical fallibilism have anything to say about itself? If it gives itself a full bill of health, that’s no more useful than the statement “This sentence is true”, which can consistently be assigned any truth value.
This is surprising to me—I’ve always thought that whatever our understanding of knowledge was should apply to itself. If it doesn’t, how do you know that your epistemology is right? An epistemology is the thing that’s meant to give you the answer to the question. So it seems like a problem if an epistemology can’t apply to itself.
One response might be ‘Bayesianism is about science, not philosophy’, but then on what basis are you adopting bayesianism? if that decision is supported by some other epistemology, then why not use that instead? (it seems like it might be more powerful because it can do something bayesianism can’t)
Yeah it does, though a ‘full bill of health’ might mean something different. CF says that we should use the best ideas we have access to. “Best” means that it works to achieve a goal and is unrefuted (no unanswered decisive criticisms). CF claims it has no decisive criticisms (at least insofar as the epistemology is concerned), but does not claim such criticisms are impossible (it’s fallibilist after all). CF is Elliot Temple’s school and he is very willing to debate (which means a forum discussion).
From Debate Policies Introduction, he says:
One of goals of CF-ists is to find and debate people who disagree with you because that’s the best way to get criticisms, which are an integral part of error correction and improvement.
But also the world isn’t very good for debate, at least not at a high level. It’s hard to find people and forums who will continue a discussion to it’s conclusion. So maybe CF isn’t as good as it could be if there was more debate.
If someone knew of a criticism of CF, then Elliot would want to know. If he disagreed, he’d want a public discussion about it.
I can expand on anything if you have questions.
Arguably, using Bayesianism to evaluate schools of epistemology (which may not be Bayesian) is a type error. However, I think I still endorse assigning tentative probabilities to beliefs about epistemology and updating them by Bayes rule in practice, so I think I don’t have a full answer for you (only a bunch of disorganized thoughts; it’s a deep question).
It’s not quite clear to me what you’re getting at.
If it’s a type error is that because bayesianism’s domain does not extend to all knowledge?
If bayesianism doesn’t cover all knowledge that would make it an incomplete epistemology. This is problematic for a few reasons (so I’m not sure this is what you mean); one is that it becomes unreliable for any knowledge because it cannot tell you what you are missing or how that missing part is involved in meaningful judgements.
If bayesianism does cover all knowledge than either epistemology (and bayesianism) seemingly doesn’t count as knowledge or we have a contradiction.
Or maybe I’m misunderstanding what you mean?
Personally, I’ve long thought that an epistemology should be able to evaluate itself (and other epistemic schools/ideas). This belief was quite formative for me and is one reason that I gravitated to fallibilism (broadly) since it seemed like any ‘true’ theory of epistemology would have this quality.
All bayesian answers to the original question (that I’ve seen in this thread) assume the bayesian already has good priors and isn’t naive. But that isn’t how we find ourselves in the world. We start from bad priors (random or worse) and naivety. So, while I agree real-world bayesians can avoid being scammed, I am not sure that bayesianism avoids it since the way of avoiding it is to have already obtained the knowledge you need to avoid it. I’m trying to see if there’s something more fundamental. That real-world bayesians don’t get scammed isn’t really a surprise, one doesn’t need a correct understanding of epistemology to reason about the world—e.g., we were able to do science and make progress before we understood science—but if we are concerned about whether bayesianism is correct or not, then it does matter.
I’m interested in how these kinds of issues (novel unintuitive systems) are dealt with when one doesn’t have obvious crutches like a known good state to start from.
The reason for asking about using Bayesianism to evaluate schools of epistemology is because epistemology is notoriously hard to make progress in. I was hoping that the idea we start with bad priors and naivety was more obvious in this context.
Hope that helps explain what I’m getting at.
Sorry for the delayed replay. My comments are being held for review.
epistemology is not complete as a project! I think meta-epistemology in particular is still pretty confusing. i suspect that the pragmatically right approach still involves Bayesian updating, because Bayesian updating is just correct in many situations.
I think you’re using the word “Bayesian” to mean “someone who believes all knowledge can be derived from Bayesian updating” and this is not really standard. There’s a lot of types of Bayesian, e.g. most Bayesians just favor Bayesian statistical methods. The claim that we ought to use Bayesian updating is probably consistent with many epistemologies.
Now you’re arguing that you have to learn to not trust mail scams. But this is true whether or not youre Bayesian, for standard no free lunch reasons. Bayesianism tells you how to work with the knowledge that you have. I feel like you’re being uncharitable here.