Regarding specialists sometimes being wrong, and in some categories being more likely to be wrong, there are two issues here. First, it might be that this should be evidence to cause us to wonder whether they are in fact wrong and the LW conventional wisdom is correct (or at least that it is nearly as obviously as correct).
Second, many of these issues becomes substantially less interesting as fields if one accepts the LW correct view. This is most obviously the issue where 79.13% of phil religion people are theists but only 13.22% of non-specialists. are theists. Simply put, philosophy of religion is much harder to justify as a subject with much content if theism is wrong.
I’m a little confused why you describe Humean as wrong. Although LW doesn’t buy into some of Hume’s ideas (e.g. the inherent unreliability of induction), a lot of what is discussed here are ideas somewhat compatible or at least not in conflict with Hume. For example, the fact that an AI might not share human values at all is pretty Humean in its approach to moral questions.
The only question which used the word “Humean” concerned the laws of nature, so none of Hume’s other views are relevant, though I admit to not being sure why RobbBB is so certain that the Humean view of laws of nature is wrong. The Humean view is sometimes called the regularity view; the short version is that it is the view that what it is for a law of nature to be true just is for there to be a pattern of events in the universe conforming to the law. The non-Humean view insists that there is something more, some additional metaphysical component, which must be present in order for it to be true that a regular pattern reveals a genuine law of nature.
One simple argument for the non-Humean view is that it seesm that some patterns are coincidences, and we need some way to distinguish the concidences from the genuine laws of nature. The standard Humean response to this is to say that there are ways for a Humean to identify something as a coincidence. A Humean identifies a coincidence by noting that the pattern fails to cohere with broader theory; laws of nature fit fairly neatly together into larger wholes up to the general theory of everything, while coincidences are patterns which in light of broader theory look more like chance. Further, the Humean will argue that when there aren’t clues like broader theoretical concerns, saying that it’s a law of nature if it’s got the right metaphysical extra and not otherwise is useless, since there’s no way to detect the metaphysical extra directly.
This is a great summary! My confidence that Humeanism is false is not extremely high; but it’s high enough for me to think philosophers of science are far too confident of Humeanism relative to the evidence. The main source of uncertainty for me here is that I’m really not clear on what it takes for something to be a ‘law of nature’. But if the essential question here is whether extremely strong correlations in fundamental physics call for further explanation, then I side with the ‘yeah, some explanation would be great’ side. The main worry is that if these events are brute coincidences, we have no Bayesian reason to expect the coincidence to continue into the future. The core intuition underlying non-Humeanism is that some simpler and more unitary mechanism is far more likely to have given rise to such empirical consistency than is such consistency to be the end of the story.
My concern with philosophers of science confidently endorsing Humeanism is that I expect this to be a case of ‘I’m supposed to think like a Scientist, and Scientists are skeptical of things, especially weird things we can’t directly observe’. Even if Humeanism itself is true, I would be very surprised if most Humeans believe in it for the right reasons.
In some ways this argument parallels the argument Eliezer has with himself over whether our universe is more like first-order logic or more like second-order; first-order logic is similar in some ways to Humeanism, because it doesn’t think we need to subsume the instances within a larger generalization in order to fully explain and predict them. (Or, more precisely, it thinks such generalizations are purely anthropocentric, human constructs for practical ends; they give us little if any reason to update in favor of any metaphysical posits.)
Hmm, I may need to reread the paper, but my understanding that they were also using Hume in the context of the question which asked which philosophers or philosophical schools people most identified with.
True, but when RobbBB comments on experts being wrong, he specifically mentions the tendency of philosophers of science to be Humeans, so I’m pretty sure he means to say they are mistaken in being Humean in this specific sense, not that it’s just generally a mistake to be influenced by Hume (I don’t get the impression that he is that down on Hume generally).
it might be that this should be evidence to cause us to wonder whether they are in fact wrong and the LW conventional wisdom is correct (or at least that it is nearly as obviously as correct).
It’s evidence for both. If philosophers disagree with us, we should be less confident that we’re right, and also less confident that they’re right. The examples I provided were ones where I had high enough priors on the issues to not be dragged into agnosticism by specialists’ disagreeing with me. (But, of course, I could still change my mind if we talked about it more. I don’t mean to encourage lock-step adherence to some vague idea of LW Consensus as a shortcut for avoiding actually evaluating these philosophical doctrines.)
Regarding Humeanism, I was voicing my own view (and Eliezer’s), not speaking for LessWrong as a whole. I’m more worried about philosophers being wrong than about their being un-LWy as such. Note that Humeanism here only refers to skepticism or reductionism about laws of ntaure; it doesn’t refer to any of Hume’s other views. (In fact, Hume himself was not a Humean in the sense used in the PhilPapers Survey. ‘Humeanism’ is like ‘Platonism’ in a modern context; a view only vaguely and indirectly inspired by the person for whom it’s named.)
Regarding specialists sometimes being wrong, and in some categories being more likely to be wrong, there are two issues here. First, it might be that this should be evidence to cause us to wonder whether they are in fact wrong and the LW conventional wisdom is correct (or at least that it is nearly as obviously as correct).
Second, many of these issues becomes substantially less interesting as fields if one accepts the LW correct view. This is most obviously the issue where 79.13% of phil religion people are theists but only 13.22% of non-specialists. are theists. Simply put, philosophy of religion is much harder to justify as a subject with much content if theism is wrong.
I’m a little confused why you describe Humean as wrong. Although LW doesn’t buy into some of Hume’s ideas (e.g. the inherent unreliability of induction), a lot of what is discussed here are ideas somewhat compatible or at least not in conflict with Hume. For example, the fact that an AI might not share human values at all is pretty Humean in its approach to moral questions.
The only question which used the word “Humean” concerned the laws of nature, so none of Hume’s other views are relevant, though I admit to not being sure why RobbBB is so certain that the Humean view of laws of nature is wrong. The Humean view is sometimes called the regularity view; the short version is that it is the view that what it is for a law of nature to be true just is for there to be a pattern of events in the universe conforming to the law. The non-Humean view insists that there is something more, some additional metaphysical component, which must be present in order for it to be true that a regular pattern reveals a genuine law of nature.
One simple argument for the non-Humean view is that it seesm that some patterns are coincidences, and we need some way to distinguish the concidences from the genuine laws of nature. The standard Humean response to this is to say that there are ways for a Humean to identify something as a coincidence. A Humean identifies a coincidence by noting that the pattern fails to cohere with broader theory; laws of nature fit fairly neatly together into larger wholes up to the general theory of everything, while coincidences are patterns which in light of broader theory look more like chance. Further, the Humean will argue that when there aren’t clues like broader theoretical concerns, saying that it’s a law of nature if it’s got the right metaphysical extra and not otherwise is useless, since there’s no way to detect the metaphysical extra directly.
Hope that helps.
Just to make everything more confusing, it turns out that David Hume was not a Humean about laws of nature.
This is a great summary! My confidence that Humeanism is false is not extremely high; but it’s high enough for me to think philosophers of science are far too confident of Humeanism relative to the evidence. The main source of uncertainty for me here is that I’m really not clear on what it takes for something to be a ‘law of nature’. But if the essential question here is whether extremely strong correlations in fundamental physics call for further explanation, then I side with the ‘yeah, some explanation would be great’ side. The main worry is that if these events are brute coincidences, we have no Bayesian reason to expect the coincidence to continue into the future. The core intuition underlying non-Humeanism is that some simpler and more unitary mechanism is far more likely to have given rise to such empirical consistency than is such consistency to be the end of the story.
My concern with philosophers of science confidently endorsing Humeanism is that I expect this to be a case of ‘I’m supposed to think like a Scientist, and Scientists are skeptical of things, especially weird things we can’t directly observe’. Even if Humeanism itself is true, I would be very surprised if most Humeans believe in it for the right reasons.
In some ways this argument parallels the argument Eliezer has with himself over whether our universe is more like first-order logic or more like second-order; first-order logic is similar in some ways to Humeanism, because it doesn’t think we need to subsume the instances within a larger generalization in order to fully explain and predict them. (Or, more precisely, it thinks such generalizations are purely anthropocentric, human constructs for practical ends; they give us little if any reason to update in favor of any metaphysical posits.)
Hmm, I may need to reread the paper, but my understanding that they were also using Hume in the context of the question which asked which philosophers or philosophical schools people most identified with.
True, but when RobbBB comments on experts being wrong, he specifically mentions the tendency of philosophers of science to be Humeans, so I’m pretty sure he means to say they are mistaken in being Humean in this specific sense, not that it’s just generally a mistake to be influenced by Hume (I don’t get the impression that he is that down on Hume generally).
Yes, thanks for pointing this out. I’m not criticizing anyone for being a follower of Hume in general. I am decidedly not down on Hume, usually.
It’s evidence for both. If philosophers disagree with us, we should be less confident that we’re right, and also less confident that they’re right. The examples I provided were ones where I had high enough priors on the issues to not be dragged into agnosticism by specialists’ disagreeing with me. (But, of course, I could still change my mind if we talked about it more. I don’t mean to encourage lock-step adherence to some vague idea of LW Consensus as a shortcut for avoiding actually evaluating these philosophical doctrines.)
Regarding Humeanism, I was voicing my own view (and Eliezer’s), not speaking for LessWrong as a whole. I’m more worried about philosophers being wrong than about their being un-LWy as such. Note that Humeanism here only refers to skepticism or reductionism about laws of ntaure; it doesn’t refer to any of Hume’s other views. (In fact, Hume himself was not a Humean in the sense used in the PhilPapers Survey. ‘Humeanism’ is like ‘Platonism’ in a modern context; a view only vaguely and indirectly inspired by the person for whom it’s named.)