1) I would not call atheism “rationality.” Atheism requires a certain degree of blind faith and accepting lack of evidence for religion as evidence of not-religion which is not in concordance with the principles of rationality. Perhaps “agnostic atheism” would be a more reasonable perspective. “There is a god” and “there is no god” are both non-falsifiable assertions, and I can think of few things that I would accept as corroborating evidence thereof. You cannot deduce atheism from the fact that religion does not seem to be correct. You can, however, reasonably state that while there might be a god, there probably is not.
2) Occam’s razor is not necessarily applicable sui generis. I would direct you to Elliott Sober’s excellent paper on the subject, “Let’s Razor Occam’s Razor.” It is a useful argument against the existence of a god, but certainly not definitive.
To a rationalist, “Thor doesn’t exist” and “Thor almost certainly doesn’t exist” are pretty much equivalent, and generally caused by “I have no good evidence that Thor exists and a low prior on complicated hypotheses like Thor”.
I would say they are equivalent “for all practical purposes,” but that qualifier is necessary. A low prior on a complicated hypothesis is not as relevant as one might think due to the washing out of priors (which is how we can have subjective priors such as “deciding” that the existence of Thor has low prior probability).
And as I said in another comment, you cannot call a probability of 0.9998 and a probability of 1 equivalent. If they were equivalent, your probability would = 1. If it is not, you cannot be rationally justified in making such an absolute statement.
I would point you over to the “Fallacy of Grey” if we are beginning to split hairs over the likelihood of Thor existing. You are right that nothing that is non-tautological can ever be proven with absolute certainty, and this category includes the existence of God and a predicted sunrise every day—however, to function rationally we must give claims the appropriate level of credence that supporting evidence demands; we must not turn a critique against binary black-white causality into a unitary embrace of the grey.
Viewed in this light, I don’t really think “gnostic” atheism has much use as a term, as it is nonexistent just as “gnostic” anything-non-tautological is nonexistent. Thus quibbling that the neo-atheists are in fact foolish gnostics is really a little petty, when really just about every rationalist falls into the agnostic category as a matter of principle; which still allows us to maintain a category of very very high unbelief in things like God.
You’re right, but that doesn’t mean what you’re saying is useful.
At some point, I have to make decisions, including what to say I believe. It is far less confusing to say ‘I believe that god does not exist’ than to say ‘I hold a vanishingly low belief in the class of things you refer to when you say ’God.″
Saying that you believe a thing is true, or don’t believe a thing is true, is not the same thing as saying that your probability estimates are 1 and 0, respectively. That would mean that no evidence would be adequate to change your beliefs, because you have infinite confidence, and that would obviously be stupid. The implication you’re inferring from the term ‘atheist’ doesn’t even really logically follow from the definition. I think the class of atheists that actually do mean that they slavishly and unquestioningly disbelieve in God is small enough to be safely disregarded. Most of the people who call themselves atheists simply hold very very low beliefs in God.
It’s worth noting that LW generally eschews an exclusive concern with falsification in favor of a concern with how various observations properly affect an observer’s confidence levels in various statements.
The canonical example around here is the entirely undetectable dragon in my garage: if I don’t observe phenomena that I would expect (probabilistically) to observe if there were a dragon in my garage, that doesn’t prove there’s no dragon, but it is grounds for me to reduce my confidence in the existence of such a dragon. Enough “missing” evidence causes my confidence in the dragon’s existence to drop to negligible levels.
A similar process can cause my confidence in any particular God’s existence to drop to negligible levels.
And “there’s no X” is a pretty standard way for humans to express the high confidence that there’s no X.
I think we share the same perspective on this issue. My main point regarding the existence/non-existence of a god was that one cannot say with a P of 1 that there is certainly no god. In fact, such an assertion seems to me to be absurd. However, given other evidence, we can have very low confidence in the existence of a god and very high confidence in the non-existence of a god.
However, as a scientist and philosopher of science, I cannot accept that missing evidence infers any one alternative hypothesis. This was one of the many criticisms of Karl Popper’s falsification theory: finding evidence that says research program T is not true may imply that not-T is true, may imply there is an auxiliary hypothesis in T that needs to be adjusted but that the other assumptions surrounding that theory are still acceptable, etc. If T is false and not-T is true, there is no immediately obvious standard by which to choose which of the theoretically infinite alternative theories is true.
This is getting more theoretical, though, and does not really apply to a binary problem like existence/nonexistence of god, so I’m afraid I’ve gone off on a bit of a tangent here. At base, I agree with you that the non-existence of god seems to have a probability very close to 1. However, it is not 1, and I would be loathe to say it is close enough to 1 for the difference to be “negligible.” If your probability is 1 (or 0), then it is 1 (or 0). If it is close to but not quite 1 (or 0), you are not justified in making an absolute statement.
As Lakatos wrote, it is not irrational to continue working within a degenerating research program, for such programs have been seen historically to have comebacks when new evidence is discovered. Personally, however, I’d place my money on the non-existence of god (which seems, to me, to be the progressive research program).
It’s a truism around here that you can’t say anything with P=1, on pain of being unable to subsequently change your mind given new evidence. Here’s a post along those lines.
Agreed that missing evidence doesn’t privilege any single alternative hypothesis, except in cases of strictly binary propositions. However, insofar as T2 and T1 are relevantly similar, events that lower my confidence in T1 will lower my confidence in T2 as well, so missing evidence can legitimately anti-privilege entire classes of explanation. That said, it’s important not to generalize over relevant dis-similarities between T2 and T1.
As far as “negligible”… well, enough “missing” evidence causes my confidence in a proposition to drop to a point where the expected value of behaving as though it were true is lower than the expected value of behaving as though it were false. For most propositions this is straightforward enough, but is insufficient when infinite or near-infinite utility is being ascribed to such behavior (as advocates of various gods routinely do)… human brains are not well-calibrated enough to perform sensible expected value calculations even on rare events with large utility shifts (which is one reason lotteries remain in business), let alone on vanishingly unlikely events with vast utility shifts. So when faced with propositions about vanishingly unlikely events with vast utility shifts, I’m justified in being skeptical about even performing an expected value calculation on them, if the chances of my having undue confidence in my result are higher than the chances of my getting the right result.
All right, I think I concede your point. (Not to say I will stop thinking about this issue, of course—have to be in a constant state of “crisis of belief” &c.) I also think we agree fundamentally about a great many of these points you made in this comment to begin with and perhaps I did not verbalize them coherently—such as “behaving for all practical purposes as if a given T were true” and so on. The majority of your last paragraph is new to me, however. Thanks.
For most propositions this is straightforward enough, but is insufficient when infinite or near-infinite utility is being ascribed to such behavior (as advocates of various gods routinely do)… human brains are not well-calibrated enough to perform sensible expected value calculations even on rare events with large utility shifts (which is one reason lotteries remain in business), let alone on vanishingly unlikely events with vast utility shifts. So when faced with propositions about vanishingly unlikely events with vast utility shifts, I’m justified in being skeptical about even performing an expected value calculation on them, if the chances of my having undue confidence in my result are higher than the chances of my getting the right result.
The inverted Pascal’s Wager.
or
Did you know that the first Matrix was designed to be a perfect human world? Where none suffered, where everyone would be happy. It was a disaster. No one would accept the program.
Just a few points:
1) I would not call atheism “rationality.” Atheism requires a certain degree of blind faith and accepting lack of evidence for religion as evidence of not-religion which is not in concordance with the principles of rationality. Perhaps “agnostic atheism” would be a more reasonable perspective. “There is a god” and “there is no god” are both non-falsifiable assertions, and I can think of few things that I would accept as corroborating evidence thereof. You cannot deduce atheism from the fact that religion does not seem to be correct. You can, however, reasonably state that while there might be a god, there probably is not.
2) Occam’s razor is not necessarily applicable sui generis. I would direct you to Elliott Sober’s excellent paper on the subject, “Let’s Razor Occam’s Razor.” It is a useful argument against the existence of a god, but certainly not definitive.
To a rationalist, “Thor doesn’t exist” and “Thor almost certainly doesn’t exist” are pretty much equivalent, and generally caused by “I have no good evidence that Thor exists and a low prior on complicated hypotheses like Thor”.
I both agree and disagree with you.
I would say they are equivalent “for all practical purposes,” but that qualifier is necessary. A low prior on a complicated hypothesis is not as relevant as one might think due to the washing out of priors (which is how we can have subjective priors such as “deciding” that the existence of Thor has low prior probability).
And as I said in another comment, you cannot call a probability of 0.9998 and a probability of 1 equivalent. If they were equivalent, your probability would = 1. If it is not, you cannot be rationally justified in making such an absolute statement.
I would point you over to the “Fallacy of Grey” if we are beginning to split hairs over the likelihood of Thor existing. You are right that nothing that is non-tautological can ever be proven with absolute certainty, and this category includes the existence of God and a predicted sunrise every day—however, to function rationally we must give claims the appropriate level of credence that supporting evidence demands; we must not turn a critique against binary black-white causality into a unitary embrace of the grey.
Viewed in this light, I don’t really think “gnostic” atheism has much use as a term, as it is nonexistent just as “gnostic” anything-non-tautological is nonexistent. Thus quibbling that the neo-atheists are in fact foolish gnostics is really a little petty, when really just about every rationalist falls into the agnostic category as a matter of principle; which still allows us to maintain a category of very very high unbelief in things like God.
You’re right, but that doesn’t mean what you’re saying is useful.
At some point, I have to make decisions, including what to say I believe. It is far less confusing to say ‘I believe that god does not exist’ than to say ‘I hold a vanishingly low belief in the class of things you refer to when you say ’God.″
Saying that you believe a thing is true, or don’t believe a thing is true, is not the same thing as saying that your probability estimates are 1 and 0, respectively. That would mean that no evidence would be adequate to change your beliefs, because you have infinite confidence, and that would obviously be stupid. The implication you’re inferring from the term ‘atheist’ doesn’t even really logically follow from the definition. I think the class of atheists that actually do mean that they slavishly and unquestioningly disbelieve in God is small enough to be safely disregarded. Most of the people who call themselves atheists simply hold very very low beliefs in God.
It’s worth noting that LW generally eschews an exclusive concern with falsification in favor of a concern with how various observations properly affect an observer’s confidence levels in various statements.
The canonical example around here is the entirely undetectable dragon in my garage: if I don’t observe phenomena that I would expect (probabilistically) to observe if there were a dragon in my garage, that doesn’t prove there’s no dragon, but it is grounds for me to reduce my confidence in the existence of such a dragon. Enough “missing” evidence causes my confidence in the dragon’s existence to drop to negligible levels.
A similar process can cause my confidence in any particular God’s existence to drop to negligible levels.
And “there’s no X” is a pretty standard way for humans to express the high confidence that there’s no X.
I think we share the same perspective on this issue. My main point regarding the existence/non-existence of a god was that one cannot say with a P of 1 that there is certainly no god. In fact, such an assertion seems to me to be absurd. However, given other evidence, we can have very low confidence in the existence of a god and very high confidence in the non-existence of a god.
However, as a scientist and philosopher of science, I cannot accept that missing evidence infers any one alternative hypothesis. This was one of the many criticisms of Karl Popper’s falsification theory: finding evidence that says research program T is not true may imply that not-T is true, may imply there is an auxiliary hypothesis in T that needs to be adjusted but that the other assumptions surrounding that theory are still acceptable, etc. If T is false and not-T is true, there is no immediately obvious standard by which to choose which of the theoretically infinite alternative theories is true.
This is getting more theoretical, though, and does not really apply to a binary problem like existence/nonexistence of god, so I’m afraid I’ve gone off on a bit of a tangent here. At base, I agree with you that the non-existence of god seems to have a probability very close to 1. However, it is not 1, and I would be loathe to say it is close enough to 1 for the difference to be “negligible.” If your probability is 1 (or 0), then it is 1 (or 0). If it is close to but not quite 1 (or 0), you are not justified in making an absolute statement.
As Lakatos wrote, it is not irrational to continue working within a degenerating research program, for such programs have been seen historically to have comebacks when new evidence is discovered. Personally, however, I’d place my money on the non-existence of god (which seems, to me, to be the progressive research program).
It’s a truism around here that you can’t say anything with P=1, on pain of being unable to subsequently change your mind given new evidence. Here’s a post along those lines.
Agreed that missing evidence doesn’t privilege any single alternative hypothesis, except in cases of strictly binary propositions. However, insofar as T2 and T1 are relevantly similar, events that lower my confidence in T1 will lower my confidence in T2 as well, so missing evidence can legitimately anti-privilege entire classes of explanation. That said, it’s important not to generalize over relevant dis-similarities between T2 and T1.
As far as “negligible”… well, enough “missing” evidence causes my confidence in a proposition to drop to a point where the expected value of behaving as though it were true is lower than the expected value of behaving as though it were false. For most propositions this is straightforward enough, but is insufficient when infinite or near-infinite utility is being ascribed to such behavior (as advocates of various gods routinely do)… human brains are not well-calibrated enough to perform sensible expected value calculations even on rare events with large utility shifts (which is one reason lotteries remain in business), let alone on vanishingly unlikely events with vast utility shifts. So when faced with propositions about vanishingly unlikely events with vast utility shifts, I’m justified in being skeptical about even performing an expected value calculation on them, if the chances of my having undue confidence in my result are higher than the chances of my getting the right result.
All right, I think I concede your point. (Not to say I will stop thinking about this issue, of course—have to be in a constant state of “crisis of belief” &c.) I also think we agree fundamentally about a great many of these points you made in this comment to begin with and perhaps I did not verbalize them coherently—such as “behaving for all practical purposes as if a given T were true” and so on. The majority of your last paragraph is new to me, however. Thanks.
The inverted Pascal’s Wager.
or