Re: knowledge of priors - are you just rehashing the problem of skepticism? If you're infinitely skeptical (eg of logic itself), then not even God can solve your problem (any proof of the existence of God depends on logic). If you're less than infinitely skeptical, I think understanding that a few hyperpriors were instilled by evolution which had an incentive to get them right, plus https://www.lesswrong.com/posts/46qnWRSR7L2eyNbMA/the-lens-that-sees-its-flaws , gets you the rest of the way. I agree you can always argue that evolution somehow found a perfectly consistent set of priors which works for all pragmatic purposes but isn't true, but at some point this because its own "psychophysical harmony" problem (why are our false priors so perfectly consistent with true reality?) to the point where the probability becomes vanishingly low.
I'm not sure why you say conscious states are epiphenomenal even if epiphenomenalism is false - you just seem to be transparently pushing epiphenomenalism. But epiphenomenalism doesn't make sense, because we *know* our conscious states have actions in the world - most obviously, the action of philosophers writing "Hey, we have conscious states, what's up with that?" The same fact about pain that causes you to write the words "pain is unpleasant" in your essay on psychophysical harmony also causes you to behave as if pain is unpleasant - they're both downstream of the actual unpleasantness of pain.
Yes, I'm asserting the thing that you call UDASSA, and not the thing that obviously doesn't work. Your "objections" to it don't make sense. You say "If the UDASSA view is to maintain a normalizable probability distribution—one like the probability distribution .5, .25, .125, etc that sums to 1—it must be that there exist some finite number of worlds such that you should think with probability 99.999999999% that you’re in one of those worlds. However, literally 100% of people do not find themselves in those worlds. This is an odd result."
This isn't an "odd result" - it's just how math works! If you divide the interval from 0 to 1 into regions, such that one region includes the first half of the space, the next the next 1/4, the next the next 1/8, etc, then you can find a finite number of regions such that 99.9999999% of the interval is in those regions, even though literally 100% of regions aren't in the interval. This may be "odd", but it's just a natural consequence of you doing the odd thing of finding a way to make an infinite number of terms sum to a finite number.
I agree that if you try to make your moral weight reflect the probabilistic weight, you get weird results, but any set of infinite universes, including those with God, require you to do this. Consider a situation where you can choose to kill a puppy. If there are infinite universes, then (absent some force pushing against this), there are an infinite number that match current-Earth atom-for-atom, and of those, there are an infinite number where you do kill the puppy, and an infinite number where you don't kill the puppy. Therefore, it seems like your choice to spare the puppy doesn't cause there to be any more living puppies than if you killed it! So who cares if you make the moral choice or not?! In order to avoid this, you need some sort of measure to make finite sense of the infinitude of worlds. Once you map worlds onto measure, sparing the puppy increases the puppy's measure (probably by quite a lot, since you're most likely to be in the most probable worlds) and can be justified again. I don't know how to justify acting morally in an infinite world without doing something like this, unless you abandon consequentialism entirely and say you should do it for the good of your soul or something.
This is part of the general concern that I don't think you've thought about the many ways that infinite worlds go wrong even *with* God and granting everything about God that you want to grant. Again, I'm not sure how you say that the probability of me being (my) world's tallest person is 1/10 billion, rather than 1/2 or undefined. In order to quantify likelihood of being in any class, including the very normal classes that we all think about every day, you need to do something about the infinite paradoxes. I'm not sure why you don't respond to this in the original post, but I think it's decisive and that a good response to this point would convince me much more than any of the other somewhat tangential things about psychophysical harmony or whatever.
Re priors: I'm not just rehashing the basic problem of priors. There's a more specific problem analogous to the moral knowledge problem. It's tenuous if you believe A based on B but B isn't explained by B. So it's unclear on what basis we hold, absent trusting we have reliable faculties for ascertaining the priors, that simpler worlds consistent with our experience are likelier than complex ones.
I'm not assuming epiphenomenalism. I agree consciousness causes stuff. The point is that there could be a different state that was disharmonious but had the same causal effects. Thus, it's a coincidence that the states that cause certain physical states pair harmoniously with them.
I agree that the UDASSA result is how the math works on the view, but I just think that's an obvious reason to reject the view. You shouldn't bet on a proposition if 100% of the copies of you betting on the proposition lose money. That's a surefire recommendation for getting money pumped. I also just find it intuitively ridiculous that you should be arbitrarily confident in a prediction when there are infinity copies of you that will make the prediction wrongly and only finitely many that will make it correctly.
My concern wasn't about moral concern tracking probabilistic weight. It was that your view requires we either give up ex ante pareto or do things that benefit finite people and harm infinite people by equal amounts. Incidentally, if every mathematical structure exists, then the same things happen no matter what you do, so I think that leads to your actions not mattering.
I don't think my view leads to moral skepticism because we have free will and causal decision theory is right. You don't cause other people with free will to act.
//Again, I'm not sure how you say that the probability of me being (my) world's tallest person is 1/10 billion, rather than 1/2 or undefined.//
I think it would be undefined by default. The idea is God uses some chancy process to place you in the world, such that the chancy process has a low probability in making you your world's tallest person. It's analogous to the Hilbert's hotel case I discussed in the article. I did, in fact, make that point in my original post.
The basic idea: if God uses a chancy process to put you in the world, then your credences are dictated by the chancy process rather than the world. Analogy: suppose God rolls a 100 sided dice and makes you the tallest person only if you get 1. Well, though there will be infinite people who get 1 and 2 and 3 and so on, you should think at 99:1 odds that you're not the tallest person. My view is that God does something analogous regarding our placement in worlds, using a random process.
I agree this is weird and it's not super obvious that this doesn't leave all probabilities irreparably fucked. But I'm sufficiently confident that UDASSA doesn't work that I'm much more willing to gamble on "eh, it's not so clear how things turn out on SIA, but it's at least not obvious that things break down, and the alternatives are much worse."
I think UDASSA is particularly bad if you grant that it's possible for there to be qualitative duplicates. These duplicates, despite being exactly the same and having exactly the same location in exactly identical worlds, just arbitrarily get radically different shares of the overall probability space. Because each person there is equally easy to specify, simplicity considerations will be of no help.
Lastly, I think UDASSA violates reflection. Suppose that there are infinite copies of you, each in an otherwise identical room with different numbers on the door. Now, UDASSA implies that you should think it's likelier you're in the rooms with lower numbers on the door. But it gets much worse! Now suppose that after a half hour, you'll change around the doors so that bigger numbers have disproportionately more complicated things written on the doors (e.g. one has something written on the door, two has a more complicated thing written on the door, and so on). Well now your credence in being in room 1 goes up, so does your credence in being in room 2, while your credence in being in later rooms goes down. But this is nuts! At the end of this process, you were in room 1 if and only if you were in room 1 at the beginning. If A and B both entail each other, you should have equal credence in them. This also sets up an easy money pump, where you offer a ticket on being in some high room (room 39738249732894) that you buy back later for a predictably lower price.
I don't want to relitigate all of these, but I'm really confused about your tallest person claim.
First, God can't roll a die and make you the tallest person iff it's 1, because He has to do this for everyone in your 100 person world, and if the die coincidentally lands on 1 twice, there are two tallest people. I think you can rescue this by saying He rolls until He gets a 1, that's the tallest person, and then He fills the world with however many other non-tallest people. This would require some ugly math, because height still has to follow a bell curve. While I suppose God could find some elegant divine way of solving this problem, I think it's more likely that He just sets everyone's height according to a Gaussian - or, more realistically, lets it follow the usual genetic laws that determine height, which naturally result in a 1/100 chance of everyone being tallest.
The bigger problem is - suppose He creates an infinite number of 100-person worlds like this. Now we're in the exact same situation we were in before, where you can move all the tallest people to one place and all the non-tallest people to one place and have two equal-cardinality infinite sets. In fact, if we imagine God lets things follow the natural genetic process, then this is exactly equal to the non-God situation where everyone has their height naturally.
I think you are just thinking of things as the multiple-worlds-are-undefined thing when it suits your arguments, and switching to the identical roll-a-die process when *that* suits your arguments. God doesn't help resolve the paradoxes either way, only the convenient paradigm-switching does.
All of your objections to UDASSA just seem like being surprised by math to me. For your last paragraph - it's not a paradox, you're just changing the probabilities, then acting surprised that the probabilities changed! Suppose I show you infinite doors. One has a car behind it, such that there's a 1/2 chance it's in the first door from the left, 1/4 chance it's in the second door, and so on. Then while you're not looking I "change the probabilities" (ie move the car) to a new distribution - 1/2 chance it's in the first door from the *right*, 1/4 chance it's in the second door, etc. Now instead of believing the car is somewhere near the left, you should believe it's somewhere near the right. This isn't a paradox, it's just that I changed the probability of where the car would be. I can't money-pump you on where the car would be, except in the fact that I can trivially charge you $1 to pick the left door, then say "Haha, I switched the car to be on the right, do you want to pay $1 to switch your choice to the right?" and keep doing that until you wise up and realize I will keep switching.
Yeah I don't think that merely having infinite people be in rooms, each with a label of 1-100, allows you to say that, at 99:1 odds, you have the label 1-99. There has to be some genuinely chancy process that applies to you specifically. It would however work out if God uses some chancy process to dictate your location in the room (e.g. randomly assigning everyone a room number so that one person is the tallest, another person the second tallest, and so on).
//The bigger problem is - suppose He creates an infinite number of 100-person worlds like this. Now we're in the exact same situation we were in before, where you can move all the tallest people to one place and all the non-tallest people to one place and have two equal-cardinality infinite sets. In fact, if we imagine God lets things follow the natural genetic process, then this is exactly equal to the non-God situation where everyone has their height naturally.//
I don't think this matters. Arrangement is irrelevant. What matters is the process that you know was applied to you.
The idea is that if there are infinite people with every property, you should have undefined credence in your having that property unless you know that some specific process was done to you that had some specific probability of giving you that property. As, for instance, in the die case. Theism can secure that kind of process.
Regarding the last objection, I think you're missing the core weirdness. Obviously your credence in some proposition at a time won't always remain the same over time as the car case illustrates. If you're flying to California, your credence in "I'm in California" should be very high but should change.
The difference is that in this case the second proposition is true if and only if the first one is! If there are two propositions which are either both true or neither true, then you should have equal credence in them. You shouldn't say "I am 90% sure there's water in my room, but only 80% sure there's H20 in my room."
The thing that's relevant is that changing what happens on your door is absolutely incapable of changing your room number so after it happens your credence in being in different rooms shouldn't change.
Definitely still not getting where you're coming from.
We got into this discussion because you said that even if there were one billion real humans and one Boltzmann brain per universe, the human:Boltzmann ratio was undefined. Why isn't whatever combination of physics, evolution, etc causes you to be either a human or a Boltzmann brain a chancey process, similar to the combination of physics, evolution etc that makes you a tall person rather than a short person? Isn't the former equivalent to rolling a one billion sided dice and making someone a BB iff they land on one? Dice work by a deterministic physical process just like physics/evolution, so how are they different? Or are you saying God uses some actual divine RNG doing things in a nonphysical way beyond human possibility in order to select the tallest person, and only results of the divine RNG are free from paradoxes of infinity? Why would this be true?
I think where we're failing to connect with the rooms thing is that you seem to think you're just changing what's on the door, but if I'm understanding you correctly, the analogy to UDASSA is changing which universes are most simple/complex, which is a genuine change that itself changes probability. Like, you say "UDASSA implies that you should think it's likelier you're in the rooms with lower numbers on the door." This is absolutely not true literally in a literal hotel (do we disagree on this?), so I'm taking it to imply that you mean it's figuratively true if the . . . numbers on the doors represent the complexity of the universe you're in? But then that's like saying "You're acting like changing the numbers on the door changes the chance of there being a car!" when the numbers represent the probability of there being a car.
By the way, I'd be totally happy to have a phone call if you want to discuss this more. If so, send me a dm. I tend to find that to often be helpful when lengthy written back and forths lead to both sides thinking the other one is missing something.
//Why isn't whatever combination of physics, evolution, etc causes you to be either a human or a Boltzmann brain a chancey process?//
Yeah this might be right. Will have to give it more thought. But the core problem is that on the Tegmark view there are infinite copies of you in different universes with your exact present experience. So figuring out which universe you're in is not subject to a similar chancy process.
(Having thought about it a bit more, I think it probably is right).
(Analogously, if you roll dice in Hilbert's hotel but there's a neighboring infinite hotel with hallucinating people of every variety, then you still should have an undefined credence in your being in the first hotel).
In a hotel, assuming you want a normalizable probability distribution over an infinite number of rooms, you will have to hold that the earlier rooms get a higher share of the probability space. But it just seems totally irrational to think you're likelier to be in room 1 than room 72.
The objection I gave to UDASSA was as follows. Imagine that you have infinite rooms or universes--doesn't matter which, just matters that you have something. Right now, universe 1 is very simple. However, in an hour, universe 1 will undergo changes which make it much more complicated (God will, by divine fiat, add a bunch of complicated stuff to universe 1).
UDASSA implies that your credence in being in universe 1 now should be higher than they will be in an hour. But you're in universe 1 now if and only if you'll be in universe 1 in an hour. So as long as you should have equal credence in mutually coentailing propositions, UDASSA has to be wrong.
Here's an analogously bad situation. Imagine that right now you think your friend's name is Fred at 60% probability. But you expect in an hour to think that his name is Fred at 80% probability. There's no chance his name will change. This seems quite suspect--your credence in some proposition shouldn't predictably go up in this way.
I still don't think you've separated out the world's-tallest-person scenario from the Boltzmann brain scenario.
Both of them involve infinite worlds. Each world has (let's say) one billion people, of whom 999,999,998 are unmarked, one is the tallest person in the world, and one is a Boltzmann brain. Both tallestness and Boltzmannness are caused by a deterministic process. In both cases your consciousness has to be placed at random into some person in some universe. What is the difference between membership in these two classes, such that tallestness plays nice with infinity but Boltzmannness doesn't?
I'm also still confused about your room scenario.
UDASSA doesn't say that universes get probability mass because of how complicated they are *now*. They get it because how complicated they are *to specify*. So even though our universe has many things, the "recipe" may take only a blackboard worth of equations to specify (it would like something like "take one point of infinite density, add gravity that works like [equation], magnetism that works like [equation], and the mass of the quark is X EV". In order to change how complex our universe is to specify, God would have to change the rules of math, which is contradictory.
I don't think God can simply go into the "material" universe, create some stuff, and then ask if the math changes too, because there is no material; it's *all* math. It's like asking "If I went onto the graph of y = x^2 on the Cartesian plane, took a giant eraser and marker, and messed with the parabola until it had a bunch of extra dents and curves in it, would that make the function y = x^2 itself have more terms?" It's not coherent enough to answer.
I guess you could phrase this as MUH not permitting the possibility of entities outside universes affecting those universes. With a marker and eraser, you can make a specific written-down graph of y=x^2 in a math textbook look different. But you can't affect the true Platonic graph of y=x^2 itself. Only the definition of the process of squaring a number can do that. But our universe isn't a specific written-down instance of the cellular automaton it represents, it's the automaton itself. So you can't affect it with normal tools (even if you're God). You just have to watch how the equation plays itself out.
(humans can affect it because we're just arbitrarily lines drawn around parts of the math, in the same way that the number five can affect the graph of y = x^2 by being 25 when it is squared rather than 26. This is basically the argument in https://www.lesswrong.com/posts/NEeW7eSXThPz7o4Ne/thou-art-physics , but with a friendly amendment changing "physics" to "math")
If you wanted to make your scenario meaningful in MUH, you could imagine God being a part of our universe, which would raise the universe's complexity a lot insofar as God is hard to describe in math. This is basically fine - there's some super low probability universe that's complex enough that it has God in it - but we should expect very low likelihood we're in that one.
Otherwise, when I try to imagine your scenario, I imagine God trying to add a very complex device on Mars or something, and then finding that oops, He's now in the different MUH universe where there was already a complex device on Mars (which is lower-probability than ours). If He objects "No! I meant to keep this in the higher-probability universe over there!" then it's fruitless - like trying to add a side to a triangle in Platonia and getting upset when you find yourself at a square rather than a 4-sided triangle.
>UDASSA implies that your credence in being in universe 1 now should be higher than they will be in an hour. But you're in universe 1 now if and only if you'll be in universe 1 in an hour. So as long as you should have equal credence in mutually coentailing propositions, UDASSA has to be wrong.
Wait, why does it imply this? The simplicity of a universe isn't a time-indexed thing in the way it prima facie sounds like you're suggesting here, though I admittedly might be misunderstanding. Rather, it will reflect the minimum description length that specifies its behavior as a whole (i.e., across all time), and that doesn't depend on what time it is right now, or how simply it happens to be behaving right now. Is there something very specific to UDASSA that's incompatible with this kind of thinking?
(Realistically, if the universe suddenly does start to act more complex-ily in a way you didn't anticipate, that should of course diachronically change your assessment of its simplicity. But it sounds like you're talking about a scenario where you do anticipate the change.)
"The idea is that if there are infinite people with every property, you should have undefined credence in your having that property unless you know that some specific process was done to you that had some specific probability of giving you that property."
How do we know, even in theism, which properties had such a process? Also, we do we deal with the fact that properties aren't independent: the tallest person is also going to have a good chance of having the longest legs, the longest arms, etc....
I'm reminded of Michael Huemer's discussion of Bertrand-type paradoxes: suppose you know a machine makes cubes whose side lengths are between 1 and 2 cm. What should you pick as the max entropy estimate of the expected volume?
On the one hand, the estimate for side length should be 1.5, halfway between the two extrema, which yields a volume of 1.5^3=3.375. On the other hand, cubes between 1 and 2 cm side lengths are the same as cubes between 1 and 8 cm^3 volume, and here the halfway point is 4.5
So which is right? Huemer says the second, because cubes, being 3d objects made out of 3d matter, have volume as a more "fundamental" property.
To avoid/resolve paradoxes like this, do we need to assume that God only applies this process to "fundamental properties"? Do we have any idea what those properties are?
I think your definition of a chancy process is going to be subject to all the problems you claim of the simplicity prior. You note that rolling a die in an infinite creation case is justified as yielding uniform probabilities of 1/6 not because of the proportion of people in each outcome set, but because "After all, as you repeat the process, it will be 1-5 most of the time."
This is the exact same limiting process you've disagreed with other people making! After all, the final "proportion" of 1s rolled won't be 1/5, it only holds if you imagine the die being rolled sequentially and take the limit of the proportion up to time t as t-> infinity.
But it's not clear why we would think of these die rolls as happening sequentially; and even if so, what if God using his all-knowingness, decides to roll the die so that two people who will get a 1 go first, then each of the other outcomes once, then repeat? You'll find that the limiting fraction is 2/7 for 1, and 1/7 for the other 5: rearranging the order of the rolls will give you other values for the probabilities.
The reason 1/6 seems natural is because you've picked a limit that feels natural, but that naturalness is grounded in intuitions from die rolls in non-infinite cases: whether I roll a finite number of dice sequentially or all at once, I get 1/6, so it seems like a natural answer to carry over to the infinite case.
But, though I'm not without some sympathy to this view, it really isn't clear that 1/6 is the correct probability, or that such a notion is defined. On the frequentist view, it seems to depend on what limiting process feels natural to take the limit of frequencies, and without knowing the exact process God uses, it's not clear there should even be a canonical limit, much less what its value is. And on a subjectivist view of probability... Can God even have subjective probabilities that aren't 0 and 1?
It feels to me like any assertion that one particular choice of limit, and hence probability, is canonical is going to feel arbitrary without a _lot_ of justification. Ironically, it strikes me as having the same issue you (correctly, IMO) identify with the simplicity prior: it won't _feel_ like a probability, it won't unambiguously do the things we want a probability to do for us; it'll just satisfy some abstract mathematical criteria.
Finally, it strikes me that we have no way of knowing what process God uses, and so even if this does ground probability in some sense, it's not clear that this is at all usable.
I think we know God cares about our cognitive faculties, so we have reason to think that we wouldn't be in a skeptical scenario! I agree that maybe some of the probabilities will turn out a bit unclear though.
The idea behind thinking 1-5 is likelier is:
1) If you take the limit as the number of die rolls approach infinity it will approach 5/6 getting 1-5.
2) It's reasonable to assign a difference principle over possible rolls that you might get--none of them are special--and so you know that in your case, there's a 5-6 chance you got 1-5.
The basic point is that if you're rolling dice, you just ignore what's happening to other people. If you didn't know about the other people, you'd assign 5/6 chance to getting 1-5.
You'd assign 5/6 if you understood the process ahead of time, and could draw the analogy to normal dice rolls. Whether that's justified is what's at stake!
For one thing, this die is being rolled by God, who presumably knows the answer to what will be rolled ahead of time.
What's more, while I certainly _can_ ignore what's happening to others, I don't see that I'm obliged to. Suppose I refuse to reason as you suggest, and endorse a different limit: God explains to me that a die will be rolled to place me in a universe, and I go ahead and calculate the probability that my die was rolled a 1 using my preferred limit, and say that my probabilities are (1/15, 2/15,3/15,4/15,5/15,6/15), not uniformly 1/6.
Am I _wrong_? What can you point to, to convince me that I've got the wrong probabilities?
My problem is that the term "genuinely chancy" seems to imply something objective here; it's presumably not genuinely chancy from God's point of view since God knows the outcomes of all the dice rolls, so it must be "genuinely chancy" from _our_ point of view, but if so, it sounds like the probabilities can't just reflect our lack of knowledge or our application of some heuristic... What is "genuine" about the chanciness? It seems like you want to ground these probabilities in some objective features of the world, but the only options I can think of are frequencies, which won't avail us in the infinite case, or some canonical model for the process... So I think you have to make an argument not just that modeling the dice roll in the way you do is "reasonable", but that there is something objective about its chanciness that is model independent, or that picks out a unique model in a strongly justified way
Re: psychophysical harmony, if the physical aversive response of pain doesn't actually require the subjective qualia of pain, then isn't the fact that pain still exists actually a point against an omnibenevolent God? We could have had an aversive response that served our survival without the innate, immediate badness of emotional suffering, but God in his arrangement of our psychophysical harmony allowed that inherently negative mental state to exist. If the qualia of suffering is inherently bad, which I believe it is, then why would an all-good God saddle us with it when it in fact isn't necessary?
Bentham's God dislikes deception and also really likes connection-building. If people didn't have qualia of suffering, helping them wouldn't be worth anything, and there'd be no reason for us to come down here from our heavenly pre-existence. "Why would God give us qualia of suffering?" seems like the same question as "Why would God make us suffer at all, given that we can?" and I think is answerable by the same theodicies. E.g., https://benthams.substack.com/p/a-new-preexistence-theodicy
Helping people would still be good if it increases their happiness! Why is pain uniquely key to connection-building? Oh right. It’s because all of these theodicies are ad hoc nonsense with no independently motivating rationale.
Sure, but they're ad-hoc nonsense that explain the *possibility* of suffering just as well as they explain the *existence* of it. My point is that (if we assume God doesn't want to deceive us) there's nothing special about this objection that isn't just the problem of evil rehashed.
"Hey, look at all this evil, isn't that evidence against theism?"
"Yeah, but here's a somewhat plausible explanation for God allowing it."
"Haha dumbass, gotcha! Why wouldn't He just *make evil impossible*?"
Well, you ignored the very first line of my comment, which pointed out that there is in fact a unique cost to maintaining pain which isn’t clearly solved by connection-building theodicies. You should answer it before being so condescending.
I don’t mean to condescend, I’m just bad at being funny…
Look, I don't think you're actually right that just increasing happiness builds the same kind of connection. Clearly there's meaning in buying a house for a poor family that isn't there when you're buying a summer home for some very wealthy guy.
I think you'd also hit a pretty strange reductio. Like, if God could just remove the qualia of pain and connection-building still happens, why can't God also remove the qualia of slight-happinesses? Why not remove all of the happy qualia that are any lesser than heavenliness?
At some point along the way, connection-building must stop happening. If it's not the point of no-suffering, where is it? Is there any other spot that isn't super arbitrary?
> Clearly there's meaning in buying a house for a poor family that isn't there when you're buying a summer home for some very wealthy guy.
Obviously a case of diminishing marginal utility from getting extra material possessions. This has nothing to do with happiness versus pain.
> I think you'd also hit a pretty strange reductio. Like, if God could just remove the qualia of pain and connection-building still happens, why can't God also remove the qualia of slight-happinesses?
Sure. Why can't he?
> At some point along the way, connection-building must stop happening. If it's not the point of no-suffering, where is it? Is there any other spot that isn't super arbitrary?
Well, A. There is presumably some fact about when it does or does not stop happening. Alternatively, if its a question with no answer like "when do grains of sand become a pile?", that's a signal that "connection-building" is fake.
B. This is a strong argument *against* God. The world *is* insanely arbitrary. The idea that its somehow the maximally perfect state of affairs is absurd for that very reason! This is distinct from evil for--in part--pleasure/pain as argued above, and every other distinction about the very arbitrary *qualities* (as opposed to magnitude) of evil.
If God dislikes deception, why is the night sky full of seemingly randomly-distributed stars, rather than e.g. a quickstart guide to telescope construction, with diagrams and multilingual text (Chinese, ancient Egyptian, and proto-indo-european), branching out to a wider variety of subjects by fractally smaller print within each astral brushstroke?
And I should add: notice how this argument breaks down on the exact same lines if you replace "people's emotional variance" with "people's emotional experience." The problem of the possibility of suffering seems equivalent to the problem of the existence of suffering.
> arranges absolute majority of the evidence such that the universe appeared to be naturalistic
You're question-begging—to a theist, the absolute majority of the evidence points to theism.
> What if people's emotional variance was moved from (suffering - neutral - pleasure) to (neutral - pleasure - absolute bliss)
Would it feel the same? Maybe see that bit below about diminishing marginal utility. You would need the highs to be much much higher than the lows, and it seems like there's a cap you'll run into ("heavenliness"). And connection-building isn't perfectly isomorphic to utility anyway—maybe God's a negative utilitarian about this. Alleviating someone's suffering creates a substantially different relationship from enhancing their pleasure—that seems very intuitive to me.
> to a theist, the absolute majority of the evidence points to theism.
Then why can't modern Abrahamic priests produce e.g. a plague of frogs on demand? Per Exodus 8:7 the Pharaoh's court wizards seemingly figured out how to duplicate that trick same-day, though not the off switch.
> I don’t quite get the complaint. I think God did set up some elegant system of psychophysical laws. I also think he had an elegant system for making it so that we’re each probably not Boltzmann brains. I don’t think God intervened to do a special miracle to solve each of these problems.
Let me try to explain. The complaint is, fundamentally about a faulty pattern in your reasoning proccess.
1. You really want something to be true - not being a Boltzman brain, being rightfully confident in inductive reasoning, moral/logical/modal facts being "stance independent", scepticism being wrong, etc.
2. You have some evidence in favor of these, but not enough to be confident in your prefered views. Actually, you have some very serious counterevidence: it would, indeed, be a huge coincidence if there were objective moral truths and we just happened to learn them via evolution, that wasn't optimizing for it at all.
3. Instead of lowering your confidence in your prefered conclusions, as one should in such circumstances, you assume it as a bottom line anyway, and then start motivatedly searching for some way to justify it.
4. You arrive to the motivated conclusion that theism solves all these problems. Therefore, not only you accept your prefered views, you also accept theism, even though it itself is questionable. P(A)P(B) < P(A). You would be less wrong by simply assuming your prefered stance A without enough justification. And yet you also accept B.
5. And then you try to find some justifications to theism, in the same motivated fashion. You start assuming theodicies that would "explain away" counterevidence to theism, even though that it actually makes your overal theory less probable P(A)P(B)P(C) < P(A)P(B) < P(A)
And so on and so forth. You keep building more and more epicycles, adding more and more details, to hide the error you did in step 3. You start inventing new problems that theism alledgedly solves. You do a tremendous effor to build an argument how naturalism fails to account for X, while simply assuming that theism surely does, without much investigation. At this point not only you are using "God" as a curriocity stopper, which prevents you from learning new things and deconfusing yourself, you are using it as a "knowledge eraser" - which makes even things we already have answers for, appear more confusing. Which is the opposite of what philosophy, ideally, is supposed to be doing.
In reality, there are people born without the ability to feel pain, and they often are prone to serious injury. We can hypothetically describe made up systems that would replace pain, but nature has not provided any that we can see.
Okay, I know this is off-topic. But this argument:
> To give an analogy, imagine that there are some goblins who worship a being called the absolute. The goblins have an intuition that worlds with the absolute have a higher prior than worlds without it. But they know that they’d have that exact intuition whether or not there was the absolute, and that intuition isn’t explained by the likelihood of the absolute. It seems they lose their justification for believing in the absolute.
seems to apply very well to religion!
There are several popular religions with fundamentally conflicting views. Therefore, if one is religious, no matter which religion one subscribes to, one should agree that most people in the world are incorrect about their religious beliefs. Yet empirically speaking, the incorrect people and the correct people acquire their beliefs from basically the same sources: faith, social proof, and philosophical arguments, combining to produce an intuition of correctness. Therefore, it's likely that one would have the same intuition about one's own religious belief whether or not it was correct, so "that intuition isn't explained by the likelihood" of the religion being right. Does one lose one's justification for believing in their religion?
Admittedly, this argument applies far more to traditional organized religion than your own position. On the other hand, your conception of God seems to be at least strongly influenced by organized religion. (But then one could say the same about philosophy as a whole, and even atheism?)
Most people are incorrect about many of their beliefs, yet that alone is not enough to lose justification in believing something. Your argument as written applies just as well to atheism: for belief in atheism comes from the same sources as belief in God, namely your own experiences, philosophical arguments, etc. If you took a particular believer and asked him why he believed he would give you specific reasons: if you convinced that believer (like the goblins) that they would still believe even if God didn't exist, then that would be very good reason for them to doubt their belief in God.
It's also worth nothing that a religious believer does not (typically) believe that their beliefs are right and every other religion is completely wrong. Though they may disagree on specific doctrines there is a lot to agree with: the existence of the supernatural, the existence of God or gods, the existence of the soul, judgement after death, etc. Lewis wrote about this in "Mere Christianity"
"If you are a Christian you do not have to believe that all the other religions are simply wrong all through. If you are an atheist you do have to believe that the main point in all the religions of the whole world is simply one huge mistake. If you are a Christian, you are free to think that all these religions, even the queerest ones, contain at least some hint of the truth. When I was an atheist I had to try to persuade myself that most of the human race have always been wrong about the question that mattered to them most; when I became a Christian I was able to take a more liberal view. But, of course, being a Christian does mean thinking that where Christianity differs from other religions, Christianity is right and they are wrong. As in arithmetic — there is only one right answer to a sum, and all other answers are wrong: but some of the wrong answers are much nearer being right than others."
> When I was an atheist I had to try to persuade myself that most of the human race have always been wrong about the question that mattered to them most;
I think he's projecting quite a bit there. For most of the human race, through most of history, question that mattered to them the most was probably
"are the kids gonna be okay" or
"will this plan bring glory, or shame, on me and my buddies" or
"where's the best beer"
or something like that, not cosmological stuff. Atheists are certainly permitted to believe that religious practices can contain useful hints toward those sorts of questions, and that people have guessed right before, without any inconsistency.
> Atheists are certainly permitted to believe that religious practices can contain useful hints toward those sorts of questions, and that people have guessed right before, without any inconsistency.
Have guessed right about what: the existence of the supernatural? Or what the best beer is?
The second one, with concrete results - though a lot of old-school animistic beliefs could also be thought of as hacking human social instincts into a useful approximation of an ecological https://en.wikipedia.org/wiki/Agent-based_model rather than really being "supernatural claims" in a strict sense.
>The basic idea: on Tegmark’s view, there will be infinite people with every conceivable property. There’s no coherent way to talk about proportions of people with the different properties—after all, you could arrange worlds so that each world is filled with any percent of people with any property. You could make every world filled almost entirely with Schizophrenics named Phillis who evolved from bats but have cognitive capacities similar to humans. Thus, all probabilities come out undefined.
But, for one thing, you haven't showed that anyone is committed to setting credences solely based on proportions and nothing else when the proportions are undefined or otherwise mathematically meaningless. This thesis seems incredibly implausible/unmotivated to me and I'm not sure anyone in the literature has ever advocated for it, for whatever that's worth (admittedly not much - there's always room for original philosophical positions). It also seems completely costless to reject; I can't think of any terrible consequences to not signing up for undefined credences of this sort when I have other epistemic principles at hand to adjudicate things, principles which in fact seem much more independently attractive to begin with than going with proportions.
Second, you yourself don't actually hold to this commitment (i.e., "undefined proportions -> undefined credences") on any other issue! When deciding based on Ptolemaic or relativistic models to explain the motion of the planets, you go with the simpler non-Ptolemaic explanations, even if there's an infinite number of people living in both Ptolemaic and non-Ptolemaic universes according to your view. And, as I remarked in a comment the other day, you can't account for this by talking about God ensuring the reliability of abductive inference, because the question being asked here is how to abduce, not how reliable that process is once we've already figured out how to do it.
Finally, once you inevitably agree that abduction works out on your view in the cosmological case via the legitimacy of a simplicity-based prior plus ignoring undefined proportions, you have to explain why Tegmarkians can't also use simplicity-based priors and ignore undefined proportions.
Sorry, one other thing, although my comment was probably already overly long:
> However, literally 100% of people do not find themselves in those worlds. This is an odd result.
This is probably not correct on Tegmark's view; I think he's going to want to say that for any universe, there are unsetly many people in structurally identical copies of that universe. (This is because mathematical objects/structures can be instantiated in many structurally isomorphic ways - for example, we can define the ordered pair (2, 6) as the set {{2}, {2,6}} as is commonly done in set theory, but we could also have defined it as {{6}, {2,6}} or indeed a ridiculously huge number of other ways.) So, if this is a correct description of Tegmark, there's no actual meaningful proportion of people in a world like ours.
It may be the case that there's a finite collection of *classes* of structurally identical universes that takes up 99.99999+% of the probability mass on Scott's view, but why pay attention to classes and not people? It seems like you want to pump intuitions based on things like "most people in the multiverse are badly wrong if the view I'm arguing against is correct."
Let's go back from the beginning, because I think people are not quite grasping the profundity of this Tegmarkian view. The basic postulates of this cosmological view are:
- self-existing mathematical structures are *already* encoding every possible computation, including the one that simulates this very universe's physical laws, as well as every other possible computable universe. (the "self-existing" part is an axiom of the theory; this universe being computable is a common hypothesis of physics; the rest is just a basic theorem of computation)
- this is enough for the biological organisms in these 'simulated' universes to be conscious of themselves and their surrounding universe, as long as they have the (simulated) physical feedback loops necessary to sustain individual sentience.
- this 'simulation' operates in the manner of an Universal Dovetailer. The Universal Dovetailer ("UD" from now on) is a computational structure (a "program" for simplicity) that enumerates and executes every possible program, step by step, in an interleaved way. See https://alwaysasking.com/why-does-anything-exist/#The_Universal_Dovetailer for details.
Now, this is not a claim that there is a big computer at the bottom of everything (in which case you'd rightly wonder who wrote the program, and why it's coded in perl, etc.). Rather, the claim is that the UD is a self-instantiating logical fixed point. Since the UD simulates every possible computation, it also simulates itself, and every possible variation of itself. But you can easily prove that, within the UD's sense of time (an ongoing computation creates a structure of time within itself), longer programs will be simulated exponentially slowly, simply because there are exponentially more of them, and they share a the UD's finite computational capabilities. You simply cannot write a Universal Dovetailer in any programming language or mathematical representation that fails to simulate longer programs exponentially slower than shorter ones. (Also, the UD's finiteness need only be local; it can possibly be globally infinite and do a spatially infinite amount of calculations at each step, while still doing those interleaved steps one by one.)
So the simplest possible UD becomes a logical fixed point, because it's the one that gets simulated the most in *every* UD, including non-simple ones with useless epicycles. This is enough to ground the structural unicity of the simplest UD. If any logical structure has the capacity to be self-instantiating and infinitely generative, the UD is it, and due to any UD's bias towards the simplest instructions, the simplest UD wins out. From here on "UD" means this simplest UD.
This gives you a good basis to ground (accursed) anthropics. The outermost, most objective reference point you can find are not conscious observers within universes, or classes of them, or instants of observer-consciousness within worlds, but instants where the UD is simulating a conscious being. The steps where the UD is simulating insert matter like a stone obviously don't count for anthropics, because the stone (presumably) doesn't have the physical feedback loops that would allow it to sustain an individual subjectivity, let alone wonder what kind of world it finds itself in. And since the UD is (here, simplisticaly) modeled as a single-threaded computation, you can do stats on those instants, and ground anthropics on them.
So if you find yourself as a conscious observer within this Tegmarkian multiverse, the question becomes, what kind of world can you expect to observe?
The answer is simple: the mathematically simplest one that is still complex enough to give rise to a rich physics that can sustain evolutionary feedback loops. That's where the vast majority of computation steps instantiating consciousness are found in an UD.
And that happens to be a great match for the world we find ourselves in! Including discoverable physics based on relatively simple mathematical laws, with apparently arbitrary constants fine-tuned to make complex life possible. Everything that your theistic hypothesis explains, this explains too.
So let's compare this kind of theory to the philosophical theism that BB defends. We can define the theistic hypothesis as postulating an entity ("God") which is 1) self-sufficient, 2) generative, and 3) good. As BB has explained in detail, this predicts many of the features we find in the world, except for the bit where we'd (naively?) expect the world to be *more* good than it appears to be.
Now, self-sufficiency and generativeness are needed features of *any* possible cosmological foundation. You can't have a foundational theory without them. In this Tegmarkian view, the UD is also self-sufficient and generative. But in the theistic view, God is self-sufficient and generative *by fiat*. Here we actually have some structural justification for the UD to be self-sufficient (it's a logical fixed point), and generative (it's literally specified as a maximally generative program).
Note that God's goodness is not only problematic (see the epicycles of theodicy), but also an extremely onerous hypothesis. The sense of 'good' is one of the most complex things in human culture. Attempts to formalize it simply are also full of problems, pace Bentham and his defenders, human and canine alike. Like Scott hinted in one of his replies, the sense of "good" is an evolutionary product, the accumulated result of millions of years of biological evolution, plus ~100k years of much faster cultural evolution, of trying out and integrating strategies of collaboration/competition. To put *goodness*, of all things, at the cosmic level, as a fundamental self-existing element of reality is the most onerous hypothesis I can think of. Literally anything else would be simpler, including a tiger jumping at you right now out of nowhere.
And that's why the Tegmarkian view wins. It can do the whole job, including predicting a fine-tuned, comprehensible universe within which complex beings develop a sense of the 'good'. And it only requires some maximally general postulates - pretty much the bare minimum to have anything at all.
Note that I haven't included any reference to Solomonoff induction here. It's not really relevant — in actual physical worlds like ours, the way we do inference is by semi-formally noticing patterns and then checking whether they generalize. It may not be perfect, but it has done wonders so far. Our best historical inferrers, people like Thales and Newton and Hawking, never once thought "let me find the shortest Turing Machine tape that matches my sensory experience". Instead they took the common shortcut of abstracting away the subject, considering the object of observation in general rather than the specific angle they were seeing it from, then abstracting away time so they could have multiple observations, and then semi-formally looking for patterns in there. It starts with simple things like "x is proportial to y", all the way up to General Relativity and the information content of black holes.
Solomonoff may well be a nice way to formalize an ideal version of induction in an induction-friendly (Ockham-friendly) world, but it's not how we actually do things, and it's not a part of our Tegmarkian theory. So the whole of Joe Carlsmith's post about to how Solomonoff induction has alleged difficulties making sense of a mathematical universe (https://joecarlsmith.com/2021/11/28/anthropics-and-the-universal-distribution), and the whole discussion of "world and claw", don't really apply to the Tegmarkian view. Because that's just NOT how we discover physics in practice within our experiential world.
> (If you’re thinking “no if physicalism is right this isn’t really imaginable I’d suggest you read the paper on the subject, as this worry is addressed at length).
I admit to only skimming this paper, but it seems like their response to physicalism is that psychophysical harmony is more likely under theism then physicalism. This is wrong. At least in *my* physicalist account of consciousness, consciousness is a *functional element of thought* and subject to evolutionary optimisation like every other part of us. There may be some ultra-contrived way to make psychophysical harmony fail anyway, but being contrived they're much less likely to arrise then more normal approaches. This whole problem is only a problem for you because you don't think that.
Do Goblins have moral knowledge? For instance, if (hypothetically) they were to attack a grove of druids, would they be doing it whilst knowing that they are committing evil? Would they simply be ignorant that causing suffering is evil? Would they argue for the morality of their actions through Divine Command from the Absolute? If so, would they be justified? Would things be different if they were Trinitarian Absolutists?
Goblinoid metaethics si such a fascinating subject of inquiry.
"So long as you keep A and C the same, B is totally irrelevant."
In the isolated incident, sure. But what about the next time I want the experience of eating a pickle? I could A) stick my hand in a blowtorch, or B) eat a pickle.
Evolution would clearly favor B and discourage A. A species of pickle-taste-maximizers that get their pickle kicks from eating pickles will outproduce a species that gets their pickle kicks from self-immolation. I'm not sure how we can consider B being irrelevant on an evolutionary scale, or even the scale of a single lifetime.
I'm saying that wanting pickle-taste would cause me to do something. If that something is light myself on fire, I'm not going to live long enough to reproduce. On naturalism, we would expect that species where "want pickle taste" is fulfilled by pickle eating, or at least some positive action, would be better at survival.
We could imagine a creature that responds to all stimulus with pickle taste. Mutations in that species that progressively loses its pickle taste response to harmful stimuli (such as blow torches) are going to out compete omni-pickle-tasters. This wouldn't cause this creature to only eat pickles, pickle taste would also come with any positive experience (reproduction, etc.). But I guess in this analogy pickle-taste is a dopamine response?
All this is complicated by the fact that I don't like pickles. Pickle taste would be a functional substitute for pain, for me.
Whether after tasting a pickle you'd want pickle taste will depend on the psychophysical laws. It would be unlikely that it does or that wanting a pickle taste is associated with getting a pickle.
Personally, my response is functionalism. And I'm not convinced by the linked paper's rebuttal, which seems to be summed up as: "if you think dualism might be true, then atheism at least potentially has a problem, whereas theism definitely doesn't have a problem; this makes theism more likely".
First of all, this is a excellent reason to discard atheistic dualism, but once you've done so, the argument no longer points towards theism.
Besides, I think that functionalism is a far more elegant solution to the problem than theism. Why do this physical thing ('the brain's perception') and this abstract thing ('conscious perception') happen to be exactly identical in all measurable ways? "God made it that way because he's good" works, but "because they're actually the same thing" is more satisfying. Plus, the same exact reasoning applies to morality, where the physical thing is 'what we evolved to feel morally good about', and the abstract thing is 'what is actually morally good'.
Finally, it also feels like a just-so argument that God fine-tunes some things extremely precisely and obviously (psychophysical harmony, moral knowledge, universal constants) but fine-tunes others indirectly and unobservably (tendency of the universe towards moral good).
I find the paper's objection to a Functionalism explanation unconvincing.
"it is crucial to the original argument that information about psycho-functional correlations is not treated as part of our background knowledge."
But why wouldn't we? Evolution "acts" a posteriori, so functional correlations would be considered. Even if "evolutionary forces cannot affect the psychophysical laws", it can select for which laws are used.
To use their boulder island analogy, I see psychophysical harmony not as encountering a Stonehenge like structure, but instead seeing all the boulders at the bottom of a ditch or other areas of low elevation. Yes, someone could have placed them there, but I see no reason to give that a higher probability than them just rolling there naturalistically.
Regarding your argument that atheism makes induction unjustified, I think there’s an issue with the argument which appears pretty similar to a problem I have with many other of your arguments and which is sort of similar to Scott’s complaint about God solving any loose ends. You appear to be arguing that induction is not justified because after all the evolutionary explanation of induction itself would be equally compatible with a universe where induction stopped working tomorrow. It is true that you don’t have this problem if God exist because he would give you a different probability if you were in that universe. But the actual evidence that either theory has to explain is your actual probability because beyond your probability distribution, you don’t actually have any evidence that the probability distribution is justified in this way. Basically, you seem to be arguing that you believe in induction. Therefore, induction must be justified in the sense. I have mentioned above, and God would provide such a justification, whereas evolution does not without the existence of God. But the only reason you think induction should be justified in the sense is because you believe in induction so you don’t actually have any independent evidence that induction is justified in this sense and therefore all that needs to be explained is the evidence that you believe in induction, which is equally likely given evolution.. also given that the only reason you have to believe that induction is justified is your belief in induction and it’s your belief in the existence of a justification which causes you to believe in God, he doesn’t actually provide any independent justifying evidence of his own that is not completely circular on your belief that induction is justified. I am having trouble putting this in words, but my basic intuition is that if you’re going to indulgent this type of reasoning where your belief can itself be used as evidence to come up with a circular justification, it seems equally good to simply take your belief has given and get rid of The unnecessary justification, which has no actual justification for believing in it beyond your belief.
I'm not sure your point about UDASSA is right if you have uncountably many worlds. If your worlds are indexed by the real numbers in [0, 1] and have a uniform (or any other continuous) probability distribution on them then the probability you're in any particular finite subset of those worlds is 0.
Another thought: does this idea of a "chancy process" dictating probabilities in an infinite world undermine the case for thirding in Sleeping Beauty? If infinitely many versions is SB will experience all possible awakenings, even in a single SB, are we obliged to take our probability from the underlying chancy process, i.e. the coin toss?
Re: knowledge of priors - are you just rehashing the problem of skepticism? If you're infinitely skeptical (eg of logic itself), then not even God can solve your problem (any proof of the existence of God depends on logic). If you're less than infinitely skeptical, I think understanding that a few hyperpriors were instilled by evolution which had an incentive to get them right, plus https://www.lesswrong.com/posts/46qnWRSR7L2eyNbMA/the-lens-that-sees-its-flaws , gets you the rest of the way. I agree you can always argue that evolution somehow found a perfectly consistent set of priors which works for all pragmatic purposes but isn't true, but at some point this because its own "psychophysical harmony" problem (why are our false priors so perfectly consistent with true reality?) to the point where the probability becomes vanishingly low.
I'm not sure why you say conscious states are epiphenomenal even if epiphenomenalism is false - you just seem to be transparently pushing epiphenomenalism. But epiphenomenalism doesn't make sense, because we *know* our conscious states have actions in the world - most obviously, the action of philosophers writing "Hey, we have conscious states, what's up with that?" The same fact about pain that causes you to write the words "pain is unpleasant" in your essay on psychophysical harmony also causes you to behave as if pain is unpleasant - they're both downstream of the actual unpleasantness of pain.
Yes, I'm asserting the thing that you call UDASSA, and not the thing that obviously doesn't work. Your "objections" to it don't make sense. You say "If the UDASSA view is to maintain a normalizable probability distribution—one like the probability distribution .5, .25, .125, etc that sums to 1—it must be that there exist some finite number of worlds such that you should think with probability 99.999999999% that you’re in one of those worlds. However, literally 100% of people do not find themselves in those worlds. This is an odd result."
This isn't an "odd result" - it's just how math works! If you divide the interval from 0 to 1 into regions, such that one region includes the first half of the space, the next the next 1/4, the next the next 1/8, etc, then you can find a finite number of regions such that 99.9999999% of the interval is in those regions, even though literally 100% of regions aren't in the interval. This may be "odd", but it's just a natural consequence of you doing the odd thing of finding a way to make an infinite number of terms sum to a finite number.
I agree that if you try to make your moral weight reflect the probabilistic weight, you get weird results, but any set of infinite universes, including those with God, require you to do this. Consider a situation where you can choose to kill a puppy. If there are infinite universes, then (absent some force pushing against this), there are an infinite number that match current-Earth atom-for-atom, and of those, there are an infinite number where you do kill the puppy, and an infinite number where you don't kill the puppy. Therefore, it seems like your choice to spare the puppy doesn't cause there to be any more living puppies than if you killed it! So who cares if you make the moral choice or not?! In order to avoid this, you need some sort of measure to make finite sense of the infinitude of worlds. Once you map worlds onto measure, sparing the puppy increases the puppy's measure (probably by quite a lot, since you're most likely to be in the most probable worlds) and can be justified again. I don't know how to justify acting morally in an infinite world without doing something like this, unless you abandon consequentialism entirely and say you should do it for the good of your soul or something.
This is part of the general concern that I don't think you've thought about the many ways that infinite worlds go wrong even *with* God and granting everything about God that you want to grant. Again, I'm not sure how you say that the probability of me being (my) world's tallest person is 1/10 billion, rather than 1/2 or undefined. In order to quantify likelihood of being in any class, including the very normal classes that we all think about every day, you need to do something about the infinite paradoxes. I'm not sure why you don't respond to this in the original post, but I think it's decisive and that a good response to this point would convince me much more than any of the other somewhat tangential things about psychophysical harmony or whatever.
Re priors: I'm not just rehashing the basic problem of priors. There's a more specific problem analogous to the moral knowledge problem. It's tenuous if you believe A based on B but B isn't explained by B. So it's unclear on what basis we hold, absent trusting we have reliable faculties for ascertaining the priors, that simpler worlds consistent with our experience are likelier than complex ones.
I'm not assuming epiphenomenalism. I agree consciousness causes stuff. The point is that there could be a different state that was disharmonious but had the same causal effects. Thus, it's a coincidence that the states that cause certain physical states pair harmoniously with them.
I agree that the UDASSA result is how the math works on the view, but I just think that's an obvious reason to reject the view. You shouldn't bet on a proposition if 100% of the copies of you betting on the proposition lose money. That's a surefire recommendation for getting money pumped. I also just find it intuitively ridiculous that you should be arbitrarily confident in a prediction when there are infinity copies of you that will make the prediction wrongly and only finitely many that will make it correctly.
My concern wasn't about moral concern tracking probabilistic weight. It was that your view requires we either give up ex ante pareto or do things that benefit finite people and harm infinite people by equal amounts. Incidentally, if every mathematical structure exists, then the same things happen no matter what you do, so I think that leads to your actions not mattering.
I don't think my view leads to moral skepticism because we have free will and causal decision theory is right. You don't cause other people with free will to act.
//Again, I'm not sure how you say that the probability of me being (my) world's tallest person is 1/10 billion, rather than 1/2 or undefined.//
I think it would be undefined by default. The idea is God uses some chancy process to place you in the world, such that the chancy process has a low probability in making you your world's tallest person. It's analogous to the Hilbert's hotel case I discussed in the article. I did, in fact, make that point in my original post.
The basic idea: if God uses a chancy process to put you in the world, then your credences are dictated by the chancy process rather than the world. Analogy: suppose God rolls a 100 sided dice and makes you the tallest person only if you get 1. Well, though there will be infinite people who get 1 and 2 and 3 and so on, you should think at 99:1 odds that you're not the tallest person. My view is that God does something analogous regarding our placement in worlds, using a random process.
I agree this is weird and it's not super obvious that this doesn't leave all probabilities irreparably fucked. But I'm sufficiently confident that UDASSA doesn't work that I'm much more willing to gamble on "eh, it's not so clear how things turn out on SIA, but it's at least not obvious that things break down, and the alternatives are much worse."
I think UDASSA is particularly bad if you grant that it's possible for there to be qualitative duplicates. These duplicates, despite being exactly the same and having exactly the same location in exactly identical worlds, just arbitrarily get radically different shares of the overall probability space. Because each person there is equally easy to specify, simplicity considerations will be of no help.
Lastly, I think UDASSA violates reflection. Suppose that there are infinite copies of you, each in an otherwise identical room with different numbers on the door. Now, UDASSA implies that you should think it's likelier you're in the rooms with lower numbers on the door. But it gets much worse! Now suppose that after a half hour, you'll change around the doors so that bigger numbers have disproportionately more complicated things written on the doors (e.g. one has something written on the door, two has a more complicated thing written on the door, and so on). Well now your credence in being in room 1 goes up, so does your credence in being in room 2, while your credence in being in later rooms goes down. But this is nuts! At the end of this process, you were in room 1 if and only if you were in room 1 at the beginning. If A and B both entail each other, you should have equal credence in them. This also sets up an easy money pump, where you offer a ticket on being in some high room (room 39738249732894) that you buy back later for a predictably lower price.
I don't want to relitigate all of these, but I'm really confused about your tallest person claim.
First, God can't roll a die and make you the tallest person iff it's 1, because He has to do this for everyone in your 100 person world, and if the die coincidentally lands on 1 twice, there are two tallest people. I think you can rescue this by saying He rolls until He gets a 1, that's the tallest person, and then He fills the world with however many other non-tallest people. This would require some ugly math, because height still has to follow a bell curve. While I suppose God could find some elegant divine way of solving this problem, I think it's more likely that He just sets everyone's height according to a Gaussian - or, more realistically, lets it follow the usual genetic laws that determine height, which naturally result in a 1/100 chance of everyone being tallest.
The bigger problem is - suppose He creates an infinite number of 100-person worlds like this. Now we're in the exact same situation we were in before, where you can move all the tallest people to one place and all the non-tallest people to one place and have two equal-cardinality infinite sets. In fact, if we imagine God lets things follow the natural genetic process, then this is exactly equal to the non-God situation where everyone has their height naturally.
I think you are just thinking of things as the multiple-worlds-are-undefined thing when it suits your arguments, and switching to the identical roll-a-die process when *that* suits your arguments. God doesn't help resolve the paradoxes either way, only the convenient paradigm-switching does.
All of your objections to UDASSA just seem like being surprised by math to me. For your last paragraph - it's not a paradox, you're just changing the probabilities, then acting surprised that the probabilities changed! Suppose I show you infinite doors. One has a car behind it, such that there's a 1/2 chance it's in the first door from the left, 1/4 chance it's in the second door, and so on. Then while you're not looking I "change the probabilities" (ie move the car) to a new distribution - 1/2 chance it's in the first door from the *right*, 1/4 chance it's in the second door, etc. Now instead of believing the car is somewhere near the left, you should believe it's somewhere near the right. This isn't a paradox, it's just that I changed the probability of where the car would be. I can't money-pump you on where the car would be, except in the fact that I can trivially charge you $1 to pick the left door, then say "Haha, I switched the car to be on the right, do you want to pay $1 to switch your choice to the right?" and keep doing that until you wise up and realize I will keep switching.
Yeah I don't think that merely having infinite people be in rooms, each with a label of 1-100, allows you to say that, at 99:1 odds, you have the label 1-99. There has to be some genuinely chancy process that applies to you specifically. It would however work out if God uses some chancy process to dictate your location in the room (e.g. randomly assigning everyone a room number so that one person is the tallest, another person the second tallest, and so on).
//The bigger problem is - suppose He creates an infinite number of 100-person worlds like this. Now we're in the exact same situation we were in before, where you can move all the tallest people to one place and all the non-tallest people to one place and have two equal-cardinality infinite sets. In fact, if we imagine God lets things follow the natural genetic process, then this is exactly equal to the non-God situation where everyone has their height naturally.//
I don't think this matters. Arrangement is irrelevant. What matters is the process that you know was applied to you.
The idea is that if there are infinite people with every property, you should have undefined credence in your having that property unless you know that some specific process was done to you that had some specific probability of giving you that property. As, for instance, in the die case. Theism can secure that kind of process.
Regarding the last objection, I think you're missing the core weirdness. Obviously your credence in some proposition at a time won't always remain the same over time as the car case illustrates. If you're flying to California, your credence in "I'm in California" should be very high but should change.
The difference is that in this case the second proposition is true if and only if the first one is! If there are two propositions which are either both true or neither true, then you should have equal credence in them. You shouldn't say "I am 90% sure there's water in my room, but only 80% sure there's H20 in my room."
The thing that's relevant is that changing what happens on your door is absolutely incapable of changing your room number so after it happens your credence in being in different rooms shouldn't change.
Definitely still not getting where you're coming from.
We got into this discussion because you said that even if there were one billion real humans and one Boltzmann brain per universe, the human:Boltzmann ratio was undefined. Why isn't whatever combination of physics, evolution, etc causes you to be either a human or a Boltzmann brain a chancey process, similar to the combination of physics, evolution etc that makes you a tall person rather than a short person? Isn't the former equivalent to rolling a one billion sided dice and making someone a BB iff they land on one? Dice work by a deterministic physical process just like physics/evolution, so how are they different? Or are you saying God uses some actual divine RNG doing things in a nonphysical way beyond human possibility in order to select the tallest person, and only results of the divine RNG are free from paradoxes of infinity? Why would this be true?
I think where we're failing to connect with the rooms thing is that you seem to think you're just changing what's on the door, but if I'm understanding you correctly, the analogy to UDASSA is changing which universes are most simple/complex, which is a genuine change that itself changes probability. Like, you say "UDASSA implies that you should think it's likelier you're in the rooms with lower numbers on the door." This is absolutely not true literally in a literal hotel (do we disagree on this?), so I'm taking it to imply that you mean it's figuratively true if the . . . numbers on the doors represent the complexity of the universe you're in? But then that's like saying "You're acting like changing the numbers on the door changes the chance of there being a car!" when the numbers represent the probability of there being a car.
By the way, I'd be totally happy to have a phone call if you want to discuss this more. If so, send me a dm. I tend to find that to often be helpful when lengthy written back and forths lead to both sides thinking the other one is missing something.
//Why isn't whatever combination of physics, evolution, etc causes you to be either a human or a Boltzmann brain a chancey process?//
Yeah this might be right. Will have to give it more thought. But the core problem is that on the Tegmark view there are infinite copies of you in different universes with your exact present experience. So figuring out which universe you're in is not subject to a similar chancy process.
(Having thought about it a bit more, I think it probably is right).
(Analogously, if you roll dice in Hilbert's hotel but there's a neighboring infinite hotel with hallucinating people of every variety, then you still should have an undefined credence in your being in the first hotel).
In a hotel, assuming you want a normalizable probability distribution over an infinite number of rooms, you will have to hold that the earlier rooms get a higher share of the probability space. But it just seems totally irrational to think you're likelier to be in room 1 than room 72.
The objection I gave to UDASSA was as follows. Imagine that you have infinite rooms or universes--doesn't matter which, just matters that you have something. Right now, universe 1 is very simple. However, in an hour, universe 1 will undergo changes which make it much more complicated (God will, by divine fiat, add a bunch of complicated stuff to universe 1).
UDASSA implies that your credence in being in universe 1 now should be higher than they will be in an hour. But you're in universe 1 now if and only if you'll be in universe 1 in an hour. So as long as you should have equal credence in mutually coentailing propositions, UDASSA has to be wrong.
Here's an analogously bad situation. Imagine that right now you think your friend's name is Fred at 60% probability. But you expect in an hour to think that his name is Fred at 80% probability. There's no chance his name will change. This seems quite suspect--your credence in some proposition shouldn't predictably go up in this way.
Sorry, I hate phone calls.
I still don't think you've separated out the world's-tallest-person scenario from the Boltzmann brain scenario.
Both of them involve infinite worlds. Each world has (let's say) one billion people, of whom 999,999,998 are unmarked, one is the tallest person in the world, and one is a Boltzmann brain. Both tallestness and Boltzmannness are caused by a deterministic process. In both cases your consciousness has to be placed at random into some person in some universe. What is the difference between membership in these two classes, such that tallestness plays nice with infinity but Boltzmannness doesn't?
I'm also still confused about your room scenario.
UDASSA doesn't say that universes get probability mass because of how complicated they are *now*. They get it because how complicated they are *to specify*. So even though our universe has many things, the "recipe" may take only a blackboard worth of equations to specify (it would like something like "take one point of infinite density, add gravity that works like [equation], magnetism that works like [equation], and the mass of the quark is X EV". In order to change how complex our universe is to specify, God would have to change the rules of math, which is contradictory.
I don't think God can simply go into the "material" universe, create some stuff, and then ask if the math changes too, because there is no material; it's *all* math. It's like asking "If I went onto the graph of y = x^2 on the Cartesian plane, took a giant eraser and marker, and messed with the parabola until it had a bunch of extra dents and curves in it, would that make the function y = x^2 itself have more terms?" It's not coherent enough to answer.
I guess you could phrase this as MUH not permitting the possibility of entities outside universes affecting those universes. With a marker and eraser, you can make a specific written-down graph of y=x^2 in a math textbook look different. But you can't affect the true Platonic graph of y=x^2 itself. Only the definition of the process of squaring a number can do that. But our universe isn't a specific written-down instance of the cellular automaton it represents, it's the automaton itself. So you can't affect it with normal tools (even if you're God). You just have to watch how the equation plays itself out.
(humans can affect it because we're just arbitrarily lines drawn around parts of the math, in the same way that the number five can affect the graph of y = x^2 by being 25 when it is squared rather than 26. This is basically the argument in https://www.lesswrong.com/posts/NEeW7eSXThPz7o4Ne/thou-art-physics , but with a friendly amendment changing "physics" to "math")
If you wanted to make your scenario meaningful in MUH, you could imagine God being a part of our universe, which would raise the universe's complexity a lot insofar as God is hard to describe in math. This is basically fine - there's some super low probability universe that's complex enough that it has God in it - but we should expect very low likelihood we're in that one.
Otherwise, when I try to imagine your scenario, I imagine God trying to add a very complex device on Mars or something, and then finding that oops, He's now in the different MUH universe where there was already a complex device on Mars (which is lower-probability than ours). If He objects "No! I meant to keep this in the higher-probability universe over there!" then it's fruitless - like trying to add a side to a triangle in Platonia and getting upset when you find yourself at a square rather than a 4-sided triangle.
>UDASSA implies that your credence in being in universe 1 now should be higher than they will be in an hour. But you're in universe 1 now if and only if you'll be in universe 1 in an hour. So as long as you should have equal credence in mutually coentailing propositions, UDASSA has to be wrong.
Wait, why does it imply this? The simplicity of a universe isn't a time-indexed thing in the way it prima facie sounds like you're suggesting here, though I admittedly might be misunderstanding. Rather, it will reflect the minimum description length that specifies its behavior as a whole (i.e., across all time), and that doesn't depend on what time it is right now, or how simply it happens to be behaving right now. Is there something very specific to UDASSA that's incompatible with this kind of thinking?
(Realistically, if the universe suddenly does start to act more complex-ily in a way you didn't anticipate, that should of course diachronically change your assessment of its simplicity. But it sounds like you're talking about a scenario where you do anticipate the change.)
"The idea is that if there are infinite people with every property, you should have undefined credence in your having that property unless you know that some specific process was done to you that had some specific probability of giving you that property."
How do we know, even in theism, which properties had such a process? Also, we do we deal with the fact that properties aren't independent: the tallest person is also going to have a good chance of having the longest legs, the longest arms, etc....
I'm reminded of Michael Huemer's discussion of Bertrand-type paradoxes: suppose you know a machine makes cubes whose side lengths are between 1 and 2 cm. What should you pick as the max entropy estimate of the expected volume?
On the one hand, the estimate for side length should be 1.5, halfway between the two extrema, which yields a volume of 1.5^3=3.375. On the other hand, cubes between 1 and 2 cm side lengths are the same as cubes between 1 and 8 cm^3 volume, and here the halfway point is 4.5
So which is right? Huemer says the second, because cubes, being 3d objects made out of 3d matter, have volume as a more "fundamental" property.
To avoid/resolve paradoxes like this, do we need to assume that God only applies this process to "fundamental properties"? Do we have any idea what those properties are?
I think your definition of a chancy process is going to be subject to all the problems you claim of the simplicity prior. You note that rolling a die in an infinite creation case is justified as yielding uniform probabilities of 1/6 not because of the proportion of people in each outcome set, but because "After all, as you repeat the process, it will be 1-5 most of the time."
This is the exact same limiting process you've disagreed with other people making! After all, the final "proportion" of 1s rolled won't be 1/5, it only holds if you imagine the die being rolled sequentially and take the limit of the proportion up to time t as t-> infinity.
But it's not clear why we would think of these die rolls as happening sequentially; and even if so, what if God using his all-knowingness, decides to roll the die so that two people who will get a 1 go first, then each of the other outcomes once, then repeat? You'll find that the limiting fraction is 2/7 for 1, and 1/7 for the other 5: rearranging the order of the rolls will give you other values for the probabilities.
The reason 1/6 seems natural is because you've picked a limit that feels natural, but that naturalness is grounded in intuitions from die rolls in non-infinite cases: whether I roll a finite number of dice sequentially or all at once, I get 1/6, so it seems like a natural answer to carry over to the infinite case.
But, though I'm not without some sympathy to this view, it really isn't clear that 1/6 is the correct probability, or that such a notion is defined. On the frequentist view, it seems to depend on what limiting process feels natural to take the limit of frequencies, and without knowing the exact process God uses, it's not clear there should even be a canonical limit, much less what its value is. And on a subjectivist view of probability... Can God even have subjective probabilities that aren't 0 and 1?
It feels to me like any assertion that one particular choice of limit, and hence probability, is canonical is going to feel arbitrary without a _lot_ of justification. Ironically, it strikes me as having the same issue you (correctly, IMO) identify with the simplicity prior: it won't _feel_ like a probability, it won't unambiguously do the things we want a probability to do for us; it'll just satisfy some abstract mathematical criteria.
Finally, it strikes me that we have no way of knowing what process God uses, and so even if this does ground probability in some sense, it's not clear that this is at all usable.
I think we know God cares about our cognitive faculties, so we have reason to think that we wouldn't be in a skeptical scenario! I agree that maybe some of the probabilities will turn out a bit unclear though.
The idea behind thinking 1-5 is likelier is:
1) If you take the limit as the number of die rolls approach infinity it will approach 5/6 getting 1-5.
2) It's reasonable to assign a difference principle over possible rolls that you might get--none of them are special--and so you know that in your case, there's a 5-6 chance you got 1-5.
The basic point is that if you're rolling dice, you just ignore what's happening to other people. If you didn't know about the other people, you'd assign 5/6 chance to getting 1-5.
You'd assign 5/6 if you understood the process ahead of time, and could draw the analogy to normal dice rolls. Whether that's justified is what's at stake!
For one thing, this die is being rolled by God, who presumably knows the answer to what will be rolled ahead of time.
What's more, while I certainly _can_ ignore what's happening to others, I don't see that I'm obliged to. Suppose I refuse to reason as you suggest, and endorse a different limit: God explains to me that a die will be rolled to place me in a universe, and I go ahead and calculate the probability that my die was rolled a 1 using my preferred limit, and say that my probabilities are (1/15, 2/15,3/15,4/15,5/15,6/15), not uniformly 1/6.
Am I _wrong_? What can you point to, to convince me that I've got the wrong probabilities?
Of course anyone can assign any credences they want. The question is simply which ones are substantively reasonable.
My problem is that the term "genuinely chancy" seems to imply something objective here; it's presumably not genuinely chancy from God's point of view since God knows the outcomes of all the dice rolls, so it must be "genuinely chancy" from _our_ point of view, but if so, it sounds like the probabilities can't just reflect our lack of knowledge or our application of some heuristic... What is "genuine" about the chanciness? It seems like you want to ground these probabilities in some objective features of the world, but the only options I can think of are frequencies, which won't avail us in the infinite case, or some canonical model for the process... So I think you have to make an argument not just that modeling the dice roll in the way you do is "reasonable", but that there is something objective about its chanciness that is model independent, or that picks out a unique model in a strongly justified way
Lol, I used the wrong triangular number in my probability vector, I obviously meant those all to be /21
Checkout my comment further down about the Universal Dovetailer if you have a moment, it goes into the details of what makes "UDASSA" tick.
Re: psychophysical harmony, if the physical aversive response of pain doesn't actually require the subjective qualia of pain, then isn't the fact that pain still exists actually a point against an omnibenevolent God? We could have had an aversive response that served our survival without the innate, immediate badness of emotional suffering, but God in his arrangement of our psychophysical harmony allowed that inherently negative mental state to exist. If the qualia of suffering is inherently bad, which I believe it is, then why would an all-good God saddle us with it when it in fact isn't necessary?
Bentham's God dislikes deception and also really likes connection-building. If people didn't have qualia of suffering, helping them wouldn't be worth anything, and there'd be no reason for us to come down here from our heavenly pre-existence. "Why would God give us qualia of suffering?" seems like the same question as "Why would God make us suffer at all, given that we can?" and I think is answerable by the same theodicies. E.g., https://benthams.substack.com/p/a-new-preexistence-theodicy
Helping people would still be good if it increases their happiness! Why is pain uniquely key to connection-building? Oh right. It’s because all of these theodicies are ad hoc nonsense with no independently motivating rationale.
Sure, but they're ad-hoc nonsense that explain the *possibility* of suffering just as well as they explain the *existence* of it. My point is that (if we assume God doesn't want to deceive us) there's nothing special about this objection that isn't just the problem of evil rehashed.
"Hey, look at all this evil, isn't that evidence against theism?"
"Yeah, but here's a somewhat plausible explanation for God allowing it."
"Haha dumbass, gotcha! Why wouldn't He just *make evil impossible*?"
Well, you ignored the very first line of my comment, which pointed out that there is in fact a unique cost to maintaining pain which isn’t clearly solved by connection-building theodicies. You should answer it before being so condescending.
I don’t mean to condescend, I’m just bad at being funny…
Look, I don't think you're actually right that just increasing happiness builds the same kind of connection. Clearly there's meaning in buying a house for a poor family that isn't there when you're buying a summer home for some very wealthy guy.
I think you'd also hit a pretty strange reductio. Like, if God could just remove the qualia of pain and connection-building still happens, why can't God also remove the qualia of slight-happinesses? Why not remove all of the happy qualia that are any lesser than heavenliness?
At some point along the way, connection-building must stop happening. If it's not the point of no-suffering, where is it? Is there any other spot that isn't super arbitrary?
> Clearly there's meaning in buying a house for a poor family that isn't there when you're buying a summer home for some very wealthy guy.
And that's why the most moral thing to do is make sure that everyone else is constantly poor and then buy them houses from time to time, right?
>At some point along the way, connection-building must stop happening.
And who decides this point, I wonder? May it be the omnipotent creator of everything?
> Clearly there's meaning in buying a house for a poor family that isn't there when you're buying a summer home for some very wealthy guy.
Obviously a case of diminishing marginal utility from getting extra material possessions. This has nothing to do with happiness versus pain.
> I think you'd also hit a pretty strange reductio. Like, if God could just remove the qualia of pain and connection-building still happens, why can't God also remove the qualia of slight-happinesses?
Sure. Why can't he?
> At some point along the way, connection-building must stop happening. If it's not the point of no-suffering, where is it? Is there any other spot that isn't super arbitrary?
Well, A. There is presumably some fact about when it does or does not stop happening. Alternatively, if its a question with no answer like "when do grains of sand become a pile?", that's a signal that "connection-building" is fake.
B. This is a strong argument *against* God. The world *is* insanely arbitrary. The idea that its somehow the maximally perfect state of affairs is absurd for that very reason! This is distinct from evil for--in part--pleasure/pain as argued above, and every other distinction about the very arbitrary *qualities* (as opposed to magnitude) of evil.
If God dislikes deception, why is the night sky full of seemingly randomly-distributed stars, rather than e.g. a quickstart guide to telescope construction, with diagrams and multilingual text (Chinese, ancient Egyptian, and proto-indo-european), branching out to a wider variety of subjects by fractally smaller print within each astral brushstroke?
> dislikes deception
You mean the same God who arranges absolute majority of the evidence such that the universe appeared to be naturalistic?
> If people didn't have qualia of suffering, helping them wouldn't be worth anything
What if people's emotional variance was moved from
suffering - neutral - pleasure
to
neutral - pleasure - absolute bliss
or even to
pleasure - absolute bliss - extremly amazing awesomeness that we do not have words for
or if at least the intensity of pleasure was not less than the intensity of pain?
If you do not have limitations imposed by natural selection of our universe and can make any rules - these are the obvious things to consider.
And I should add: notice how this argument breaks down on the exact same lines if you replace "people's emotional variance" with "people's emotional experience." The problem of the possibility of suffering seems equivalent to the problem of the existence of suffering.
> arranges absolute majority of the evidence such that the universe appeared to be naturalistic
You're question-begging—to a theist, the absolute majority of the evidence points to theism.
> What if people's emotional variance was moved from (suffering - neutral - pleasure) to (neutral - pleasure - absolute bliss)
Would it feel the same? Maybe see that bit below about diminishing marginal utility. You would need the highs to be much much higher than the lows, and it seems like there's a cap you'll run into ("heavenliness"). And connection-building isn't perfectly isomorphic to utility anyway—maybe God's a negative utilitarian about this. Alleviating someone's suffering creates a substantially different relationship from enhancing their pleasure—that seems very intuitive to me.
> to a theist, the absolute majority of the evidence points to theism.
Then why can't modern Abrahamic priests produce e.g. a plague of frogs on demand? Per Exodus 8:7 the Pharaoh's court wizards seemingly figured out how to duplicate that trick same-day, though not the off switch.
More generally, why the historical trend away from big, easily verifiable miracles, toward increasingly obscure examination-resistant claims and emphasis on inward, *evidence-denying* faith? https://acoup.blog/2019/10/25/collections-practical-polytheism-part-i-knowledge/
> I don’t quite get the complaint. I think God did set up some elegant system of psychophysical laws. I also think he had an elegant system for making it so that we’re each probably not Boltzmann brains. I don’t think God intervened to do a special miracle to solve each of these problems.
Let me try to explain. The complaint is, fundamentally about a faulty pattern in your reasoning proccess.
1. You really want something to be true - not being a Boltzman brain, being rightfully confident in inductive reasoning, moral/logical/modal facts being "stance independent", scepticism being wrong, etc.
2. You have some evidence in favor of these, but not enough to be confident in your prefered views. Actually, you have some very serious counterevidence: it would, indeed, be a huge coincidence if there were objective moral truths and we just happened to learn them via evolution, that wasn't optimizing for it at all.
3. Instead of lowering your confidence in your prefered conclusions, as one should in such circumstances, you assume it as a bottom line anyway, and then start motivatedly searching for some way to justify it.
4. You arrive to the motivated conclusion that theism solves all these problems. Therefore, not only you accept your prefered views, you also accept theism, even though it itself is questionable. P(A)P(B) < P(A). You would be less wrong by simply assuming your prefered stance A without enough justification. And yet you also accept B.
5. And then you try to find some justifications to theism, in the same motivated fashion. You start assuming theodicies that would "explain away" counterevidence to theism, even though that it actually makes your overal theory less probable P(A)P(B)P(C) < P(A)P(B) < P(A)
And so on and so forth. You keep building more and more epicycles, adding more and more details, to hide the error you did in step 3. You start inventing new problems that theism alledgedly solves. You do a tremendous effor to build an argument how naturalism fails to account for X, while simply assuming that theism surely does, without much investigation. At this point not only you are using "God" as a curriocity stopper, which prevents you from learning new things and deconfusing yourself, you are using it as a "knowledge eraser" - which makes even things we already have answers for, appear more confusing. Which is the opposite of what philosophy, ideally, is supposed to be doing.
>The same point, by the way, applies just as much to other facts about math, modality, logic, and what priors are reasonable.
As I have pointed out elsewhere, the sky doesn't fall in if you accept nonrealism about all four.
In reality, there are people born without the ability to feel pain, and they often are prone to serious injury. We can hypothetically describe made up systems that would replace pain, but nature has not provided any that we can see.
Okay, I know this is off-topic. But this argument:
> To give an analogy, imagine that there are some goblins who worship a being called the absolute. The goblins have an intuition that worlds with the absolute have a higher prior than worlds without it. But they know that they’d have that exact intuition whether or not there was the absolute, and that intuition isn’t explained by the likelihood of the absolute. It seems they lose their justification for believing in the absolute.
seems to apply very well to religion!
There are several popular religions with fundamentally conflicting views. Therefore, if one is religious, no matter which religion one subscribes to, one should agree that most people in the world are incorrect about their religious beliefs. Yet empirically speaking, the incorrect people and the correct people acquire their beliefs from basically the same sources: faith, social proof, and philosophical arguments, combining to produce an intuition of correctness. Therefore, it's likely that one would have the same intuition about one's own religious belief whether or not it was correct, so "that intuition isn't explained by the likelihood" of the religion being right. Does one lose one's justification for believing in their religion?
Admittedly, this argument applies far more to traditional organized religion than your own position. On the other hand, your conception of God seems to be at least strongly influenced by organized religion. (But then one could say the same about philosophy as a whole, and even atheism?)
Most people are incorrect about many of their beliefs, yet that alone is not enough to lose justification in believing something. Your argument as written applies just as well to atheism: for belief in atheism comes from the same sources as belief in God, namely your own experiences, philosophical arguments, etc. If you took a particular believer and asked him why he believed he would give you specific reasons: if you convinced that believer (like the goblins) that they would still believe even if God didn't exist, then that would be very good reason for them to doubt their belief in God.
It's also worth nothing that a religious believer does not (typically) believe that their beliefs are right and every other religion is completely wrong. Though they may disagree on specific doctrines there is a lot to agree with: the existence of the supernatural, the existence of God or gods, the existence of the soul, judgement after death, etc. Lewis wrote about this in "Mere Christianity"
"If you are a Christian you do not have to believe that all the other religions are simply wrong all through. If you are an atheist you do have to believe that the main point in all the religions of the whole world is simply one huge mistake. If you are a Christian, you are free to think that all these religions, even the queerest ones, contain at least some hint of the truth. When I was an atheist I had to try to persuade myself that most of the human race have always been wrong about the question that mattered to them most; when I became a Christian I was able to take a more liberal view. But, of course, being a Christian does mean thinking that where Christianity differs from other religions, Christianity is right and they are wrong. As in arithmetic — there is only one right answer to a sum, and all other answers are wrong: but some of the wrong answers are much nearer being right than others."
> When I was an atheist I had to try to persuade myself that most of the human race have always been wrong about the question that mattered to them most;
I think he's projecting quite a bit there. For most of the human race, through most of history, question that mattered to them the most was probably
"are the kids gonna be okay" or
"will this plan bring glory, or shame, on me and my buddies" or
"where's the best beer"
or something like that, not cosmological stuff. Atheists are certainly permitted to believe that religious practices can contain useful hints toward those sorts of questions, and that people have guessed right before, without any inconsistency.
> Atheists are certainly permitted to believe that religious practices can contain useful hints toward those sorts of questions, and that people have guessed right before, without any inconsistency.
Have guessed right about what: the existence of the supernatural? Or what the best beer is?
The second one, with concrete results - though a lot of old-school animistic beliefs could also be thought of as hacking human social instincts into a useful approximation of an ecological https://en.wikipedia.org/wiki/Agent-based_model rather than really being "supernatural claims" in a strict sense.
>The basic idea: on Tegmark’s view, there will be infinite people with every conceivable property. There’s no coherent way to talk about proportions of people with the different properties—after all, you could arrange worlds so that each world is filled with any percent of people with any property. You could make every world filled almost entirely with Schizophrenics named Phillis who evolved from bats but have cognitive capacities similar to humans. Thus, all probabilities come out undefined.
But, for one thing, you haven't showed that anyone is committed to setting credences solely based on proportions and nothing else when the proportions are undefined or otherwise mathematically meaningless. This thesis seems incredibly implausible/unmotivated to me and I'm not sure anyone in the literature has ever advocated for it, for whatever that's worth (admittedly not much - there's always room for original philosophical positions). It also seems completely costless to reject; I can't think of any terrible consequences to not signing up for undefined credences of this sort when I have other epistemic principles at hand to adjudicate things, principles which in fact seem much more independently attractive to begin with than going with proportions.
Second, you yourself don't actually hold to this commitment (i.e., "undefined proportions -> undefined credences") on any other issue! When deciding based on Ptolemaic or relativistic models to explain the motion of the planets, you go with the simpler non-Ptolemaic explanations, even if there's an infinite number of people living in both Ptolemaic and non-Ptolemaic universes according to your view. And, as I remarked in a comment the other day, you can't account for this by talking about God ensuring the reliability of abductive inference, because the question being asked here is how to abduce, not how reliable that process is once we've already figured out how to do it.
Finally, once you inevitably agree that abduction works out on your view in the cosmological case via the legitimacy of a simplicity-based prior plus ignoring undefined proportions, you have to explain why Tegmarkians can't also use simplicity-based priors and ignore undefined proportions.
Sorry, one other thing, although my comment was probably already overly long:
> However, literally 100% of people do not find themselves in those worlds. This is an odd result.
This is probably not correct on Tegmark's view; I think he's going to want to say that for any universe, there are unsetly many people in structurally identical copies of that universe. (This is because mathematical objects/structures can be instantiated in many structurally isomorphic ways - for example, we can define the ordered pair (2, 6) as the set {{2}, {2,6}} as is commonly done in set theory, but we could also have defined it as {{6}, {2,6}} or indeed a ridiculously huge number of other ways.) So, if this is a correct description of Tegmark, there's no actual meaningful proportion of people in a world like ours.
It may be the case that there's a finite collection of *classes* of structurally identical universes that takes up 99.99999+% of the probability mass on Scott's view, but why pay attention to classes and not people? It seems like you want to pump intuitions based on things like "most people in the multiverse are badly wrong if the view I'm arguing against is correct."
Let's go back from the beginning, because I think people are not quite grasping the profundity of this Tegmarkian view. The basic postulates of this cosmological view are:
- self-existing mathematical structures are *already* encoding every possible computation, including the one that simulates this very universe's physical laws, as well as every other possible computable universe. (the "self-existing" part is an axiom of the theory; this universe being computable is a common hypothesis of physics; the rest is just a basic theorem of computation)
- this is enough for the biological organisms in these 'simulated' universes to be conscious of themselves and their surrounding universe, as long as they have the (simulated) physical feedback loops necessary to sustain individual sentience.
- this 'simulation' operates in the manner of an Universal Dovetailer. The Universal Dovetailer ("UD" from now on) is a computational structure (a "program" for simplicity) that enumerates and executes every possible program, step by step, in an interleaved way. See https://alwaysasking.com/why-does-anything-exist/#The_Universal_Dovetailer for details.
Now, this is not a claim that there is a big computer at the bottom of everything (in which case you'd rightly wonder who wrote the program, and why it's coded in perl, etc.). Rather, the claim is that the UD is a self-instantiating logical fixed point. Since the UD simulates every possible computation, it also simulates itself, and every possible variation of itself. But you can easily prove that, within the UD's sense of time (an ongoing computation creates a structure of time within itself), longer programs will be simulated exponentially slowly, simply because there are exponentially more of them, and they share a the UD's finite computational capabilities. You simply cannot write a Universal Dovetailer in any programming language or mathematical representation that fails to simulate longer programs exponentially slower than shorter ones. (Also, the UD's finiteness need only be local; it can possibly be globally infinite and do a spatially infinite amount of calculations at each step, while still doing those interleaved steps one by one.)
So the simplest possible UD becomes a logical fixed point, because it's the one that gets simulated the most in *every* UD, including non-simple ones with useless epicycles. This is enough to ground the structural unicity of the simplest UD. If any logical structure has the capacity to be self-instantiating and infinitely generative, the UD is it, and due to any UD's bias towards the simplest instructions, the simplest UD wins out. From here on "UD" means this simplest UD.
This gives you a good basis to ground (accursed) anthropics. The outermost, most objective reference point you can find are not conscious observers within universes, or classes of them, or instants of observer-consciousness within worlds, but instants where the UD is simulating a conscious being. The steps where the UD is simulating insert matter like a stone obviously don't count for anthropics, because the stone (presumably) doesn't have the physical feedback loops that would allow it to sustain an individual subjectivity, let alone wonder what kind of world it finds itself in. And since the UD is (here, simplisticaly) modeled as a single-threaded computation, you can do stats on those instants, and ground anthropics on them.
So if you find yourself as a conscious observer within this Tegmarkian multiverse, the question becomes, what kind of world can you expect to observe?
The answer is simple: the mathematically simplest one that is still complex enough to give rise to a rich physics that can sustain evolutionary feedback loops. That's where the vast majority of computation steps instantiating consciousness are found in an UD.
And that happens to be a great match for the world we find ourselves in! Including discoverable physics based on relatively simple mathematical laws, with apparently arbitrary constants fine-tuned to make complex life possible. Everything that your theistic hypothesis explains, this explains too.
So let's compare this kind of theory to the philosophical theism that BB defends. We can define the theistic hypothesis as postulating an entity ("God") which is 1) self-sufficient, 2) generative, and 3) good. As BB has explained in detail, this predicts many of the features we find in the world, except for the bit where we'd (naively?) expect the world to be *more* good than it appears to be.
Now, self-sufficiency and generativeness are needed features of *any* possible cosmological foundation. You can't have a foundational theory without them. In this Tegmarkian view, the UD is also self-sufficient and generative. But in the theistic view, God is self-sufficient and generative *by fiat*. Here we actually have some structural justification for the UD to be self-sufficient (it's a logical fixed point), and generative (it's literally specified as a maximally generative program).
Note that God's goodness is not only problematic (see the epicycles of theodicy), but also an extremely onerous hypothesis. The sense of 'good' is one of the most complex things in human culture. Attempts to formalize it simply are also full of problems, pace Bentham and his defenders, human and canine alike. Like Scott hinted in one of his replies, the sense of "good" is an evolutionary product, the accumulated result of millions of years of biological evolution, plus ~100k years of much faster cultural evolution, of trying out and integrating strategies of collaboration/competition. To put *goodness*, of all things, at the cosmic level, as a fundamental self-existing element of reality is the most onerous hypothesis I can think of. Literally anything else would be simpler, including a tiger jumping at you right now out of nowhere.
And that's why the Tegmarkian view wins. It can do the whole job, including predicting a fine-tuned, comprehensible universe within which complex beings develop a sense of the 'good'. And it only requires some maximally general postulates - pretty much the bare minimum to have anything at all.
Note that I haven't included any reference to Solomonoff induction here. It's not really relevant — in actual physical worlds like ours, the way we do inference is by semi-formally noticing patterns and then checking whether they generalize. It may not be perfect, but it has done wonders so far. Our best historical inferrers, people like Thales and Newton and Hawking, never once thought "let me find the shortest Turing Machine tape that matches my sensory experience". Instead they took the common shortcut of abstracting away the subject, considering the object of observation in general rather than the specific angle they were seeing it from, then abstracting away time so they could have multiple observations, and then semi-formally looking for patterns in there. It starts with simple things like "x is proportial to y", all the way up to General Relativity and the information content of black holes.
Solomonoff may well be a nice way to formalize an ideal version of induction in an induction-friendly (Ockham-friendly) world, but it's not how we actually do things, and it's not a part of our Tegmarkian theory. So the whole of Joe Carlsmith's post about to how Solomonoff induction has alleged difficulties making sense of a mathematical universe (https://joecarlsmith.com/2021/11/28/anthropics-and-the-universal-distribution), and the whole discussion of "world and claw", don't really apply to the Tegmarkian view. Because that's just NOT how we discover physics in practice within our experiential world.
Edit: NOT, rather than "now"
> (If you’re thinking “no if physicalism is right this isn’t really imaginable I’d suggest you read the paper on the subject, as this worry is addressed at length).
I admit to only skimming this paper, but it seems like their response to physicalism is that psychophysical harmony is more likely under theism then physicalism. This is wrong. At least in *my* physicalist account of consciousness, consciousness is a *functional element of thought* and subject to evolutionary optimisation like every other part of us. There may be some ultra-contrived way to make psychophysical harmony fail anyway, but being contrived they're much less likely to arrise then more normal approaches. This whole problem is only a problem for you because you don't think that.
"This is surprising, given that there are infinite possible correlations between the mental and the physical,'
There are,assuming they are completely disjoint realms, as you do and Scott doesn't.
Do Goblins have moral knowledge? For instance, if (hypothetically) they were to attack a grove of druids, would they be doing it whilst knowing that they are committing evil? Would they simply be ignorant that causing suffering is evil? Would they argue for the morality of their actions through Divine Command from the Absolute? If so, would they be justified? Would things be different if they were Trinitarian Absolutists?
Goblinoid metaethics si such a fascinating subject of inquiry.
"So long as you keep A and C the same, B is totally irrelevant."
In the isolated incident, sure. But what about the next time I want the experience of eating a pickle? I could A) stick my hand in a blowtorch, or B) eat a pickle.
Evolution would clearly favor B and discourage A. A species of pickle-taste-maximizers that get their pickle kicks from eating pickles will outproduce a species that gets their pickle kicks from self-immolation. I'm not sure how we can consider B being irrelevant on an evolutionary scale, or even the scale of a single lifetime.
You're assuming that wanting to eat a pickle would cause you to eat a pickle. But the whole point is that's improbable!
I'm saying that wanting pickle-taste would cause me to do something. If that something is light myself on fire, I'm not going to live long enough to reproduce. On naturalism, we would expect that species where "want pickle taste" is fulfilled by pickle eating, or at least some positive action, would be better at survival.
We could imagine a creature that responds to all stimulus with pickle taste. Mutations in that species that progressively loses its pickle taste response to harmful stimuli (such as blow torches) are going to out compete omni-pickle-tasters. This wouldn't cause this creature to only eat pickles, pickle taste would also come with any positive experience (reproduction, etc.). But I guess in this analogy pickle-taste is a dopamine response?
All this is complicated by the fact that I don't like pickles. Pickle taste would be a functional substitute for pain, for me.
Whether after tasting a pickle you'd want pickle taste will depend on the psychophysical laws. It would be unlikely that it does or that wanting a pickle taste is associated with getting a pickle.
Personally, my response is functionalism. And I'm not convinced by the linked paper's rebuttal, which seems to be summed up as: "if you think dualism might be true, then atheism at least potentially has a problem, whereas theism definitely doesn't have a problem; this makes theism more likely".
First of all, this is a excellent reason to discard atheistic dualism, but once you've done so, the argument no longer points towards theism.
Besides, I think that functionalism is a far more elegant solution to the problem than theism. Why do this physical thing ('the brain's perception') and this abstract thing ('conscious perception') happen to be exactly identical in all measurable ways? "God made it that way because he's good" works, but "because they're actually the same thing" is more satisfying. Plus, the same exact reasoning applies to morality, where the physical thing is 'what we evolved to feel morally good about', and the abstract thing is 'what is actually morally good'.
Finally, it also feels like a just-so argument that God fine-tunes some things extremely precisely and obviously (psychophysical harmony, moral knowledge, universal constants) but fine-tunes others indirectly and unobservably (tendency of the universe towards moral good).
That's not the paper's objection! Functionalism doesn't help for the reason explained in the paper.
I find the paper's objection to a Functionalism explanation unconvincing.
"it is crucial to the original argument that information about psycho-functional correlations is not treated as part of our background knowledge."
But why wouldn't we? Evolution "acts" a posteriori, so functional correlations would be considered. Even if "evolutionary forces cannot affect the psychophysical laws", it can select for which laws are used.
To use their boulder island analogy, I see psychophysical harmony not as encountering a Stonehenge like structure, but instead seeing all the boulders at the bottom of a ditch or other areas of low elevation. Yes, someone could have placed them there, but I see no reason to give that a higher probability than them just rolling there naturalistically.
🔥
Regarding your argument that atheism makes induction unjustified, I think there’s an issue with the argument which appears pretty similar to a problem I have with many other of your arguments and which is sort of similar to Scott’s complaint about God solving any loose ends. You appear to be arguing that induction is not justified because after all the evolutionary explanation of induction itself would be equally compatible with a universe where induction stopped working tomorrow. It is true that you don’t have this problem if God exist because he would give you a different probability if you were in that universe. But the actual evidence that either theory has to explain is your actual probability because beyond your probability distribution, you don’t actually have any evidence that the probability distribution is justified in this way. Basically, you seem to be arguing that you believe in induction. Therefore, induction must be justified in the sense. I have mentioned above, and God would provide such a justification, whereas evolution does not without the existence of God. But the only reason you think induction should be justified in the sense is because you believe in induction so you don’t actually have any independent evidence that induction is justified in this sense and therefore all that needs to be explained is the evidence that you believe in induction, which is equally likely given evolution.. also given that the only reason you have to believe that induction is justified is your belief in induction and it’s your belief in the existence of a justification which causes you to believe in God, he doesn’t actually provide any independent justifying evidence of his own that is not completely circular on your belief that induction is justified. I am having trouble putting this in words, but my basic intuition is that if you’re going to indulgent this type of reasoning where your belief can itself be used as evidence to come up with a circular justification, it seems equally good to simply take your belief has given and get rid of The unnecessary justification, which has no actual justification for believing in it beyond your belief.
I'm not sure your point about UDASSA is right if you have uncountably many worlds. If your worlds are indexed by the real numbers in [0, 1] and have a uniform (or any other continuous) probability distribution on them then the probability you're in any particular finite subset of those worlds is 0.
Another thought: does this idea of a "chancy process" dictating probabilities in an infinite world undermine the case for thirding in Sleeping Beauty? If infinitely many versions is SB will experience all possible awakenings, even in a single SB, are we obliged to take our probability from the underlying chancy process, i.e. the coin toss?