You unexpectedly find yourself sitting in a windowless room across from a gray-haired gentleman. You didn't wake up there; you were walking down the street and cut to camera two, a white windowless room with a table and two chairs. After a moment, the gentleman speaks:

"You are dead, killed instantly by a small meteorite. Incidentally," he smirks, "you have lost Pascal's Wager. You may 'cross-over' once you can accept that you are dead. I am here to help in that endeavor and can present any evidence you desire."

You, being a stone-cold rationalist, will only reach this conclusion on the basis of solid evidence. He, being extremely ethical, will neither present false evidence nor attempt to undermine your rationality. What can he do to convince you that you have died?

I suspect there is nothing he could say or do to convince you of this. Rather, for any sufficiently "final" definition of physical death, there's no way he can demonstrate that you have somehow come out the other side. That's my wager: there is no sound way to convince someone, even while in the afterlife, that there is such a thing; thus, we should never believe in an afterlife knowing that we could never accept it even if actually there.

Am I wrong? Has this been proposed before? Is there any thing which, while actually true, could never be demonstrated in this manner?

I think that, if correct, this may point to a special class of untruths. Sort of... Bayesian contradictions, things which could never be sufficiently demonstrated.

Naturally, lukeprog's earlier post has me thinking on religious lines.

New to LessWrong?

New Comment
27 comments, sorted by Click to highlight new comments since: Today at 6:37 AM

I'm pretty sure I could be convinced of an afterlife pretty quickly. Obvious things would be allowing me to see my dead body and funeral, and the other events surrounding my death. I'd also like some evidence that I'm talking to a being with powers that humans generally lack. The most obvious ways to test that is to pick some list of statements that probably have proofs (e.g. the Riemann hypothesis, whether there are infinitely many Mersenne primes, possibly a few substantially weaker statements thrown in) and to have it pick one and present a proof or disproof to me. It is possible that it is wrecking with my brain to think that I'm seeing a valid proof when it isn't but I don't assign that a very high probability. I'd also be interested in seeing personal information that no human has an easy way of knowing (and I have a few obvious examples ready on hand). This all wouldn't make me certain that I'm in an afterlife- a sufficiently powerful alien force could duplicate this sort of thing, but it would be assigned a pretty high probability.

Incidentally, I'm not sure I agree with the statement that the situation demonstrates a failure of Pascal's Wager by itself. Nothing in the hypothetical says that there's an afterlife where one is rewarded iff one believed in the right deity. By itself, finding out that there's actually an afterlife might even be a pleasant surprise. If for example we're in a simulation and the simulator cares enough about intelligence that all intelligent life is not only backed up but gets to keep processing, that would be a good thing. Even if the afterlife is due to a more classical theistic universe where it turns out that we really have "souls" or some other ontologically irreducible portion of our consciousness, that seems not obviously bad.

[-][anonymous]13y120

Well, one question is to define specifically what death he wishes me to accept before I leave the room. So when he says "You are dead." And I ask "Which kinds of dead am I?" If he says:

"You are information theoretically dead. You're also brain dead and clinically dead." then that leads to an entirely different kind of much more confusing discussion. (What exactly is he arguing with? If the answer is my soul, what is the nature of my soul? An 'Elan Vital' style answer doesn't clarify anything unless learning to accept an 'Elan Vital' is the point of locking me into a room, in which case that will cause me to consider the room in an entirely different light.)

"You are information theoretically alive, but you are brain dead and clinically dead." (Your pattern of experiences has been teleported by some means to a locked room. You can resume experiencing things that are not the locked room when yoou have seen sufficient evidence that you accept you are brain dead and clinically dead.)

Or perhaps, "You're information theoretically dead, brain dead, but clinically alive." (You are a comatose vegetable with irreversible brain damage.)

Or perhaps even, "You're information theoretically dead, brain alive, but clinically dead." (It might be more accurate to say you've suffered massive trauma, exploding your body and head and throwing your intact brain from it's body. There is no state of affairs where you don't die, but technically your brain is still oxygenated for a few seconds and is producing a few last gasp hallucinations.)

Also, another question is "Do I even WANT to 'cross-over?' " I mean, he says he is here to help me in that endeavor, but I don't know if every rationalist would even necessarily be endeavoring to do that. Particularly not if told that by someone who also told them they lost Pascal's wager. If I HAVE been transported into something similar to the Christian afterlife, accepting what might be the Devil's implied axioms without double checking them seems likely to be a mistake.

Another pertinent question is "Which religion on earth is this situation in this afterlife closest to and how close is it?"

If the answer is "Catholicism. Mostly accurate with a few details wrong." That may also lead to an entirely different approach from "Atheism. And really, not even Atheism is that close."

I also want to note something else about your original post:

Is there any thing which, while actually true, could never be demonstrated in this manner?

Gödel's first incompleteness theorem talks about things that relate to this concept. See http://en.wikipedia.org/wiki/G%C3%B6del%27s_incompleteness_theorems for details.

In order for this to make any sense, you have to taboo some words and expand out the metaphysics, and in particular what's meant by the word "death". The scenario you've described sounds more like "teleported into an alternate universe" (or copied into the new universe after being destroyed in the old one), with the destination universe being an afterlife. The scenario says that you're having a conversation, and you can't do that without a working brain, which is my definition of being alive. But the "teleported into an alternate universe" hypothesis isn't very hard to test; you could look for differences in the laws of physics, or look through a portal into the original universe, for example.

The way to convince me that an afterlife exists would be to negate the evidence that currently convinces me that there isn't one. This is essentially that :

  • Physics is pretty well understood and there's no evidence that following the laws of physics from the starting conditions of the universe would create an afterlife.
  • There's not a clean cut between "brain damage" and "death" nor is there a clean cut between "brain damage" and "experience". How would an afterlife know when to scoop someone up from this world? How would it know what personality to give them in the afterlife? Especially if someone dies gradually of a mental illness.

If I were shown why these reasons were wrong or confused then I might be convinced.

Even is there was something that I couldn't become Bayesianly convinced of, this still wouldn't show it to be false. If such a thing existed it would be a problem with Bayesian reasoning.

Logic stays true, wherever you may go,
So logic never tells you where you live.

From the Parable of Hemlock.

[-][anonymous]13y70

I think this is some kind of absurd hypothetical situation that is not relevant to practical decision making.

[-][anonymous]13y60

If you aren't starting with a prior of 0 or demanding a posterior of 1, what's the problem?

If you are, you're doing it wrong.

If you aren't starting with a prior of 0 or demanding a posterior of 1, what's the problem?

Strictly speaking this isn't the only way you can be not convinced. I can assign a non-zero prior to something even if there's no evidence that can cause me to update. Pick some non-falsifiable hypothetical entity whose sole behavior is that it doesn't interact with the universe in any testable way. I can consistently assign a non-zero probability to its existence (and in fact can consistently assign any probability I choose) but no matter what probability I start with you won't be able to make me update my probability.

Not necessarily. For example, if I showed you a theory of physics that's simpler if you postulate such an entity's existence, you might very well update your estimate due to Occam's razor.

This, incidentally, is what Eliezer's argument for many worlds boils down to.

Well, that's kind of the issue I kept thinking about: does the situation secretly rely on a contradiction or tautology to drive the result?

[-][anonymous]13y30

As other people have pointed out, there's evidence like your funeral and explanations of the actual laws of physics. Is there some reason to believe that things like that couldn't bring your posterior probability up to a reasonable level?

I don't see how this would sufficiently raise my PP, however. It seems like I'd get a stronger belief in a simulation argument in tandem with any increased belief in an afterlife. Rules of reason would have me siding with simulation every time, it seems,

I would ask the deity what computational substrate it was running on. Then the conversation would branch depending on its answer.

I feel that some of this hinges on the meanings of "dead" and "afterlife".

Since "dead" means I cannot experience anything, there is absolutely no way I could be convinced that I am dead -- the mere idea is self-contradictory and logically impossible. No one can ever convince me I am currently dead.

"Alive, but in an afterlife" is different. This I think I could be convinced of, in principle, but I'm not yet sure how I would go about distinguishing this from the dream hypothesis.

Pascal's wager is about belief in God, not about belief in the afterlife, so that phrase is a bit clumsily inserted there; unless he's telling us that we're going to hell as soon as we believe we're dead, which is a rather strong disincentive to pursue the truth of the matter...

Anyway, I'm not certain I understand the scenario. If we assume he provides no false evidence, then the mere statement "You are dead" should be already considered sufficient evidence, no? (For the sake of simplicity I'm currently ignoring minor scenarios where he's himself mistaken or deluded)

Now, if we don't assume perfect truthfulness, the question becomes more difficult. But at the point where we can be allowed to witness our funeral, converse with the souls of other dead (and ask them in regards to several historical mysteries as well), etc and see that the whole thing fits together as well and as badly as history should, then certainly we'd have increased reason to believe in the existence of afterlife.

We know he doesn't provide false evidence, but the person in the scenario doesn't know that. How could they distinguish between that scenario and the scenario where the gentleman lies when says he will always tell the truth.

If we assume he provides no false evidence, then the mere statement "You are dead" should be already considered sufficient evidence, no? (For the sake of simplicity I'm currently ignoring minor scenarios where he's himself mistaken or deluded)

I think this is a meta-level restriction. That is, we (outside the hypothetical) know that he isn't going to present any false evidence. But the you who is sitting in the room doesn't know that.

I'm not sure what the no false evidence rule adds to the situation.

It's base covering, there may be false statements he could make to convince you but being uninterested in them, I decided to exclude any which may exist. The non-undermining rule serves the same purpose.

. If we assume he provides no false evidence, then the mere statement "You are dead" should be already considered sufficient evidence, no?

Even if not, the next natural inclination would be to demonstrate power over souls/minds. This would most easily be done simply by causing me to believe in the afterlife directly and making me aware of this causation of belief while doing so. (This doesn't, I feel, impede rationality because it does not prevent me from reassessing the situation based on my newfound set of priors.) Alternatively, causing me to experience mindspaces hitherto unavailable to me would also help in demonstrating a separation of traditional physiology and mentality. Witnessing all places on Earth simultaneously, or other such parlor tricks, would also do nicely there.

Would they prove the legitimacy of the supernal claim of an afterlife? No, not at all. But they would demonstrate a separation between cognition and physiology, from which the extraordinary nature of a specific afterlife would be lessened to the point where merely being shown it would qualify as justification for naive belief.

If the gray-haired gentleman is the Simulator of our universe, then presumably he could demonstrate this by allowing you to "view" the simulation, in particular your decomposing corpse, grieving loved ones &c. Also, he could further replay to you events from the past: events you remember vividly, and perhaps some historical ones too. And also, he could allow you to spectate on the ongoing simulation.

Of course, he would no doubt have to provide a considerable number of bits of information in order for the Afterlife hypothesis to become more likely than the alternatives such as dreaming, drugs, major brain malfunction, &c.

This also doesn't differentiate 'dead but sentient' from 'brain in a vat', and I think the prior for the latter is a bit higher than the prior for the former.

I'm not sure what the difference is between 'dead but sentient' and 'brain in a vat'.

I was assuming that we want to distinguish between:

  1. The universe is simulated and the Simulator has the power to preserve minds even after their bodies in the simulation die. (This may or may not include brains being in vats.)

  2. You are still in this universe and someone is trying to trick you into thinking that (1) is true.

For me, the entity would have to give evidence which makes his explanation more likely than other hypotheses, one of which being that I clinically and brainically but not information theoretically died, suffered some retrograde but no anterograde amnesia in the process, was cryopreserved and then woken up again in the future by a post-singularity society, while the AI extrapolated my volition that I'd rather believe I was dead and in some sort of afterlife. That, for example, has a far higher prior than some sort of genuine afterlife.

Convincing me that I was clinically (and to a slightly lesser extent also brain) dead would be quite possible, though.

Your argument seems to be that it is impossible to convince someone that they do not exist. It's really just a restatement of I Think Therefore I Am. And persuading someone otherwise is indeed something that likely cannot be done, since even if you are dead, if you are considering these propositions, you still "exist" in some sense. The reason why this question is tricky is that we don't exactly know what we mean when we say the word "exist".

If I am understanding you correctly, then I think you picked a very bad example, since the whole basis of afterlife-ism beliefs is that your body can cease functioning, but you can still exist without your body. In which case, all he has to do is provide enough evidence that your body is now destroyed (which would probably be harder than it seems).

This is another possible response I considered, but could not decidee whether it actually stacks up. There are differences in the cases, but I don't know whether they're relevant.

Extensive knowledge of my private thoughts, revulsions and desires, in particular the ones embarrassing enough that I've never cared to share them with anyone.

Why would it be more likely that you're speaking to a deity than that you are in a simulation speaking to the principal investigator of an experiment or some other non-theistic scenario?

The difficulty I have with this thought experiment is that I can't decide how to distinguish between the hypothesis that there is a deity with whom I'm now conversing, and the many hypotheses that preserve a purely naturalistic universe in which my brain (or a simulation of my brain) is receiving coherent sensory inputs that make it seem like I'm interacting with a deity who can read my mind and show me absolutely anything I ask for -- he could even give me the memories of having proven the Riemann hypothesis to my satisfaction, of having taken me to my funeral...

My gut feeling is that the simulation hypothesis and some other non-theistic hypotheses have higher prior probability for me, and any evidence for the theistic alternative is also consistent with the simulation and some other hypotheses -- which I guess indicates with a problem with simulation hypothesis in this case, since it's not falsifiable.

One of the primary ways to distinguish between a simulator and a deity is simply what they claim to be. It seems unlikely to me that a deity would claim to be a simulator. I can more see reason why a simulator would claim to be a deity but it still seems not like a likely course of action.

Assume for a minute that Simulator and Deity are the only two hypotheses with substantial probability mass. Then P(Simulator|says it is a simulator)> P(simulator). So by conservation of evidence, P(~ Simulator|says ~ simulator) > P(~Simulator) so, P(Deity|says is a deity)> P(deity). The question becomes then by how much?