dadadarren

Comments

Leslie's Firing Squad Can't Save The Fine-Tuning Argument

Well, in that case, our arguments actually have a lot in common. My position regarding anthropics is that perspectives are axiomatic in reasoning. So a valid argument/notion must be formulated from one single perspective. This postulate refutes the probability shift in fine-tuning. It also invalidates the notion of self-locating probabilities like in the case of the doomsday argument.

Leslie's Firing Squad Can't Save The Fine-Tuning Argument

Very interesting...

I think proponents of the fine-tuning arguments for design are saying there doesn't need to be a reliable way to assign a prior. You can assign any prior you deem reasonable. Nonetheless, after considering our seemingly unlikely existence, the probability would greatly shift towards a teleological conclusion that the universe is designed for life.

So unless you are willing to commit that not only there is no reliable way to assign a prior, but also assigning a probability in this situation is invalid in itself, it doesn't counter their argument per se. It would be just pointing out even with the probability shift we should still be skeptical that the universe is designed due to the unknowns (but less skeptical than before considering fine-tuning). Are you saying what I think you are saying?

Just to be clear, I am not against the invalidity of probability in this situation. In fact, I probably support it more than you would like. I just choose to counter the proposed probability update because that is more direct.

Anthropic Reasoning and Perspective-Based Arguments

At this point, I think it might be more productive to list our differences rather than try to convince each other.

1. I say a probability cannot be formulated from a single perspective is invalid. You say it doesn't matter.

BTW, you said Question 2 and 3 are no different if both labeling outcomes actualizes in two parallel universes. Yes, if that is the case Question 2 and Question 3 are the same. They are both self-locating probabilities and both invalid according to PBA. However, what you are saying is essentially the MWI. Given I already argued against MWI's origin of probability that is not a counter-argument. It is just a restatement of what I have said.

2. I say SIA is based on an imagined sampling from a reference class. You say it is not.

Here I am a little upset about the selective quoting of the lesswrong wiki to fit your argument. Why not quote the definition of SIA? "All other things equal, an observer should reason as if they are randomly selected from the set of all possible observers." The set of all possible observers being selected is the reference class. Also, you have miss-interpreted the part you quoted "SIA is not dependent on the choice of reference class, as long as the reference class is large enough to contain all subjectively indistinguishable observers." It is saying the choice of reference class, under some conditions, would not change the numerical value of probability. Because the effect of the choice cancels out in problems such as sleeping beauty and doomsday argument. Not that there is no reference class to begin with. Also, just a suggestion, given the ongoing nature of the debate of anthropic principles and the numerous paradoxes, try not to take a single source, even the lesswrong wiki, as given facts.

3. You think SIA solves all anthropic paradoxes. I think not.

To name a few problems that I can think of right away: Does every single observation we make confirms the MWI? Refer to the debate between Darren Bradley and Alastair Wilson if you are unfamiliar. Does my simple existence confirm the existence of the multiverse? Applying SIA to the simulation argument. Wouldn't my existence alone confirm there are numerous ancestor simulations? In that case, wouldn't "I" be almost certainly simulated? Contrary to the simulation argument, applying SIA would suggest the great filter is most likely ahead. So we should be pessimistic about reaching a technologically mature state as described by the simulation argument. In Dr. Evil and Dub, what conclusion would SIA make? Is the brain arm race correct? Does my existence already confirm the arms race has already happened? Can all the above questions be satisfactorily answered using one consistent choice of reference class?

Basing on your opinion previously given on some paradoxes, it seems you think they have idiosyncratic explanations. Which is not wrong per se. But it does seems ad hoc. And if the idiosyncratic reasons are so important, are the paradoxes really solved by SIA or those individual explanations?

Anthropic Reasoning and Perspective-Based Arguments

I don't think I can give further explanations other than the ones already said. But I will try.

As for the difference between Questions 2 and 3, the fundamental difference is the probability cannot be formulated from a single perspective for 3. It requires the assumption of an innate reference class for the indexical "I". Both different from Question 2 which is regarding a random/unknown experiment. Again the argument has nothing to do with Omega can or cannot give definitive and differentiable answers to either of them.

Bayesianism is perfectly capable of assigning probabilities here. You haven't actually argued for this claim, you're just asserting it.

I only asserted that perspective is an important starting point of reasoning like an axiom. Arguments cannot be formulated from one consistent perspective is therefore false. That includes SIA, SSA, FNC, any notion of reference class for indexicals. And of course self-locating probabilities. I have also shown why self-locating probabilities cannot be formulated with a frequentist model. The same assertion about perspective's axiomatic importance also leads to other conclusions such as rational perspective disagreement. Whether or not my position is convincing is of debate. But I feel it is unfair to say I just asserted self-locating probabilities' invalidity without arguing for it.

You can, of course, do this for any question. You can refuse to make any predictions at all. What's unclear is why you're ok with predictions in general but not when there exist multiple copies of you.

I am not refusing to make a prediction. I am arguing in these cases there is no rational way to make predictions. And keep in mind, the nature of probability in MWI is a major ongoing debate. The fact that the probability comes from a complete known experiment with a deterministic outcome is not easily justifiable. So I think self-locating probabilities' validity should be at least debatable. Therefore I do not think your assertion that "Bayesianism is perfectly capable of assigning probabilities here" can be regarded as an obvious truth.

I don't see any paradoxes. SIA is the natural setup.

I think this is our fundamental disagreement. I do not think all anthropic paradoxes are settled by SIA. Nor do I think SIA is natural. Whatever that means. And I am pretty sure there will be supporters of SIA who's unhappy with your definition of the reference class (or lack thereof).

Anthropic Reasoning and Perspective-Based Arguments

I am not sure about "expectations" in this context. If you meant mathematics expectations, i.e. the arithmetic means of large independent repetitions, then as I have demonstrated with the frequentist argument, the relative frequency does not converge on any value. So expectation does not exist for self-locating probabilities. If the"expectation" here just means the two alternatives, being the original vs the clone, are meaningful facts then I agree with this assessment. In fact, I argue the fact about the perspective center is extremely meaningful that they are axiomatic. So there is no rational way to assign probabilities to propositions regarding it. My argument against self-locating probability does not depend on the observer never being able to find out the answer. It does not hinge on the lack of record or differentiable observables. The answer (e.g. "original" vs "clone") could be on the back of a piece of paper laying right in front of me. There is still no way to make a probability out of what it says.

If you are familiar with the Many-Worlds Interpretation (MWI), its proponents often use self-locating probability to explain the empirical randomness of quantum mechanics. The argument typically starts with experiments with highly symmetrical outcomes (e.g. measuring the spin of an electron at along a perpendicular axis). Then after the experiment but before observing the outcome, it is argued the probability of "me" in a particular branch must be 1/2. However, one counter-argument is to question the validity of this claim. Why there has to be a probability at all? Why can I just say "I don't know"? Sean Carroll (one of the most famous supporters of MWI) calls it" the simple-minded argument, yet surprisingly hard to counter." (not verbatim) In the MWI construct, the experiment outcome is definitively observable. Yet that does not automatically justify a probability to it.

PBA starts with the plausible assumption of the importance of perspectives, the invalidity of self-locating probability is one of its conclusions. I think it is much less ad hoc than simply making the judgment call of saying there should/shouldn't be a probability in those situations. If we say there should be a probability to it, then it comes to the question of how. SSA or SIA, what counts as an observer, which reference class to use in which case etc. There's one judgment call after another. Paradoxes ensue.

Regarding the individual analysis of the paradoxes, I understand your position. If you do not agree with the invalidity of self-locating probabilities you would not agree with the takedowns. That is the nature of PBA. There is no flexibility in the argument such as the choice of reference classes like other schools of thought. Yet I would consider it an advantage rather than a weakness.

Hello ordinary folks, I'm the Chosen One

Basing on the reply I am not very certain of your exact position. I kind of suspect it is implying the multiverse response to fine-tuning. It suggests the reason for observed fine-tuning is because there are many universes in total, and only in the ones compatible with life can give rises to observers pondering upon the parameters. Therefore finding ourselves in a universe compatible with life is not a surprise. I.e. It is not statistically incredible because of the huge number of universes out there.

I have to say that answer is very problematic. It treats "I" or "us" as the outcome of a sampling process subjected to survivorship bias, which interprets the WAP as an Observer Selection Effect (OSE). This conceptual selection has to be done from a god's eye perspective. It makes the same mistake as the fine-tuning argument by mixing first-person reasoning with objective reasoning.

In my opinion, this actually justifies the fine-tuning argument. Furthermore, it hijacks the anthropic rebuttal (which should be a simple tautology based on consistent perspective thinking). It also leaves the door open for rebuttals such as Leslie's firing squad and the fine-tuned multiverse.

In a sense, the fine-tuning argument is still an ongoing debate because currently, anthropic reasoning is inadequate. It is filled with paradoxes and controversies. All popular assumptions (SSA, SIA) treats indexicals as the outcome of some sampling process, implying the OSE. My Perspective-Based Argument (PBA) is an attempt to change that.

Anthropic Reasoning and Perspective-Based Arguments

"The drivers in the next lane are going faster" is true both from a driver's first-person view and from a god's eye view. However, none of those two are self-locating probabilities. This is explained by PBA's position on self-locating probabilities, by the link mentioned above.

The lane assignment can be regarded as an experiment. The lane with more vehicles assigned to it moves slower. Here, from a god's eye view, if a random car is selected then the probability of it from the slow lane is higher. From a driver's first-person view, "I" and the other drivers are in symmetrical positions in this lane assigning experiment. So the probability of me being in the slow lane is higher. According to PBA, both probabilities are valid. However, they are not the same concept, though they have the same value. (This point have been illustrated by Question 1 and Question 2 in the link above)

However, neither of them are self-locating probabilities in the anthropic context. Some anthropic reasoning suggests there is an innate reference class for indexicals like "I". E.g. SSA assumes "I" can be considered a randomly chosen human from all humans. This requires both the first-person to identify "I" and a god's eye view to do the choosing. It does not depend on any experiment. Compare this with the driver's first-person view above, the reference class is all drivers on the road. It is defined by the lane assigning experiment. It does not even matter if other drivers are humans or not. They could be all pigs and they would still be in symmetrical positions with me. The PBA argues the self-locating probabilities are invalid. (This point has been demonstrated by Question 3 in the link above.)

Since we are discussing Nick Bostrom's position, he made the explicit statement in "The Mysteries of Self-Locating Belief and Anthropic Reasoning" that an experiment is unnecessary in defining the reference class. We can always treat "I" as the result of an imaginary sampling process. This is in direct conflict with my PBA. According to PBA, anthropics is not an observation selection effect, just recognizing the perspective of reasoning.

Lastly, you are clearly interested in this topic. I just find the questions you raised have already been covered by the argument presented on my website. I can only recommend you give it a read. Because this question and answer model of communication is disproportionally effort heavy on my part. Cheers.

Anthropic Reasoning and Perspective-Based Arguments

It may seem very natural to say "I" am a randomly chosen observer (from some proposed reference class). But keep in mind that is an assumption. PBA suggests that assumption is wrong. And if we reason from one consistent perspective such kind of assumptions are unnecessary.

Anthropic Reasoning and Perspective-Based Arguments

The answer is simple yet unsatisfying. In those situations, assuming the objective is simple self-interest, there is no rational choice to be made.

If we assume the objective is the combined interest of a proposed reference class, and we further assume every single agent in the reference class follows the same decision theory, then there would be a rational choice. However, that does not correspond to the self-locating probability. It corresponds to a probability that can be consistently formulated from the god's eye view. E.g. the probability that a randomly chosen observer is simulated rather than the probability that "I" am simulated. Those two are distinctly different unless we mix the perspectives and accept some kind of anthropic assumption such as SSA or SIA.

Anthropic Reasoning and Perspective-Based Arguments

First of all, I am pessimistic about finding evidence of the multiverse. That being said, If we take the multiverse as given the WAP is still not the complete picture. Because there are two separate questions here. And the WAP answers only one of them. Let me show this with an example.

Say the subject is my parents' marriage. There are two ways to think about it. One way is I take my first-person view and ask a perspective dependent question "why (do I find) they married each other?" Here a WAP type answer is all that's needed. Because if they didn't I wouldn't exist. However, if the question is formulated impartially/objectively (e.g. from a god's eye view): "why did they marry each other?". Then it calls for an impartial answer, maybe a causal model. The WAP doesn't apply here. The key is to keep reasoning from different perspectives separate. Back to the fundamental parameters. The WAP explains why we find the parameters compatible with our existence. Yet that is not the scientific (impartial) explanation for their values. (If the multiverse is confirmed then the scientific answer could be it's just random). If we do not recognize the importance of perspective to reasoning, we would mix the above two questions and treat it as one problem. By doing so teleological conclusions can always be made. Instead of the fine-tuned universe, they would just argue for a fine-tuned multiverse. Which has already been done by intelligent design proponents IIRC.

Load More