posted
So atheism does require more faith than theism.

I might check this book out, but I doubt it will be very satisfying. Results from Bayes analyses can change wildly based on minute changes to the starting probabilities. Given his starting point, he would have had to be almost impossibly rigorous to avoid tilting the starting probabilities minutely in his favor.

It'd be interesting if someone made a spreadsheet that allowed tweaking his starting probabilities to see the different results.

posted
While I haven't read it myself, according to someone (who is educated in statistics) who has, the person basically assumes that there's an base 50% probability for God and allows it to be influenced either way by essentially arbitrary suppositions about what the world would be like without God (that is, he assumes that natural disasters would cause more harm, that sort of thing).

IOW, he assumes God to prove God.
Posts: 15770 | Registered: Dec 2001
| IP: Logged |

posted
You mean he assumes 50% probability of God - which is actually neither favoring God's existence of nonexistience.
Posts: 2432 | Registered: Feb 2001
| IP: Logged |

posted
What a ridiculous "study." *laugh* It's up there with that beer bubble bit, only less scientific.
Posts: 37449 | Registered: May 1999
| IP: Logged |

posted
First, that statement was predicated on the sort of suppositions he makes, not just the 50% part, which certainly did involve assuming God.

Second, assuming a 50% possibility for something is certainly favoring his existence or nonexistence. Any assumed probability for something is complete bias, and while assumptive bias can be useful in determining a starting place for investigation, its presence in results, as the 50% consitutes part of in this case, is farcical.

To make this clearer, assuming there's a 50% probability you will die in the next five seconds is highly biased, and if I just start adding and subtracting probabilities based on my perception of your risk I'm going to end up with a number that's far too high because my starting point was an arbitrarily high, ridiculous probability. It makes as much sense as assuming there's a 95% chance you will die; not much at all.
Posts: 15770 | Registered: Dec 2001
| IP: Logged |

posted
I think your example is flawed Fugu. To say that you have a 50% chance of dying in the next 5 seconds in abscence of all information is actually about the best you can do. Then things like life length trends, your current health, what you are doing now and will do in the next 5 seconds then gets added and and (hopefully ) make the probability of you dying in those 5 seconds neglagible.

Not that I'm saying anything about this book, but I think a 50% chance of something in abcence of all knolwedge is a good assumption in general.

posted
I think I disagree strongly with his weightings. For example, starting with a 50% possibility of unicorns....
Posts: 37449 | Registered: May 1999
| IP: Logged |

posted
No, its not the best you can do. The best you can do is not making the assumption. Making an assumption as a basis for a number is a logical error.

To put it simply, an unknown probability does not equal a 50% probability.

Assuming a 50% probability in all cases of unknown probability will lead to rather extreme biases quite frequently. Lets try it:

I hereby assume there is a 50% probability I will spontaneously combust in the next five minutes.

Well, since not many people have spontaneously combusted, I'll remove, say, 98% of the probability. Okay, so I have a 1% chance of spontaneously combusting.

Tres is going to jump on and cry foul at the probability I chose to modify it by, quite rightly. The key thing is, modifying an assumed probability by other probabilities with their own guesswork built in only magnifies both the original and their own assumptions.

The proper way to procede is not to assume any probability at all. Instead, examine several things: the actual frequency of spontaneous combustion in a sample population, the likely causes of spontaneous combustion, such things as those, and consider the probability based on the sample, and based on the likelihood of the causes arising, et cetera. There is no place for an assumed base probability in science.

For one thing, then you get assumed base probabilities in complex situations for mutually exclusive things which exceed 100%. That should make it even more obvious how ever assuming a base probability is silly.
Posts: 15770 | Registered: Dec 2001
| IP: Logged |

posted
How about the probability of more than one god?

Supposing that there is already one god who created and organized OUR observable universe, and that god started at a point in space and moved outward from that point, organizing matter along an ever-expanding, spherical front, what are the odds that another god made a similar start elsewhere in infinite space and set a similar, expanding organizational sphere in motion? Will the spheres ever collide? What are the chances that both gods subscribe to the same set of values?
Posts: 2655 | Registered: Feb 2004
| IP: Logged |

posted
Something about which nothing is known, you can only say "it either does or does not exist, no evidence leads me to believe one above the other" thus, 50%. If you start with 0 knowledge than 50% chance is accurate. The trick is modifying your new probability right. The way you do it you'd be copmletely correct in saying doesn't work, it magnifys your choice of 50/50.

98% of 50 goes to 1%, but if you assumed 100% chance orginal it would go to 2% and the numbers then show the bias of your choice. However, that isn't the only way of doing it. The way I see it is this:

Begin with each side having an equal opportunity being true in a lack of all information: 1 to 1 (true is first false is second). Then weight the probability by the amount of information supporting it. Of course then the question of how do you weight the importance of information comes up but that's why I would never write sucj a book. However, in a complete lack of information we have

lim x->0 x*1 to x*1

divide these two to get the ratio and we have

(x*1)/(x*1)

and thus a 1 to 1 chance. As soon as any information enters we then have infinite on side and zero on the other.

lim x->0 [(x+C)*1] to (x*1)

where C is the weight of the information that has been gained (assumed positive, non-zero value) all of a sudden there is a 100% chance and the initial assumption of 1 to 1 dissapears.

quote:Having tackled what he means by God, Unwin moves onto to explicate what he means by probability. The simply answer is Bayes Theorem. He distinguishes Bayesian approach to probability from a frequentist one. The frequentist approach is best described by the coin flip. From this perspective we expect the coin to land heads 50% of the time so its probability is 50%. Roll a six-sided die and the probability of any side coming up is 16.7 or one-sixth. In contrast, Unwin asserts that in "the Bayesian world, a probability is an expression of a degree of belief." Using his background in quantum physics, he uses the following example:

When the statement is made by the physicist that there is a 5 percent probability neutrinos are massless, she means that on a scale of 0 to 100, her degree of belief that the said particles are massless is 5.

Unwin does a good job of laying out the Bayesian approach and any attempt on my part to replicate that explanation would be redundant and inelegant. The basic idea, however, is to take the evidence one has and to convert the "data" into a mathematical formula and thus a probability. Like the scientist with the neutrinos, Unwin attempts to arrive at a numerical probability. Instead of sub-atomic particles, the subject is God's existence.

Unwin starts with the proposition that he is neutral on God's existence. In other words the starting point is 50/50 on God's existence. This is the a priori probability from a neutral perspective but he goes further.

This isn't enough to totally satisfy fugu's objection, but it puts it in a much less damaging light I think. I'll have to reserve further judgement until I read the book.

posted
I can't believe you all are complaining about the initial probability, when the much more blatant problem is that he's just making up a bunch of modifications to the initial number out of thin air. I mean, how could anyone say a given observation X makes God P more likely. Where could those numbers (P) possibly come from?

Bayesian analysis, as I understand it, requires some way to know how a new observation effects the previous theory, probability-wise. But does the existence of the Bible make God 10% more likely? 30%? 1%? You can't know that.

posted
For example, how much of a weighting to the potential existence of unicorns is it that almost no one claims to have seen a unicorn? Or that no unicorn bones have ever been found?

When it comes to God, how much of a penalty is it that many people claiming to have spoken to God think they've spoken to a DIFFERENT God?

These entirely arbitrary numbers are still pretty darn relevant, even if -- as I suspect -- he's using physics for most of his assumptions.
Posts: 37449 | Registered: May 1999
| IP: Logged |

posted
I agree (no, not with you Richard ), which is why I didn't even bother arguing, besides which, I don't think anyone here has read the book, so there's really no way to know how he weighted things, and thus no way to really say anything about it.

posted
Something tells me Unwin and I assign different prior probabilities to the proposition that God exists.

You can't just say your priors for a proposition are 50-50 if you're neutral on whether it's true. For example, say you're neutral about the color of a flag before you look at it. I ask you, "What's your prior probability for the flag being red?"

You say 50%.

I say, "What's your prior probability for the flag being green?"

You say 50%.

I say, "What's your prior probability for the flag being blue?"

You say 50%.

These are all mutually exclusive propositions, so your priors now add up to 150%. Oops!

So I would have to weigh my prior credence about God's existence against that of all the other propositions that could explain the existence of the universe: brane collision, quantum tunnelling from nothing, spontaneous creation from a black hole in another universe, etc. Seems to me like my prior credence for God's existence would end up much lower than 50%.
Posts: 4600 | Registered: Mar 2000
| IP: Logged |

posted
But if you define it right, there are only two possibilities. I'm not saying there are only two possiblities for God (I know, He exists, He doesn't, multiple Gods exist, so on), but that in a strictly boolean question 50/50 is good.

I'm really not trying to defend the book, I haven't read it so I don't actually know how well constructed it is, but from what I've heard I'd have to say it's rather absurd. I'm just commenting on the 50/50 assumption in general.

quote:So I would have to weigh my prior credence about God's existence against that of all the other propositions that could explain the existence of the universe: brane collision, quantum tunnelling from nothing, spontaneous creation from a black hole in another universe, etc. Seems to me like my prior credence for God's existence would end up much lower than 50%.

To major objections to this statement. First, brane collision, quantum tunnelling, and the rest are not exclusive of God's existence and creation of the Universe. Second, if you're going to throw in alternative physical explanations, you'd have to throw in alternative theistic explanations, too.

quote:But if you define it right, there are only two possibilities. I'm not saying there are only two possiblities for God (I know, He exists, He doesn't, multiple Gods exist, so on), but that in a strictly boolean question 50/50 is good.

Same goes for colors. The flag is either red or non-red. But it'd be ridiculous for you to assign a prior prob of 50% to the flag's being red when there are so many other actual possibilities.

EDIT: You're right, Dag. My point is that the priors concerned are so horrendously underdetermined by our poor understanding of the question that I don't have any idea what to use for them, and so applying Bayes's theorem to this question is stupid.

posted
As I said, I have not read it myself, however, the person I know who has read it is statistically well versed.

For one thing, the neutral position in science is not 50%, its that you don't know what the probability is.

Also Hobbes, you just made a circular mathematical statement:

You write

quote:lim x->0 x*1 to x*1

Which assumes the ratio is 1:1 (that is what a 1 factor means, after all).

Then you write:

quote:divide these two to get the ratio and we have

(x*1)/(x*1)

and thus a 1 to 1 chance.

Assuming the result isn't logic.

The proper assumption in science is that the probability of an positive unknown is x, where x is unknown and falls in [0, 1) (some would assert (0, 1), but scientists are generally willing to entertain impossibilities, so lets include 0) (the possibility of negative unknown is, of course, in the range (0, 1], if we accept that the range for a positive unknown's probability is [0, 1), and (0,1) if that is also the range for a positive unknown's probability).
Posts: 15770 | Registered: Dec 2001
| IP: Logged |

posted
Actually Fugu, it wasn't circular logic, because what I was showing was the way in which an assumption of a 1 to 1 relationship could not influence future probability. As soon as any information enters the equation the initial assumption basically becomes meaningless as it simply multiplies that information by 1.

Dest, the problem there is you're starting with information. Either it is red or it isn't. Then the revelation that there are actually many colors becomes a new piece of information that affects the probability.

posted
I agree with that last part Fugu, a probability x lies within the line (0,1). As information is added the probability is moved from on place to another. If you want to determine if you will die in 5 seconds and you are given the piece of information that .01% of the population will die in the next 5 seconds you would then set x = .0001 When you add more information you continually move x around accordingly. However, before that information all you know is that x is somewhere in (0,1). It is perfectly appropriate to say I simply do not know where it is, no information is given to me. However, most people don't like saying that, and it is also possible to say that the average of all possible positions (each one just as likely as the next) between (0,1) is .5 Or 50%.

posted
You know, my understanding of this Bayesian method was that regardless of the starting point, we'd end up at the correct probability after enough trials. 50% is just the quickest starting point.
Posts: 2432 | Registered: Feb 2001
| IP: Logged |

posted
As the number of trials approaches infinity, that's true, the dependence on priors goes away. In Unwin's case, it seems as though only a few "trials" have been used (however many pieces of "evidence" he considers), and so prior probabilities will be very important to the end result.

quote:Then the revelation that there are actually many colors becomes a new piece of information that affects the probability.

Yeah, same with the revelation that there are more than two possible ways for a universe to be created. I doubt that Unwin incorporated that piece of info into his study.

posted
Destineer, you haven't made a credible case for there being "multiple ways to make a universe" in this scenario.

The reason the flag analogy comes to the conclusion it does is because the colors are exclusive. However, none of the "world-creation" scenarios you described are exclusive of an atheistic universe. So the analogy is utterly unrelated.

posted
Dag, the examples I gave are (underdeveloped/speculative) ideas about how the universe could have arisen from nothing, i.e. without an intelligent creator. I think they are mutually exclusive with a universe created by a god of the traditional theistic sort.

This is especially true of the quantum tunnelling case (Alexander Vilenkin has a paper about it, can't remember the full reference but it's not available online anyway, I had to copy it in a library). Basically, the idea is that you start out with literally nothing -- a 0-dimensional spacetime -- and there is a nonzero probability that this will evolve into a de Sitter spacetime resembling the early universe.

quote:The reason the flag analogy comes to the conclusion it does is because the colors are exclusive. However, none of the "world-creation" scenarios you described are exclusive of an atheistic universe.

None of the non-red colors are exlusive of a non-red flag, either. I can say of the flag, "It's either red or a non-red color," and of the universe, "It was created by God, or by some non-God process." Why should the colors be given equal priors, but not the creation processes?

The creation processes are mutually exclusive of each other. They all fit under the "non-God" heading, just as the other colors fit under the "non-red" heading. Individuating all possible creation stories into God and non-God, and saying that somehow we should assign each of these categories equal proper credence, seems completely unmotivated. Not that I have a better solution. I just think we're ignorant enough on this issue that we shouldn't make silly attempts to pretend we understand it yet.
Posts: 4600 | Registered: Mar 2000
| IP: Logged |