What are the strongest arguments for atheism?

[harvey1]

You know, come to think of it. I haven't seen any arguments that support the atheist claim that God doesn't exist. Why is that? So, let's turn the tables for a second, and ask, what are the strongest arguments in support of atheism?

Btw, don't bother answering if you either don't have an argument or don't feel that you are required to support your philosophical position.

[spetey]

Again, hard questions, and ones that I don't think need to be solved fully in order to form reasonable beliefs. Nonetheless I do think it's important to be able to give reasons even for your views on what reasons are, when they are questioned or disagreed upon. My best guess is minimum message length as a good starting point for what it is for one hypothesis to make the evidence more "plausible" than another (roughly, P(E|H1) > P(E|H2)), and for what it is to have one theory to be simpler than another (roughly, -log(P(H1)) < -log(P(H2)), and for how these two trade off each other (minimize -log(P(H)) + -log(P(E|H)).

If you have standards that differ, and that show my reasons for believing that there is no God are poor, I'd be interested to hear them. Until then I will hold my beliefs according to the reasons I now see.

And incidentally what I consider "reasonable to believe" is the same for "a metaphysical reality that is not itself given to physical examination" as for all other cases of evaluating reason to believe. So I was answering your question. I evaluate them all on roughly IBE grounds.


spetey

[spetey]

Also, it seems that we differ again in our concepts of 'God'. I tend to define God as the underlying, unifying principle to the workings of reality and of the universe. Do reality and the universe have some kind of basic underlying unity? The answer is yes (this is the extent of my belief, and I don't think it epistemically irresponsible to maintain it; the rest is just my attempt at metaphysical conjecture), but then the question becomes, what is it? What is the nature of it? This, I think, is ground for a more useful debate.

Ugh. This is as bad as defining 'God' as "all matter in the universe". If God is just the basic physical laws, then of course I believe in a God so defined. But notice this "God" doesn't have any beliefs or desires, can't absolve anyone of sins, can't bring about the universe itself...

Also, "evil" in the context of PoE does not just mean "moral evil" (evil done by people). It means badness of any sort, like the horrible death many children experienced in the recent tsunami. Why did God do that? Did the children deserve those horrible deaths? That is one of the hard questions of the PoE. If you want to talk more about PoE in particular, please do so on another thread.


spetey

[harvey1]

Question 4: All you've done by suggesting "some tradoff of making the available explananda more plausible than other available explanations" is refer to yet another set of criteria that defines 'reasonable' but you haven't told me what it is. What criteria makes an explananda more plausible? Does that criteria provide a means by which to define "reasonable to believe"...itself not given to physical examination? 'Being simpler' also begs the question since you haven't told me what a simpler criteria is and how that criteria provide a means by which to define "reasonable to believe"etc...

My best guess is... minimum message length as a good starting point for what it is for one hypothesis to make the evidence more "plausible" than another

Question 5: Okay, I don't want to get too picky, but this approach looks like it has zero possibility of being used in science, much less philosophy. But, it occurred to me that this approach presents some hope to getting a criteria of a reasonable belief from you. If this view represents a reasonable belief which you wouldn't deny, can you tell me what attributes there exist about MML that make it reasonable to you and that if you name those attributes without using MML, then why can't you just skip MML and use those attributes for all reasonable beliefs?

If you have standards that differ, and that show my reasons for believing that there is no God are poor, I'd be interested to hear them. Until then I will hold my beliefs according to the reasons I now see.

I'm an epistemic pragmatist don't forget! I base my views of reasonableness based largely on how I interpret their success of implementation... The "MML approach" seems to be a foundationalist approach in sheep's clothing (it doesn't possess successful implementation as far as I'm concerned so it wouldn't pass pragmatic muster in my viewpoint)...

[spetey]

If this view represents a reasonable belief which you wouldn't deny, can you tell me what attributes there exist about MML that make it reasonable to you and that if you name those attributes without using MML, then why can't you just skip MML and use those attributes for all reasonable beliefs?

Harvey, I did! I initially said I prefer hypotheses that lend credence to the data and that are simpler. You asked for details about these criteria, and then I gave a potential formalization of these notions for the details. You're right, real scientists don't (typically) code up their info and check the log-likelihoods. Most probably aren't aware of MML. They just make estimations of those results based on the informal guidelines of data-fit and simplicity. But you asked for the details beyond this. I took this as asking: "when there's disagreement about how well the data fit the theory, and how simple a theory is, how can that be settled?" This is a very hard question, but I think MML is a great first step at a natural answer. If two scientists agreed on data-fit and simplicity as criteria, but disagreed on which theory fared better on those grounds, I would recommend to them they try the MML measure.

I'm an epistemic pragmatist don't forget! I base my views of reasonableness based largely on how I interpret their success of implementation... The "MML approach" seems to be a foundationalist approach in sheep's clothing (it doesn't possess successful implementation as far as I'm concerned so it wouldn't pass pragmatic muster in my viewpoint)...

(For anyone else reading: this leans on an odd dichotomy Harvey pushes between what he calls "pragmatism" and what he calls "foundationalism". I still can't make out what he means by these terms; at any rate it's little like how the philosophers I've read use them. But if you want to try to make sense of it, it's discussed here.)


spetey

[harvey1]

If this view represents a reasonable belief which you wouldn't deny, can you tell me what attributes there exist about MML that make it reasonable to you and that if you name those attributes without using MML, then why can't you just skip MML and use those attributes for all reasonable beliefs?

I did! I initially said I prefer hypotheses that lend credence to the data and that are simpler. You asked for details about these criteria, and then I gave a potential formalization of these notions for the details. You're right, real scientists don't (typically) code up their info and check the log-likelihoods. Most probably aren't aware of MML. They just make estimations of those results based on the informal guidelines of data-fit and simplicity. But you asked for the details beyond this. I took this as asking: "when there's disagreement about how well the data fit the theory, and how simple a theory is, how can that be settled?" This is a very hard question, but I think MML is a great first step at a natural answer. If two scientists agreed on data-fit and simplicity as criteria, but disagreed on which theory fared better on those grounds, I would recommend to them they try the MML measure.

That's not quite what I asked. I want to know criteria that isn't itself begging the question. Referring to IBEs, best plausible explanations, simplest theories, etc, is all begging the question on what criteria makes for a reasonable belief since...

1) IBEs refer to 'best explanation' (what criteria makes it the best??)
2) Best plausible explanation refer to plausibility (what criteria makes it the most plausible?)
3) Most simplist theory (what criteria makes it the simplest?)
4) MML measure theory (what criteria do you use to select this as a reasonable theory when in fact it doesn't appear to be reasonable - no one in science/philosophy uses such hypothetical reasoning methods)

I'm not asking for anything too complex Spetey. All I want to know is what criteria do you use that makes beliefs such as MML measure theory preferable?

(For anyone else reading: this leans on an odd dichotomy Harvey pushes between what he calls "pragmatism" and what he calls "foundationalism". I still can't make out what he means by these terms; at any rate it's little like how the philosophers I've read use them. But if you want to try to make sense of it, it's discussed here.)

Don't you think you're too caught up in the 5-10 people who might read half of our posts? More than once you've addressed others rather than me, and it leaves me baffled. Who cares who is reading our rants? I'm content just to have these pleasant discussions and find friendly correspondence on enjoyable discussions. Aren't you?

[spetey]

That's not quite what I asked. I want to know criteria that isn't itself begging the question. Referring to IBEs, best plausible explanations, simplest theories, etc, is all begging the question on what criteria makes for a reasonable belief since...

To beg the question is to argue for a conclusion by using that conclusion as a premise. You'll have to be more explicit about where I'm doing that.

1) IBEs refer to 'best explanation' (what criteria makes it the best??)

Simplicity, fit with data.

2) Best plausible explanation refer to plausibility (what criteria makes it the most plausible?)

Simplicity, fitting with data. If you're saying that I haven't explained what it is to "fit the data", I'd first want to know why you ask--do you disagree that we should try to get theories to fit the data? It's hard to imagine you do, and if not, then why press the issue? But anyway I would try to answer by saying that the relative fit of data to a theory can be measured by how much the theory increases the probability of the data, ie the size of P(E|H) - P(E) (if I understand this Bayesian stuff right, it's still somewhat new to me).

3) Most simplist theory (what criteria makes it the simplest?)

Informally: paucity of concepts or types of entities posited. More formally, as I say, in the size of -log(P(H)), the minimal length it would take to encode the theory. Again, do you not think simplicity is a good criterion for theories? It's hard to imagine you don't. (Do you prefer the theory that posits more entities, like hidden IPUs and such?) If you do think simplicity is a good standard, why do you challenge me on it? We have found a premise we share.

4) MML measure theory (what criteria do you use to select this as a reasonable theory when in fact it doesn't appear to be reasonable - no one in science/philosophy uses such hypothetical reasoning methods)

As I say, I like MML since it seems like a good formalization of a way to assess the simplest and best-fitting theory for given data. What is inadequate in any of these responses?

I'm not asking for anything too complex Spetey. All I want to know is what criteria do you use that makes beliefs such as MML measure theory preferable?

I think I've explained that for the third time now. Where are you going with this?

Don't you think you're too caught up in the 5-10 people who might read half of our posts? More than once you've addressed others rather than me, and it leaves me baffled. Who cares who is reading our rants? I'm content just to have these pleasant discussions and find friendly correspondence on enjoyable discussions. Aren't you?

I don't like to exclude people from our conversations, and to join this particular topic would require knowledge specific to our discussions. I am content to just discuss this pleasantly with you, but I would welcome other opinions (though I wasn't so nice to MagusYanam I admit). And it's hard to say how many "lurkers" there are. Our posts get more views than seem to be explained by you and me, Harvey.


spetey

[bernee51]

Question 1: Do you believe it is unreasonable to be an atheist if you have strong reason to believe that a God (of any of the definitions which you refused) is one of a few reasonable explanations to the universe existing as we observe it to exist?

Can I play too? What if you have no strong reason to believe that a god is one of the few reasonable explanations to the universe existing as we observe it to exist? But I see that is not what you are asking. If I had a strong reason to believe such a thing I don't think I could honestly call myself an atheist. No my answer to your question - as you ask it would be 'yes'. If you are allowing for any possible definition of god I would have to call myself agnostic.

Well, that was a quick game (that is, I'm assuming you accepted the possible definitions of God that Spetey *accepted*)!

Silly me...I said any possible definitions of god...I was not specifically referring to Spetey's acceptable or otherwise defintions. I shall read back and review.

I would be loathe to grant you so easy a 'conversion'

....

Some time later...having re-read your question and subsequent posts of both your self and Spetey I must recant. Sorry - so far I am still an atheist - where I define athesut as one you does not have a god-belief.

Now if you want to talk about a specific definiton of god....

BTW Harvey do you have a belief in Zeus as a deity?

[harvey1]

beg the question is to argue for a conclusion by using that conclusion as a premise. You'll have to be more explicit about where I'm doing that.

In your argument for what criteria exists for something to be the most reasonable you assumed the definition of "best explanation", "most plausible explanation", "most simplest explanation", etc. I want the criteria for the assumption. I know that once we lock in on our assumptions of reasonableness that it is straightforward to show which explanation is reasonable. I want to know what to do before that.

Let me give you an example. Spet1 says that a reasonable theory is one that is the 'most weird' and that Spet1 is able to deduce which hypothesis is the most reasonable and likely to be true by appealing to the Inference of the Most Weird Explanation. If you ask Spet1 what is the criteria he uses to establish the 'most weird' (and clarify what is weird from what is not weird), he comes back with MWL theory that shows that the shortest statement is most likely to be the most weird and therefore true. If you ask Spet1 more about 'most weird' he just asks you why you wouldn't accept it...

Do you see the problem in Spet1's reasoning?

Simplicity, fitting with data. If you're saying that I haven't explained what it is to "fit the data", I'd first want to know why you ask--do you disagree that we should try to get theories to fit the data? It's hard to imagine you do, and if not, then why press the issue? But anyway I would try to answer by saying that the relative fit of data to a theory can be measured by how much the theory increases the probability of the data, ie the size of P(E|H) - P(E) (if I understand this Bayesian stuff right, it's still somewhat new to me).

That's just pushing things up a notch. How did you come to accept Bayesian probability as reasonable? What criteria did you use? Just saying 'fit the data' is not exactly what I'm looking for. Of course good theories fit the data, but every model can fit the data if you amend your model to do so. It means nothing in terms of deciding if a theory/belief is reasonable unless the criteria of reasonableness is more explicit that that.

Again, do you not think simplicity is a good criterion for theories? It's hard to imagine you don't. (Do you prefer the theory that posits more entities, like hidden IPUs and such?) If you do think simplicity is a good standard, why do you challenge me on it? We have found a premise we share.

Of course I think simplicity is a good criterion for theories, but the question is why do you accept it as a reasonable criterion for a model? What criteria do you use to accept simplicity, plausibility, etc in order to establish these concepts as reasonable? I want to know how you determine that something is reasonable.

As I say, I like MML since it seems like a good formalization of a way to assess the simplest and best-fitting theory for given data. What is inadequate in any of these responses?

That's fine, but it is not the basis of your concept of reasonableness otherwise you would need the MML to have once decided that the MML was reasonable!

I think I've explained that for the third time now. Where are you going with this?

It's the best place to start when someone says something isn't reasonable, I want to know what criteria they use to establish something as reasonable. If the criteria they offer is something they couldn't have possibly known prior to deciding on the issue, then obviously there must be a more primitive criteria that I think plays the largest role in what they consider to be reasonable. I want to know what that is since it is the criteria that I need to satisfy in order to show that God is a reasonable belief.

[harvey1]

Silly me...I said any possible definitions of god...I was not specifically referring to Spetey's acceptable or otherwise defintions. I shall read back and review. I would be loathe to grant you so easy a 'conversion' Some time later...having re-read your question and subsequent posts of both your self and Spetey I must recant. Sorry - so far I am still an atheist - where I define athesut as one you does not have a god-belief. Now if you want to talk about a specific definiton of god....

Bernee, I'm confused, I thought you already acknowledged that you were an agnostic earlier in this thread?

BTW Harvey do you have a belief in Zeus as a deity?

No. I don't.

[spetey]

beg the question is to argue for a conclusion by using that conclusion as a premise. You'll have to be more explicit about where I'm doing that.

In your argument for what criteria exists for something to be the most reasonable you assumed the definition of "best explanation", "most plausible explanation", "most simplest explanation", etc.

This is not begging the question. If you ask why I think Socrates is mortal, and I say I think so because Socrates is human and all humans are mortal, I have not "begged the question" by not having defined 'mortal' or 'human', or by not arguing for those premises immediately in turn. I might suppose you agree to the premises, and know what it is to be human and mortal. Similarly I did not beg the question by not immediately defining 'simplicity' or 'data-fit', or by not further justifying those premises. I was supposing you knew roughly what simplicity and data-fit are, and why they might be relevant to a better explanation. Of course you can challenge those premises, that better explanations fit the data and are simpler. Do you?

Let me give you an example. Spet1 says that a reasonable theory is one that is the 'most weird' and that Spet1 is able to deduce which hypothesis is the most reasonable and likely to be true by appealing to the Inference of the Most Weird Explanation. If you ask Spet1 what is the criteria he uses to establish the 'most weird' (and clarify what is weird from what is not weird), he comes back with MWL theory that shows that the shortest statement is most likely to be the most weird and therefore true. If you ask Spet1 more about 'most weird' he just asks you why you wouldn't accept it...

Do you see the problem in Spet1's reasoning?

Yes, he is appealing to implausible premises. Why would weirdness make for a better explanation? Why are shorter messages weirder? These would certainly be premises worthy of challenge.

But I think it's much more plausible to say that simple and data-fitting explanations are good. No? Do you deny either? If so, I will argue for them. If not, let's move on.

Just saying 'fit the data' is not exactly what I'm looking for. Of course good theories fit the data, but every model can fit the data if you amend your model to do so.

Good. First of all, you seem to agree that fitting data is a good mark of a reasonable theory. Now of course you can amend any theory to fit the data. You seem to agree that this would be a bad thing to do--to keep adding ad hoc bits to your theory until it fits all data perfectly. Is this perhaps because you also value simplicity in a theory, and think that they should trade off...?

Of course I think simplicity is a good criterion for theories, but the question is why do you accept it as a reasonable criterion for a model?

This is hard to say, though I have my own guesses. Why do you think so? Look, I don't have to justify all premises in order to get an argument off the ground. They only have to bottom out in premises that others will accept. It seems we have found some: a better explanation will (perhaps among other things, on your view) fit the data well and be relatively simple.

What criteria do you use to accept simplicity, plausibility, etc in order to establish these concepts as reasonable? I want to know how you determine that something is reasonable.

A belief is more reasonable if it has a better explanation, and an explanation is better if it is simple and fits the data. And it's hard to say why that is, the same way it's hard to justify a Peano axiom or something. But since you agree I don't see how it's at issue. If you genuinely disagree then I'd like to hear why, and we can discuss that.

Bernee, I'm confused, I thought you already acknowledged that you were an agnostic earlier in this thread?

He said he recants. Though actually, Bernee, from your own words it sounds to me like you're still agnostic, since you say you "do not have a god-belief". It sounds like you mean you do not have a strong belief with respect to the existence of God, which is agnosticism. An atheist has a god-belief: she believes that there is no God.


spetey