Saturday, September 05, 2009

Book Review: The Existence of God by Richard Swinburne

The Existence of God (2nd Edition) by Richard Swinburne is an impressive and significant philosophical work on the existence of God. Swinburne’s premise is to weigh the evidence for and against theism and determine which way the balance falls. This review will only summarize his approach and some of the ideas developed in the book.

Swinburne introduces his approach. He is not trying to come to an indubitable conclusion; his goal is a probabilistic one based upon confirmation theory. So, of the fourteen chapters, the first four describe inductive arguments, the nature of explanation, justification of explanation, and complete explanation. The author first aims to show the need to take the arguments for God’s existence together:

One unfortunate feature of recent philosophy of religion has been a tendency to treat arguments for the existence of God in isolation from each other…But clearly the arguments may back each other up or alternatively weaken each other, and we need to consider whether or not they do.1
Swinburne rejects a piecemeal criticism of arguments for God:
Sometimes, however, philosophers consider the arguments for the existence of God in isolation from each other, reasoning as follows: the cosmological argument does not prove the conclusion, the teleological argument does not prove the conclusion, etc. etc., therefore the arguments do not prove the conclusion. But this ‘divide and rule’ technique with the arguments is inadmissible.2
Again, the author emphasizes the need to assess the weight of all the evidence:
The crucial issue, however, is whether all the arguments taken together make it probable that God exists, whether the balance of all the relevant evidence favours the claim of theism or not…we ought to act on a hypothesis in so far as it is rendered probable by the total evidence available to us – all we know about the world, not just some limited piece of knowledge.3
Swinburne presents an analogy of a scientific theory to explain the need to assess the arguments together: “Thus each of the various pieces of evidence that are cited as evidence in favour of the General Theory of Relativity do not by themselves make it very probable, but together they do give it quite a degree of probability.”4 He says that the probability of a theory being true rests partially in its explanatory power: “…for a theory to have great explanatory power, the phenomena that it predicts must be such that but for it they would not otherwise be expected.”5

As Swinburne sets the stage for his cumulative assessment, he also shows that an explanation does not itself need to be explained before it can be a satisfactory explanation. He thus puts the “who designed the designer?” objection in its place:
Now a full explanation really does by itself explain why something happened. It does so quite independently of whether or not there is an explanation of how any states it cites came to be…or why any reasons that it cites operate… To suppose otherwise is to commit a fallacy that we may call ‘the completist fallacy’. Clearly it is a fallacy. For if it were really the case that F could not explain E unless there is an explanation of F, nothing in the universe could be explained, unless there were explanations of such things as the origin of our galaxy – which is absurd.6
After his preliminary exposition of his inductive approach (which includes a treatment of Bayes’ theorem) Swinburne begins with what he calls the inherent probability of theism. This hinges on the idea that theism is, in itself, a very simple hypothesis: “Theism postulates God as a person with intentions, beliefs, and basic powers, but ones of a very simple kind, so simple that it postulates the simplest kind of person that there could be.”7 He develops this idea in much depth, as simplicity is the major determinant of intrinsic probability: “So in postulating a person with infinite power the theist is postulating a person with the simplest kind of power possible.”8 Swinburne explains that unlimited power is simpler to posit than power that is limited to some finite degree:
To postulate a G of very great but finite power, much but not all knowledge, etc., would raise the inevitable questions of why he has just that amount of power and knowledge, and what stops him from having more, questions that do not arise with the postulation of God.9
Before moving on to the classical arguments for theism, Swinburne reminds the reader that his assessment is to show the greater explanatory power of theism: “how much more likely does the existence of God make the occurrence of those phenomena than it would be if we do not assume the existence of God.”10 It is at this point that the book shifts into exploring one or two major arguments for the existence of God per chapter. These include: the cosmological argument, teleological arguments, arguments from consciousness and morality, the argument from providence, the problem of evil (which counts against the theistic hypothesis), arguments from history and miracles, and the argument from religious experience.

In dealing with the cosmological arguments, the author shows that either the universe is the “stopping point” or God is the “stopping point.” As for design arguments, Swinburne treats them “not as an argument from analogy (the way typical of the eighteenth century) but in the way in which other arguments are treated in this book, as an argument from evidence that it would be probable would occur if theism is true, but not otherwise.”11 Noted here is the fact that Swinburne shows the greater probability of theism given evidence for design even though Swinburne seems to concede macroevolutionary theory: “So our question becomes – why are there not just any laws of nature, but laws of a particular kind such that together with the initial matter-energy at the time of the ‘Big Bang’ would lead to the evolution of human bodies.”12

Swinburne includes an evaluation of evidence from the fine-tuning of the universe for life, which entails 15 or so pages. He criticizes multiverse theory:
…it is the height of irrationality to postulate an infinite number of universes never causally connected with each other, merely to avoid the hypothesis of theism. Given that simplicity makes for prior probability, and a theory is simpler the fewer entities it postulates, it is far simpler to postulate one God than an infinite number of universes, each differing from each other in accord with a regular formula, uncaused by anything else.13
Topping of his design argument, the author also includes the argument from beauty. His two chapters dealing with consciousness, morality, and providence follow. Within these chapters the reader will find amazing philosophical work. At this point Swinburne again concludes that each one of these arguments are more probable if theism is true, while emphasizing their overall cumulative effect: “as we consider more and more phenomena, the probability that they will occur if there is no God gets less and less.”14

Swinburne’s assessment is not all positive. Chapter eleven’s presentation of the issue of evil is sober and honest. Common defenses against evil are presented, but the author also shows where they may fall short. Chapter twelve then explores the argument from history and miracles: “if there is a God, there exists a being with power to set aside the laws of nature that he normally sustains”15 Swinburne presents four types of evidence that can be furnished to support miracles. Chapter thirteen is in a similar vein, dealing with the argument from religious experience. Here the author explores five kinds of religious experience and determines how one can determine the veracity of such claims.

The book ends with a final chapter weighing the probabilities. Swinburne’s conclusion: “On our total evidence theism is more probable than not.”16 Whether one agrees or disagrees with Swinburne’s conclusions or methods, this is a praiseworthy book from a strictly philosophical point of view. Nevertheless, The Existence of God will prove to be both a challenging and rewarding book for the reader.

32 comments :

Anonymous said...

This is in fact a very influential book. The only thing I don't like about it is that it's a little dated, but if I recall correctly there recently was a new edition that came out.

Jonathan West said...

The problem with most reviews I have read of "The Existence of God" (including this one) is hat they appear to have been written by people who do not know enough of science or mathematics to judge the quality of Swinburne's methods and arguments.

Swinburne unfortunately is using lots of long words and equations to give the impression that he is being logical, rigorous and scientific, whereas he is in fact making it up as he goes along, and making use of statistical mathematics in ways which are entirely inappropriate.

If you use statistical tests in a circumstance in which the mathematics are invalid (and Swinburne does this many times) then you results are entirely worthless, though anybody unfamiliar with statistical theory will be unaware of the fact.

Swinburne makes constant use of Bayes' Theorem. Bayes' theorem is only valid when dealing with situations which can result from a variety of known causes of known probabilities of events which which can occur in known combinations, so that conditional probabilities can be calculated. Swinburne tries to apply it to a situations of unknown causes and unknown linkages between events. Moreover he is dealing with matters for which no statistical data is available at all.

If you wish to read a rather more detailed chapter-by-chapter review of "The Existence of God", highlighting the issues described above and the precise issues with Swinburne's methods, come over to my blog. I have so far reviewed the first five chapters, and will be reviewing the rest over the next month or so.

http://scepticalthoughts.blogspot.com/

Haecceitas said...

Jonathan West,

I'd be very surprised to find out that the editor of Bayes's Theorem (Proceedings of the British Academy) from Oxford University Press just makes stuff up as he goes along and uses statistical mathematics in ways which are entirely inappropriate.

I suppose you could tone those comments down a little bit and make them more believable (like, "controversial" instead of "entirely inappropriate"). The range of legitimate applications of Bayes' Theorem is a matter that's subject to debate between the competing schools of thought.

If you want to be a fair-minded sceptic, you should avoid the very common tactic of choosing a side on a debate that pertains to some technical issue with the primary motivation of avoiding the force of theistic arguments and then proceeding to claim that the chosen side is obviously and uncontroversially right.

Jonathan West said...

Haecceitas
I've described why I think Swinburne's use of Bayes' theorem is inappropriate - briefly in my previous comment and in much more detail in my blog. If you think I am wrong, then you will be extremely welcome to leave a comment on my blog explaining what mistakes I have made in the line of reasoning which led me to that conclusion. If I am wrong, then I would like nothing more than to have my mistake pointed out to me so that I can cease being wrong. However, simply making an appeal to authority is not going to be good enough for the purpose.

If Swinburne had been arguing against the existence of God making the same inappropriate use of statistical theory, I would have criticised him in exactly the same terms.

I could rewrite Swinburne's book just changing a number here or there and conclude that the balance of probability is strongly against God's existence. However such an exercise would be just as invalid as Swinburne's original. Made-up numbers remain made up no matter which conclusion they are chosen to point to.

A badly-formed argument helps nobody who is actually interested in the truth.

Jonathan West said...

Haecceitas
By the way, I was very much struck by the irony of a situation in which you accuse me of "choosing a side on a debate that pertains to some technical issue with the primary motivation of avoiding the force of theistic arguments" in a comment on a blog on apologetics!

Haecceitas said...

Jonathan West,

Thanks for your reply. I hope you don't mind if I ask for clarification on some of your views.

- Let's examine the following statement more closely:
"It is a priori unlikely that the editor of a book that is a collection of specialist essays on subject x and is published by the academic press of a prestigious university, is totally incompetent when (s)he deals with the subject x."
Does this statement sound reasonable to you? Obviously, this is not intended to be the end of them matter, but it is a reasonable starting point, isn't it?

- Since you are an atheist, would it be fair to characterize your position on the existence of God by saying that you think that it is improbable (perhaps very improbable) that God exists?

- Do you accept the concept of epistemic probability? Or are you a strict frequentist?

Haecceitas said...

"By the way, I was very much struck by the irony of a situation in which you accuse me of "choosing a side on a debate that pertains to some technical issue with the primary motivation of avoiding the force of theistic arguments" in a comment on a blog on apologetics!"

I'm sure that the same "sin" is committed by some apologists as well. But two wrongs don't make a right.

Jonathan West said...

Haecceitas
I was commenting more on the fact that apologetics consists in whole of doing precisely what you accused me of - but in reverse. I described this kind of behaviour by theists in some detail on "The conflict between science and religion"
http://scepticalthoughts.blogspot.com/2009/06/conflict-between-science-and-religion.html

As for your detailed question. My answers are as follows:

1. Does this statement sound reasonable to you?
As I said, your statement is an appeal to authority. If you understand in what way I am wrong, with me having provided my line of reasoning directly, then you ought to be able to say so by reference to my specific arguments, and without need for any reference to authority.

2. Since you are an atheist, would it be fair to characterize your position on the existence of God by saying that you think that it is improbable (perhaps very improbable) that God exists?
It would be more accurate to characterise me as a skeptic. If and when decent evidence for God's existence appears, then I will give it all due credit. I read "The Existence of God" in the hope that Swinburne might be an academic capable of marshalling the best of that evidence and that a coherent case might emerge. I was extremely prepared to give him all possible benefit of the doubt. But when I found how thoroughly flawed his methods were, it became clear that no conclusions based on them could have any value at all, no matter what they might ultimately be. At this point I hadn't read to the end of the book and therefore didn't know what his conclusions were.

3. Do you accept the concept of epistemic probability? Or are you a strict frequentist?
I'm well aware of the concept of epistemic probability. The term was coined by the statistician Ian Hacking, and is an entirely subjective or qualitative concept - in essence describing the strength with which you hold a belief (or if you prefer, the degree of confidence you have in the truth of a proposition). As he is a statistician, Hacking of course knows far better than to attempt to confuse epistemic probability with statistical (or frequentist) probability by applying statistical mathematics to epistemic issues as if the subjective probabilities could have objective numbers applied to them. So for instance, when Swinburne in 2002 gave a lecture at Yale in which he calculated the probability of the Resurrection as being 97%, I know he was either mistakenly or deliberately conflating the two concepts.

Haecceitas said...

"I was commenting more on the fact that apologetics consists in whole of doing precisely what you accused me of - but in reverse."

How so? I see no reason why apologetics as a discipline can't be understood as simply employing the best methods that there are in evaluating the arguments. One may be subjectively motivated by a desire to demonstrate the truth of Christianity (just as one can pursue the equivalent of apologetics with the aim of defending Naturalism, Buddhism, or whatever). That's not what I'm objecting to.

IMO, your blog post didn't demonstrate any conflict between science and religion in general. Sure, there may be forms of religion that are in conflict with forms of science (and indeed, forms of science that are in conflict with forms of science), but that doesn't show that science and religion are by their very nature in conflict. You concluded your article by saying that "The crux of the conflict between science and religion is the conflict between evidence and authority." But if the religious authority is grounded with evidence, then there is no conflict.

"As I said, your statement is an appeal to authority. If you understand in what way I am wrong, with me having provided my line of reasoning directly, then you ought to be able to say so by reference to my specific arguments, and without need for any reference to authority."

Given your comments about Swinburne, the issue of authority is perfectly legitimate for me to raise. Unless you claim to have better credentials on Bayesian probability theory than Swinburne does, it is at least as probable that you are the one who's mistaken as it is that Swinburne is. I wouldn't make this an issue if your comments had been a bit more reserved, or if you had included a disclaimer that's similar to what you do include in your blog post when you write: "There is another possibility: that I’m mistaken in my own understanding of Bayes’ Theorem, and that Swinburne’s use of it is in fact justified. I’ve given my reasons for thinking I’m right. Bayes’ Theorem is well enough documented. You can look it up and judge for yourself."

Since my post was too long for the blog software to accept, I'll split it. Part 2 below.

Haecceitas said...

Part 2

"It would be more accurate to characterise me as a skeptic. If and when decent evidence for God's existence appears, then I will give it all due credit."

Given this attitude, you do seem to presuppose that it is meaningful to make probabilistic judgements pertaining to God. How would you go about making such judgements, given your criticisms of Swinburne's approach?


"I'm well aware of the concept of epistemic probability. The term was coined by the statistician Ian Hacking, and is an entirely subjective or qualitative concept - in essence describing the strength with which you hold a belief (or if you prefer, the degree of confidence you have in the truth of a proposition). As he is a statistician, Hacking of course knows far better than to attempt to confuse epistemic probability with statistical (or frequentist) probability by applying statistical mathematics to epistemic issues as if the subjective probabilities could have objective numbers applied to them. So for instance, when Swinburne in 2002 gave a lecture at Yale in which he calculated the probability of the Resurrection as being 97%, I know he was either mistakenly or deliberately conflating the two concepts."

Well, if your objection is that such probability judgments are subjective, then you have the entire school of subjectivist bayesians to deal with. In that case, your beef isn't just with Swinburne.

I don't know what Swinburne said in that particular lecture, but what makes you think that he isn't aware of the subjectivity element involved in his argument? It would seem to me that his argument isn't intended as establishing that there is some kind of "objective probability" of 97% for the resurrection. But neither does he pull the probabilities out of thin air in his argument. He argues for them in a rigorous manner in his book, and tries to err on the side of caution. In principle, it can be no more objectionable for a philosopher to argue that the probability of the truth of a proposition is 60% than it is for him to argue that it's "somewhat more probable than not". This holds equally well even if the proposition is of the type that can't be examined in an empirical manner or on the basis of statistical data.

Jonathan West said...

Haecceitas
Unless you claim to have better credentials on Bayesian probability theory than Swinburne does, it is at least as probable that you are the one who's mistaken as it is that Swinburne is.
It should be a straightforward matter for you to look up those authorities and see what they say about the specific objection I have made. But also ensure that you consult authorities on statistics concerning their use of Bayes' theorem and on the limitations on its usefulness.

Remember that Bayes' theorem needs numbers. Those numbers are either derived from real-world statistics objectively obtained, or they are made up. In the latter case, processing them through Bayes' theorem simply means that the result is also made up.

When you have done your research, you will be in a position to say on the basis of evidence (rather than by appeal to authority) whether or not I am wrong.

Take as long as you need to gather the evidence.

Given this attitude, you do seem to presuppose that it is meaningful to make probabilistic judgements pertaining to God. How would you go about making such judgements, given your criticisms of Swinburne's approach?
Swinburne's approach is essentially to say of his hypothesis (I'm paraphrasing here) "this seems sensible to me, so I'll call it probable" and he then goes on to apply an arbitrary number to that probability. Now, Hacking's definition of epistemic probability makes it clear that the assessment of probability is subjective. You cannot make an objective assessment of epistemic probability because you do not have the data to work from, since the assessment is too bound up with the complex of your other beliefs.

The mathematics of probability theory has demonstrated very effectively that our intuitions in respect of probability assessments are deeply flawed. Innumerable scientific experiments have been performed which demonstrate this. If you can get hold of a copy of Ben Goldacre's Bad Science (I don't know if it is on sale in the US) I would strongly recommend you read chapter 12 "Why clever people believe stupid things". This link describes the relevant chapter in brief.
http://2020science.org/2008/11/09/why-clever-people-believe-stupid-things/

So, my criticism if Swinburne is twofold. First of all, he is trying to obtain for himself a spurious and unwarranted degree of authority for his views by passing them off as more precise and justified than they really are, in that he is arbitrarily assigning numbers to his intuitions. Swinburne has this in common with the entire school of subjectivist Bayesians, though given the manifest nonsense they peddle, to describe them as a "school" is an insult to education.

Secondly, he is taking the anti-scientific approach of making himself a theory (his hypothesis of God) working it out in great detail (God's omniscience, omnipotence, omnipresence, perfect goodness, eternal existence and so on) and only then looking (highly selectively) at the evidence in his subjectively bayesian manner, and of course finding that his intuitive evaluation of the evidence confirms his intuitively accepted hypothesis. How could it be otherwise?

The correct approach (in scientific terms) is to see what phenomena are currently unexplained by existing theories, and see what additional refinements or new ideas might explain those phenomena, and also predict the results of observations not yet made.

In all the history of science since it started to be a coherent and organised activity, every scientific discovery concerning a phenomenon previously thought to be evidence of God working in the world, the phenomenon has turned out to be explained by the consistent operation of unchanging natural laws. There hasn't been a single case where the scientists have concluded "Oh, it turns out that this is God working in the world after all". Now, this is not proof that God doesn't exist, but the evidence forms a valid C-inductive argument for the hypothesis of his nonexistence.

Haecceitas said...

"It should be a straightforward matter for you to look up those authorities and see what they say about the specific objection I have made."

This still misses my point. If it is a straightforward matter for anyone to see what the authorities say about this, it is just as straightforward for Swinburne as it is for you and me. Since it is unreasonable to think that a person with Swinburne's credentials would be unaware of what the authorities say, it's probably the case that your disagreement with him boils down to some issue in which he is not self-evidently mistaken. The point remains that unless you claim to be better credentialed in this field than Swinburne, you have no grounds for thinking that it's more probable that Swinburne is the one who's in error here - even if you remain confident that he's in error after reviewing some of the literature to the best of your ability. You even acknowledge this possibility in the post that you made on your own blog.


"Remember that Bayes' theorem needs numbers."

Obviously, no one is disputing that.


"Those numbers are either derived from real-world statistics objectively obtained, or they are made up."

Seems like a false dichotomy, unless you intend to expand the definition of "made up" to include much more than it typically does.

The simple fact of the matter is that we all make probability judgements all the time without relying on "real-world statistics objectively obtained", and we manage to do this with pretty good results. There are people who are pushing the Bayesian approach to fields of study such as history, where it's simply unrealistic to suppose that the numbers can be anything but "made up" (in your sense of the term). So again, your beef isn't just with Swinburne, it's with all of the people who are making such applications of Bayes' theorem.

"In the latter case, processing them through Bayes' theorem simply means that the result is also made up."

I would certainly agree that the "garbage in, garbage out" principle applies to Bayes' theorem, but I'm disputing the claim that anything but "real-world statistics objectively obtained" will fall within the category of garbage.

Here's the crux of the matter as I see it. You can't really even get off the ground in the whole project of epistemology if you don't grant that the human mind is capable of gaining prima facie reliable knowledge about the world. Since the world is contingent rather than necessary, such knowledge must be probabilistic in nature, rather than simple deduction of logically necessary truths. Therefore, either we are left in the total darkness of global scepticism, or else we are indeed capable of making at least semi-reliable probabilistic judgements on grounds other than statistical analysis. But if this is the case, then I see no way around the fact that these ( at least semi-) reliable probabilistic judgements can be legitimately represented in numerical form. If I'm a person whose cognitive faculties are working in a more or less reliable way, then I can conclude with some epistemic justification that A is very improbable, B is moderately improbable, C is very probable, etc. But if this is the case, there can be no objection to representing these in numerical form (especially if one errs on the side of caution).

Haecceitas said...

"Swinburne's approach is essentially to say of his hypothesis (I'm paraphrasing here) "this seems sensible to me, so I'll call it probable" and he then goes on to apply an arbitrary number to that probability."

You didn't actually answer my question. You said that you're open-minded towards finding any evidence for God's existence when such is produced. But presumably you're not open-minded towards evidence for the existence of square circles. Indeed, you're probably reasonabe enough to not waste much time with evaluating any alleged evidence for the existence of square circles. Therefore, your own comments presuppose that the whole concept of evidence for the existence of God is in a different category. I fail to see how you can maintain this openness while holding to all of your criticisms toward Swinburne's approach, since presumably they can be reformulated against any inductive arguments for theism.

"The mathematics of probability theory has demonstrated very effectively that our intuitions in respect of probability assessments are deeply flawed."

You certainly won't get an objection from me if your point is just that intutive probability assessments are much more fallible than those that are based on solid data. But where has Swinburne claimed otherwise?

Also, keep in mind that both sides of the theism vs. naturalism debate are equally vulnerable to this objection, since the most fundamental metaphysical questions can't possibly be examined in the strictest scientific way. That doesn't mean that one should not aim at the maximal level of rigour that is attainable.


"Secondly, he is taking the anti-scientific approach of making himself a theory (his hypothesis of God) working it out in great detail (God's omniscience, omnipotence, omnipresence, perfect goodness, eternal existence and so on) and only then looking (highly selectively) at the evidence in his subjectively bayesian manner, and of course finding that his intuitive evaluation of the evidence confirms his intuitively accepted hypothesis. How could it be otherwise? The correct approach (in scientific terms) is to see what phenomena are currently unexplained by existing theories, and see what additional refinements or new ideas might explain those phenomena, and also predict the results of observations not yet made."

One can hardly be faulted for being "anti-scientific" if one doesn't apply the scientific method as such to questions of metaphysics! Metaphysical Naturalism and Theism are on par in terms of not being hypotheses that are beyond scientific verification or falsification. In essence, it seems like you're now disqualifying metaphysics as a whole after undermining epistemology. What's left is apparently some form of naive scientism that builds sandcastles in mid-air without realizing the need for philosophical foundations.


"In all the history of science since it started to be a coherent and organised activity, every scientific discovery concerning a phenomenon previously thought to be evidence of God working in the world, the phenomenon has turned out to be explained by the consistent operation of unchanging natural laws. There hasn't been a single case where the scientists have concluded "Oh, it turns out that this is God working in the world after all"."

Depending on how you define the key terms here, this is either false or trivially true. Obviously, nothing could count as a scientific discovery that verifies God's action in the world when science is pursued under the constraints of methodological naturalism. But clearly there have been many and are scientists who have concluded that some particular aspects of science are evidence of God. Also, it's simply mistaken to assume that theism is necessarily committed to some particular interventionist type of mechanisms in explaining events in the world, so it won't do to just claim that the existence of scientific explanation for various aspects of the world will automatically show that the theists were wrong to see God's handiwork in them.

Haecceitas said...

There are a few typos in my previous posts, but I'll just fix this one:

"Metaphysical Naturalism and Theism are on par in terms of not being hypotheses that are beyond scientific verification or falsification."

Should be "in terms of being" rather than "in terms of not being".

PS. Brian, are you sure that the max 4000 characters per post limitation is the best way to go?

Brian said...

Haecceitas,

I am not sure the 4000-character limit is something that can be removed from blogger. But if you know of a way, please provide a link.

Thanks.

Jonathan West said...

Haecceitas,

The point remains that unless you claim to be better credentialed in this field than Swinburne, you have no grounds for thinking that it's more probable that Swinburne is the one who's in error here - even if you remain confident that he's in error after reviewing some of the literature to the best of your ability.

Let us conduct a thought experiment. Imagine for the moment that I am an extremely eminent astrologer. I have written lots of learned books about astrology, have taught lots of students who have gone on to become professional astrologers, and I know all the details of the differences between classic tropical astrology, sidereal astrology, the Sarjatak system, and many of the hundreds of other astrological systems that have been used by men down the ages.

Would you bow before that authority if I were to claim - on the basis of astrology - that the current position of Jupiter in the sky will have a direct and discernable impact on your job and marriage prospects?

I rather suspect you would not. No matter how impressive my astrological credentials were, and no matter how great my knowledge of the different astrological systems, you would still conclude that the primary claims of astrology are bunkum, and you would not regard it as necessary to know the finer points of tropical or Sarjatak astrology to be confident of maintaining your view in the face of my expertise.

The question therefore arises as to what relationship with authority and credentials you choose to maintain. If (as I suspect) you do not unquestioningly accept all claims of authority, then the issue is how you choose which claims to accept and which to reject.

Now, there are two broad classes of approach here. One is to believe those who say things similar to the views you already hold. The other is to do what you can to ascertain the reliability of different authorities, by making a direct examination of the evidence for yourself.

If you are going to choose the first option, then this conversation will have to come to an end, because you will simply disbelieve anything I say that conflicts with your chosen authorities, and I am in no mood to conduct such a futile conversation.

If you are going to choose the second option, then you are going to have to do a bit of research into statistics yourself and find out whether my comments are right or not.

I can address your other points afterwards, if and when I am convinced that it is worth my effort continuing with this conversation. Persuade me.

Haecceitas said...

It seems to me that your analogy isn't very good one. The rational merits of astrology as a field of study are questionable in the extreme. Therefore, my reluctance to accept any astrologer's authority isn't just based on my questioning of his credentials as an astrologer. It has more to do with my distrust in astrology in general. But sure, there are certain aspects of astrology (such as knowledge of its history, its common practises, etc) that I could take on the authority of someone who has "credentials" on such a field.

In order to contextualize this to our discussion, this particular point in our interaction has been about Swinburne's credentials in Bayesian probability theory. Unlike in the case of astrology, the validity of the subject matter itself isn't in question. It is at this point that your analogy totally breaks down.

So, the issue is twofold:

- Do you claim to have better credentials than Swinburne in Bayesian probability theory?
- If not, then what are your grounds for thinking that on any particular issue where your view radically differs from that of Swinburne, it is more likely that you are correct and he's wrong?

By the way, I wonder if you've tried to contact Swinburne directly. His e-mail address is available at his Oxford faculty page.

Also, one thing that might bring some more clarity to the issue would be his book on Epistemic Justification. I admit that I haven't read it yet, but am hoping to read it sooner or later.

Jonathan West said...

The point is that I do not accept that subjective Bayesian inference and the assignment of numerical probabilities to subjective epistemological issues has any value or validity.

Therefore, it doesn't matter in the least how eminent Swinburne is in the field of subjective Bayesian inference. It is as flawed as astrology. If you are not prepared either to agree with this or at least to explore with me the issues I have raised on this question, but will instead continue to assume that Swinburne's expertise by definition trumps my own views, then we cannot continue this conversation.

Let me offer you a short example from the conclusion of chapter 6 of The Existence of God to illustrate what I mean.

I have argued in this chapter that there is a modest probability intermediate between 1 and 0, to which I will give the artificially precise value of 1/2, that a God will create humanly free agents located in a beautiful physical universe, perhaps also containing animals.

If you understand probability and statistics it is possible to decode this. It means that by his own admission none of his arguments are sufficient to make any kind of estimate of probability in the matter (it's impossible by definition to have a probability outside the range 0 to 1, so he is in fact ruling nothing out at all), but because he needs a number in order to proceed to the next chapter, he's going completely arbitrarily to decide to use 1/2.

But if you aren't familiar with probability and statistics (including Bayes' theorem amongst other statistical techniques), you might think that his statement sounds all very scholarly and sensible. It is in fact 46 words of concentrated drivel, and I have no hesitation in saying so.

But it is very educated and reasonable sounding drivel. Note the tasteful use of the term "modest intermediate probability" as if he were being really very cautious and conservative about the conclusions that could be drawn from his data. Nothing could be further from the truth.

(By the way, on the available evidence, I think Bayes himself would not be regarded as a Bayesian by the modern definition. He was writing about statistical probability, not subjective inference or epistemic probability.)

Haecceitas said...

Well, I think that most of the points you jut brought up are either directly or indirectly answered in my posts #12 and #13 which you haven't dealt with yet. So if you want to continue this discussion, you might want to respond to them.

Jonathan West said...

Haecceitas
Does that mean that you accept the validity of my criticisms, or at least are prepared to investigate them with me on the basis of evidence directly presented?

If so, then I would like you to explain in your own words, without reference to authority, what you think is mistaken about my criticism of Swinburne's use of Bayes' Theorem.

I'm not prepared to debate with you if you are going to run off and appeal to authority for any point you find it difficult to deal with from your own knowledge.

Haecceitas said...

Any neutral observer can go through the previous posts and see that I haven't supported my point of view merely by appeals to authority. So you're very close to setting up a straw man here. Only one of my several points had anything to do with the issue of authority.

Jonathan West said...

The key point and the basis f my criticism of Swinburne is his misuse of Bayes' Theorem and his assignment of numerical values to subjective assessments of the strength of his beliefs.

It is that specific point where as far as I can tell your sole response has been to claim that Swinburne has more credentials than I have, and that you are therefore disounting my view on the basis of your perception of Swinburne's greater authority.

On this specific point, as far as I am aware, you have made no attempt to engage with the substance of my criticism, relying solely on your acceptance of Swinburne's authority and your consequent assumption that I am wrong and he is right. If you aren't prepared to address the substance of this point, I might as well hold a conversation with a speak-your-weight machine.

Since this point acts as the foundation for all my other criticisms of Swinburne, unless we can get past the question of evidence versus authority on this issue, there is no purpose in progressing to any other.

If I'm wrong, then all my other criticisms become null and void. If I'm right but you refuse to engage the point, then discussion of the other topics will be a dialog of the deaf.

So, I'll ask one last time. Using your own words, please express your opinion as to what errors of fact or reasoning I have made in my criticism of Swinburne's use of statistical techniques such as Bayes' Theorem.

Haecceitas said...

"On this specific point, as far as I am aware, you have made no attempt to engage with the substance of my criticism, relying solely on your acceptance of Swinburne's authority and your consequent assumption that I am wrong and he is right."

This is very interesting. I wonder if anyone who happens to be reading this thread could give his/her honest opinion. I'm simply amazed if you have read all of what I've written and still think that the only points that were relevant to this specific point were the ones that dealt with the issue of authority. But if we get a few comments from others and they think the same, then I might conclude that I haven't communicated as clearly as I thought I have.

Jonathan West said...

Haecceitas
Humor me. Copy & paste the exact bit where you described my errors of reasoning. I must have missed or misunderstood that bit.

Haecceitas said...

"The key point and the basis f my criticism of Swinburne is his misuse of Bayes' Theorem and his assignment of numerical values to subjective assessments of the strength of his beliefs."

Let's split this to two logically distinct points.

(A) OK, first of all, Bayes' Theorem pertains to the relations between probabilities, and thus it contains no inherent limitation as to what must be the source of the assigned probability values. So one can't violate the theorem itself by using any source imaginable. The point is that if the values are representative of actual meaningful probabilities, the theorem can be used to get reliable results about the implications of these probabilities.

(B) This leaves another issue, namely, whether or not there are legitimate ways of estimating (at least rough) probabilities in situations where statistics can't determine them. In this thread, I've made several points in favour of the view that such probability estimations can be made (which you seem to dispute). Here are just a few (plus one that I didn't make).

1. I made the point that a denial of this fact would logically lead to global scepticism. Human knowledge is not possible without prima facie reliable beliefs about what's likely to happen in particular circumstances.
2. I made the point that in actual fact, Bayes' Theorem is used by (for example) some scholars of history and some seem to think that this will become more common. Historians don't have the luxury of relying on statistical data for most of their crucial estimations of probability/plausibility.
3. I could add yet another example. Law courts regularly make judgements about what's probable "beyond reasonable doubt" and these judgements are heavy enough to send people to prison for life (or even to death).

These (especially #1 and #3) will constitute in essence a reductio ad absurdum argument against your view (as I understand it). The consequences of maintaining this view (global scepticism in epistemology, total unreliability of the procedures used to make very significant verdicts in the courts of law, etc.) seem too absurd for your view to be taken seriously. So you should either show how those consequences don't follow from your view, or else you could try to justify why accepting those consequences is more reasonable than rejecting the view that entails them.

Based on some of your online writings, you seem to have no objection to the way that most atheists claim God's existence to be very improbable. (You note the distinction between very improbable and impossible approvingly in one of your articles.) Do you think that these atheist are basing their views on objectively obtained real-world statistics?

Jonathan West said...

Haecceitas
That's much better. Now I know what it is we disagree about.

OK, first of all, Bayes' Theorem pertains to the relations between probabilities, and thus it contains no inherent limitation as to what must be the source of the assigned probability values. So one can't violate the theorem itself by using any source imaginable. The point is that if the values are representative of actual meaningful probabilities, the theorem can be used to get reliable results about the implications of these probabilities.
You are making the same mistake as Swinburne himself. The weakness of the argument is the phrase "if the values are representative of actual meaningful probabilities". They aren't. It is only possible to assign meaningful numerical values to probabilities if you have a population of entities on which you can perform statistical calculations.

And this is recognised by creationists who look for disproofs of Darwin's theory of evolution. It doesn't matter how many fossils or bits of related DNA are found that are consistent with the theory, only one confirmed out-of-sequence fossil or one piece of irreducible complexity would be sufficient to overthrow Darwin. Behe's bacterial flagellum of course isn't it. But as you can see, it would be meaningless to try and say that Darwin's theory is 99.7% probably true, or any other figure you care to arbitrarily select. As Darwin's theory is in effect a C-inductive argument as described by Swinburne (though one with a huge number of confirming premises), all we can say is that all the available evidence supports it. We cannot estimate the chance that some future unknown piece of evidence will change that situation.

1. I made the point that a denial of this fact would logically lead to global scepticism. Human knowledge is not possible without prima facie reliable beliefs about what's likely to happen in particular circumstances.
Reliable beliefs can be based on valid C-inductive arguments without being able to put numbers on them. In fact, a significant proportion of our beliefs are not reliable but we hold them even so. In evolutionary terms what matters is not whether a belief is held with complete reliability, but whether a mistake poses a risk to survival. So if for instance you see a lion looking hungrily towards you, you would think it a very good idea to be somewhere else very quickly. The mere fact that occasionally you would be incorrect and the lion has already eaten recently doesn't matter - better to be on the safe side. But a mistaken belief that the lion isn't hungry would be disastrous.

2. I made the point that in actual fact, Bayes' Theorem is used by (for example) some scholars of history and some seem to think that this will become more common. Historians don't have the luxury of relying on statistical data for most of their crucial estimations of probability/plausibility.
Actually, there are quite a lot of statistics available to historians, and where they are available it is perfectly valid to use statistical techniques to analyse them. But it is not valid to use statistical analysis when you are inventing the numbers, no matter how justified you feel by doing so.

3. I could add yet another example. Law courts regularly make judgements about what's probable "beyond reasonable doubt" and these judgements are heavy enough to send people to prison for life (or even to death).
In the majority of cases, the cases are presented in terms of C-inductive arguments (though that term is not actually used when describing things to the jury. In cases where statistics have been used in court cases, there have been some notable miscarriages of justice, for instance in the Sally Clark case. As it happens this is a case where Bayes' Theorem should have been used in analysing and presenting the evidence, as the case hinged on the statistical improbability of two cases of SIDS being suffered by the same family, and this is something which could be calculated and given a numerical estimate.

Ryan said...

(part 2 of 2)

Second, you’re argument so far is an undercutting defeater for Swinburne’s project. That is to say, even if you’re right that Swinburne’s project fails due to his use of Bayes’s theorem, that doesn’t rebut his conclusion (i.e. that God probably exists.). It simply reduces the warrant for accepting that conclusion. Therefore, in order for your debating partners to bolster that warrant, we can simply offer an undercutting defeater-defeater of our own to your proposed defeater. Thereby, we reduce the warrant for accepting you’re critique and simultaneously restore the original warrant (to the degree there was such warrant) for Swinburne’s approach. Hacceitas has offered several undercutting defeaters: (1) that you’re approach leads to global skepticism and (2) that Bayes’s theorem is used reliably by in other fields. In addition to those undercutting defeaters, I take my first point, in the earlier post, to be a rebutting defeater that destroys any warrant for your position at all. Thus, to the extent my critic is successful, you’re argument has no warrant, or to the extent Hacceitas’s critics are successful the warrant for your position is reduced.

Third, you’re also mistaken about the relationship between C-inductive and P-inductive arguments. Again, Swinburne deals with exactly this point in his book in the first chapter. I’m starting to wonder how closely you’ve read his material. In the legal analogy, you’re correct to point out that a single piece of evidence (say a finger print, or eyewitness testimony) is a C-inductive argument. That piece of evidence is probative of the defendant’s guilt to the extent it raises the probability that he committed the crime. But you’re incorrect to the extent you suggest that the story ends there. All the individual pieces of evidence, when considered together as a set, as juries and judges are required to do, amount to a P-inductive argument for the defendant’s guilt or innocence. Swinburne discuses this (precisely in relation to evidence of guilt or innocence) on pp. 13, 16-17.

In conclusion, you’re mistaken about Swinburne’s use of values for the probability calculus, you’re mistaken about the relationship between correct C-inductive and P-inductive arguments. And you’ve failed to rebut Hacceitas’s undercutting defeater regarding legal evidence.

Ryan said...

[Looks like my posts went in backward.]

Jonathan West:

Three points. First, I think you’ve mistakenly characterized Swinburne’s project in his book The Existence of God. You indicated that Swinburne gives arbitrary numerical values to the various elements of Bayes’s theorem and then concludes that the God hypothesis is >.5. You argue that this method of assigning arbitrary numerical values and then deriving probabilistic conclusions results in a misleading confidence in the conclusion.

But you’ve mistaken Swinburne’s view here. Swinburne does not think that one must, necessarily, assign numerical values to Bayes’s theorem. He makes this very clear on pp. 17-18, 68. Swinburne: “I have claimed Bayes’s theorem is true, but I had better make clear what I mean by saying this….In so far as they [i.e. the various e, h, and k, elements] cannot be given precise numerical values, my claim that Bayes’s theorem is true is simply the claim that all statements of comparative probability that are entailed by the theorem are true. By statements of comparative probability I mean statements about one probability being greater than, or equal to, or less than another probability. (Such statements are sometimes all that we can justifiably assert about some probabilities….) (p.68). And, again, on pp. 17-18, Swinburne explains: “In using the symbols of confirmation theory I do not assume that an expression of the form P(p/q) always has an exact numerical value. It may merely have relations of greater or less value to other probabilities, including ones with a numerical value, without itself having a numerical value….Clearly, for example, we may judge one scientific theory to be more probable than another on the same evidence while denying that its probability has an exact numerical value; or we may judge a predication to be more probable than not and so to have a probability of greater than ½, while again denying that that probability has an exact numerical value.”

Thus, you’re assertion, if I’ve read you correctly, that we (and Swinburne) must assign specific numerical values to the elements of the probability calculus is simply mistaken.

(part 1 of 2)

Jonathan West said...

Ryan
Swinburne's statement can't be used in support of your point. Bayes' Theorem is true, as it is a mathematical theorem for the manipulation of numerical probabilities which has been proved with the same degree of rigour that has proved that there are an infinite number of prime numbers. It can therefore be used to make statements of comparative probability. However, in order to make such statements, some assignment of numerical values has to be made. It may be that the assignment is to an undetermined value within a determined range, but if you say that event x has a greater probability than event y, you are making statements about the relative numerical values of x and y. You can't have a "greater than" comparison except between two quantities that have been expressed, however vaguely, in numerical terms. As I have pointed out in comments here and in more detail in my own blog article on Chapter 1, that is an invalid approach when dealing with C-inductive arguments.

As for your second point, that my line of argument is an undercutting defeater of Swinburne's project, I would say two things. First, I accept the principle of using both P-inductive and C-inductive arguments in pursuit of the truth of the matter. It is Swinburne's claim that C-inductive arguments can be used to make numerical probability comparisons which is invalid. Second, I haven't yet blogged any of the chapters in which he examines specific arguments for or against the existence of God, so I ask your patience in reading my comments on the rest of the book, once I get round to writing them) before reaching any conclusion as to the totality of my argument.

As for being mistaken about the relationship between C-inductive and P-inductive arguments, I have to point out that I am not. In a court case where lots of pieces of evidence are assembled, these pieces of evidence usually do not form a population of entities on which you can perform statistical calculations to assess the strength of a P-inductive argument. Rather, they form multiple premises which all individually contribute to a C-inductive argument. No matter how many pieces of evidence place somebody at the scene of a crime, a single confounding piece (e.g. that it was known that the person was in hospital immobile with his leg in plaster at the time) would rule out any possibility of his involvement.

Haecceitas said...

Sorry for the delay in replying, but it'll probably continue that way. Luckily Ryan is here as well.

"The weakness of the argument is the phrase "if the values are representative of actual meaningful probabilities". They aren't. It is only possible to assign meaningful numerical values to probabilities if you have a population of entities on which you can perform statistical calculations."

I split the contention into two distinct points (A and B). Since your only criticism of A is about the truth of the condition that this conditional claim requires, your criticism actually falls within B (since it doesn't apply to the statement in its conditional form). So we can focus on B.


"Reliable beliefs can be based on valid C-inductive arguments without being able to put numbers on them."

Valid c-inductive arguments can give reliable beliefs only if we do make at least the roughest kind of probability estimates. Upon a moment's consideration, one can see that unless there is a way to get at least semi-reliable knowledge pertaining to the cumulative effects of the c-inductive arguments, it remains unclear as to whether they make the conclusion overall probable (and to what extent).


"In fact, a significant proportion of our beliefs are not reliable but we hold them even so."

Sure, but one can't remain consistent while exaggerating this problem. This is essentially the same problem that the extreme sceptic has to face in epistemology. One can't point to cases of error as evidence of fundamental unreliability of our belief-forming faculties since total unreliability of these faculties would not even allow for any justification for the claim that these are known cases of error.


"Actually, there are quite a lot of statistics available to historians, and where they are available it is perfectly valid to use statistical techniques to analyse them. But it is not valid to use statistical analysis when you are inventing the numbers, no matter how justified you feel by doing so."

Yes, but what I'm talking about is a more comprehensive Bayesian approach to history. That can't possibly be grounded on what you'd take to be a perefeclty valid use of statistical techniques.


"In the majority of cases, the cases are presented in terms of C-inductive arguments (though that term is not actually used when describing things to the jury."

But unless there is a somewhat reliable way to convert the cumulative effect of the c-inductive arguments to an estimation of whether or not the evidence not only confirms the hypothesis of the person's guilt, but also makes it very probable (beyond reasonable doubt), I fail to see how the view that results from the application of your critcisms to the context of law courts can be anything but rather deep distrust in the validity of the system.

Haecceitas said...

"No matter how many pieces of evidence place somebody at the scene of a crime, a single confounding piece (e.g. that it was known that the person was in hospital immobile with his leg in plaster at the time) would rule out any possibility of his involvement."

Firstly, I don't see how this defeats the objection. Surely there can be court cases where there's no such clear confounding evidence. How can the system be reliable in those cases if c-inductive arguments can't be converted to the equivalent of a srong p-inductive argument?

Second (and perhaps mildly but not more than mildly nitpicking) point would be that at least the example that you gave is disputable. There are many conceivable scenarios where one could have prima facie reliable evidence that the person was immobile at the hospital at that time that would in fact be compatible with the person's actual guilt of a murder (or whatever) in another location (think of twin brothers, paid actors, bribed nurses, etc.). So one would have to weigh the positive evidence against this piece of confounding evidence and get a somewhat reliable estimatimation of their relative strength.

Jonathan West said...

Haecceitas
All your points seem to me to be largely variations on the same theme, which can be described by the following question you put.
How can the system be reliable in those cases if c-inductive arguments can't be converted to the equivalent of a strong p-inductive argument?

And all I can say is that statistics don't work that way. Don't take my word for it, as I claim no authority in this matter. Instead, buy yourself a textbook on statistics and look it up.

There is a distinction to be made between qualitative and quantitative estimates of probability. On the basis of a C-inductive argument, you can decide beyond reasonable doubt that a person is guilty of a crime, or that Darwin's theory of evolution is true. As it happens, the confirming evidence for the latter is far more extensive and comprehensive than anything ever brought before a court. But we cannot put numbers on it, because to put numbers on it would be to assign an arbitrary numerical value to the chance that confounding evidence as yet unknown will be found.

Comparing qualitative and quantitative estimates of probability is like comparing apples and oranges. We make those rough estimates all the time in order to determine things to subjective standards such as "Beyond reasonable doubt".

But those involved in running clinical trials do have statistics to work from, and there are a whole battery of statistical tests that they can use, to determine the probability that a particular observed proportion of some effect is down to mere random variation. If the chance if it being merely down to random variation is low (the cutoff point is usually set at 5%), the result is described as being statistically significant.

"Beyond reasonable doubt" is a subjective criterion". "Statistically significant to 95%" is an objective one.

Where you can obtain and make use of objective probabilities derived from statistics and processed using statistical analytical techniques, it is generally better to do so, and this is recognised by the public, even when most members of the public have not the faintest idea of how the statistics are derived and what degree of certainty there actually is in the figures. The public therefore tend to treat opinions backed by numbers as carrying more authority.

And that is why many people with an agenda to push are desperately keen to turn their hunches and guesses (and even their C-inductive arguments) into statistics, so that they can claim the authority of numbers for their view.

If you understand how statistics work, you can know better. I strongly recommend gaining an understanding of statistics. It will give you a most marvellous insight into all sorts of matters, from the claims made by advertisers, political polls, op-ed pieces in the papers, and all varieties of snake-oil.

Post a Comment

Thanks for taking the time to comment. By posting your comment you are agreeing to the comment policy.

Blog Archive

Amz