Paul Draper’s burden of proof for the theist
May 11, 2012 — 13:09

Author: Helen De Cruz  Category: Atheism & Agnosticism  Tags: , , , ,   Comments: 17

[x-posted on Newapps] A few days ago, I had the privilege of attending a lecture by Paul Draper, probably one of the most prominent atheist philosophers of religion today. His lecture had a wealth of ideas (including a proposed solution to Hume’s problem!), but I’d like to focus on one tiny piece of the lecture, viz. his argument that the burden of proof is on the theist, and not on the atheist.

Here goes the argument, which Paul was kind enough to discuss with me, prior to posting it. I apologize if there are any remnant misrepresentations.

Let’s assume that there are a number of epistemically possible world views: some are naturalistic, some are supernaturalistic, let’s even grant there are others (non-supernatural, non-natural, but some third, unknown view). Then we can see that the following diagram exhausts all epistemic possibilities: N (naturalism), S (supernaturalism) and not-N and not-S.

Now, according to Draper S and N are equally epistemically modest (see below). There is no clear definition for epistemic modesty, except perhaps a comparative sense: a statement p is epistemically more modest than a statement q if (or iff?) p is less informative than q (e.g., p says I have a car, q says I have a red toyota; p says linda is a bank teller, q says she’s a bank teller and a feminist etc.) In the absence of prior information, epistemically modest statements are more likely to be true than epistemically less modest statements.

  • S holds that the original cause of the world is mental.
  • N holds that the original cause the world is physical.

Thus according to N the physical causes the physical and the mental (except in the case of eliminative physicalism, where the physical only causes the physical). According to S the mental causes the mental and the physical (except in the case of eliminative idealism, where it only causes the mental).

These two world views seem equally epistemically modest to Draper. They both have a burden of proof. Theism (as opposed to supernaturalism more generally) is a much less epistemically modest position than both generic S and N. Theism makes claims about the nature of the supernatural (God is one person, omniscient, omnipotent, perfectly good, etc.) and is thus considerably less epistemically modest than atheism (see figure). Atheism is per definition everything that is not-theism (for Draper), so it encompasses the entire gray area on the figure. Theism is a small speck in the realm of supernatural epistemic possibilities. So according to Draper, the atheist’s burden of proof is lifted. Since the theist argues for a position that is less epistemically modest than the atheist, the theist carries the burden of proof.

My main qualm with this argument is that Draper represents the atheist as the holder of a generic worldview, compared to the epistemically immodest theist. However, I think this is a misrepresentation of how most atheists would view their position. I doubt it if it’s a kind of generic view that is in principle compatible with the existence of all sorts of supernatural beings like elves and demons as long as they are not God. I think the atheism that is usually argued for is a form of scientific naturalism, which, like theism is a small part of all epistemic possibilities (it excludes supernaturalism, not-supernaturalism and not-naturalism, and non-scientific naturalism).

Draper concedes that a scientific naturalist does have a burden of proof, but argues that his main point remains that one can be justified in rejecting theism without evidence but one cannot be justified in accepting theism (or any other specific view like scientific naturalism or even naturalism) without evidence.

  • Trent Dougherty

    Lots to be said here, and I’m in the foyer about to head into vigil Mass, but one thing I want to say right away. Even if the generic/specific distinction holds up, it will only be significant to the degree that any other specific supernaturalism takes up any significant probability space. Many theists–myself, and Swinburne seems to agree–that this is not so. That is, even their union doesn’t have much measure. Add up the space taken by polytheism, fairieism, etc. and you just don’t have much. There will be countability issues since there is no upper limit on the number of gods in a polytheism, but they also get correspondingly more improbable as the number goes up. Furthermore, some of that already sad lot will be ruled out by basic empirical evidence in a heartbeat. Then after we renormize, their significance will be even less. There’s the bells, I gotta go!

    May 12, 2012 — 15:58
  • Trent Dougherty

    Hopefully it is not a sin to blog in Mass LOL but it’s just announcements. This issue is of course purely theoretical. We don’t start w a pure ur-probability corrseponding to the logical/a priori prior. We come to the issue w scads of evidence in hand. In reality, I am prepared to argue, the atheist has the burden of proof. Atheism is the anomaly.

    May 12, 2012 — 16:03
  • Mike Almeida

    Seems almost certainly true, I think, that bland atheism (let that be atheism + anything consistent with atheism) will take up more probability space than non-generic theism. But claiming that atheism is therefore the prima facie more rational view is a little like saying that bland naturalism is prima facie more rational than any specific scientific theory S. Surely the probability space that is occupied by the view that the correct scientific view is a naturalistic theory is much larger than the space of S, for any specific S. But why would anyone take the disjunctive theory of all scientific theories consistent with naturalism as a position competing with specific theories? Similarly, why would anyone take bland atheism as a competing position one might take??

    May 12, 2012 — 16:48
  • Joshua Rasmussen

    I’m sympathetic with Draper’s overall approach. I hesitate on the conclusion only because I’m not sure that the prior probability of there being a maximal being is lower than the proper probability of any hypothesis that entails otherwise (excluding obviously false hypotheses, such as that there is nothing.) I could go either way.

    May 12, 2012 — 17:04
  • Joshua Rasmussen

    above: proper = prior

    May 12, 2012 — 17:06
  • jordan.nwc

    Did Drapermove from agnosticism to atheism just recently?

    May 12, 2012 — 18:08
  • It would be really weird to call a polytheist an atheist.

    May 12, 2012 — 21:22
  • Helen De Cruz

    It’s my sense he has. In any case, he hasn’t objected to me calling him such in the preview of the post I showed him. But I’m not sure.

    May 13, 2012 — 2:05
  • Helen De Cruz

    It’s my sense he has. In any case, he hasn’t objected to me calling him such in the preview of the post I showed him. But I’m not sure.

    May 13, 2012 — 2:06
  • Helen De Cruz

    It’s really weird, but there are precedents. The Romans called the Christians atheists. Muslims call polytheists unbelievers.

    May 13, 2012 — 2:10
  • Paul Draper

    I’ve recently become an atheist in one sense: I believe there is no omnipotent, omniscient, omnibenevolent creator of the universe. I’m an agnostic about whether numinous experiences are veridical or delusory. I reject reductive naturalism and all popular religions. I tentatively affirm ietsism.

    May 13, 2012 — 20:09
  • Trent Dougherty

    And for the record, I take theism to be much, much a priori simpler than any version of atheism beyond pure nothingness, if that is metaphysically possible.
    The Draper thesis has strong uniformity assumptions.

    May 13, 2012 — 20:12
  • Helen De Cruz

    Dear Paul: thank you, that was very helpful. I find it interesting that ietsism is getting to be a mainstream English word (it originated in Dutch, my first language, but has apparently become a generally accepted loanword)

    May 14, 2012 — 8:15
  • Jeremy Gwiazda

    It’s been some time since I’ve thought about some of these issues, but a concern of mine is the prior probabilities of one versus two gods. Swinburne places the prior probability of God’s existence at 1 in 100, and the prior of polytheism (many and/or finite god(s)) at 1 in 2000 (bottom of page 395, here):
    But for that to be correct, 1 person must be much more likely, prior-wise, than 2 people. I just don’t think that it is. There are ditheists.
    I wonder if it would be possible to get at this experimentally. Have some story where either one person or two people did something. Then, ask people questions to determine the N where:
    1. The probability that it was done by one person given the evidence equals the probability that it was done by two people given the evidence, and
    2. the probability of the evidence given that it was done by two people is N times the probability of the evidence given that it was done by one person.
    Then we have data on the subjective views of (the ratio of) the priors of one person versus two people. I think. 1 is saying that P(1 person | evidence) = P(2 people | evidence). 2 gives us that P(evidence | 2 people) = N * P(evidence | 1 person). P(evidence) is what it is, and so from 1 and 2 we should be set to compute the ratio of the priors of one person and two people, or P(2 people) and P(1 person).
    But if you told me a story where the footprints are twice as likely to come from 2 people as to come from 1 person (that is, N above is 2), am I really going to come back to you and say ‘I still think it is 10 times more likely that one person did it given the evidence, because the prior of 1 person is 20 times the prior of 2 people’? That just sounds daffy to me. As these thoughts are competely off-the-cuff, I may be missing something though.

    May 14, 2012 — 14:52
  • Helen:
    “The Romans called the Christians atheists.” Well, from the Roman point of view, Christians were practically atheists: they believed in almost no gods, as compared to the Romans. 🙂
    Whether Mormons are or are not Christians, they sure aren’t atheists.
    What if we use complexity-type reasoning? Let’s do it in a finite context.
    Take a formal language all of whose terms have perfectly natural content (no gerrymandering allowed). Suppose that in addition to a finite stock of symbols, we add a period (“.”) that cannot be used in a grammatically correct sentence. Say that an infinite sequence of symbols “inscribes a sentence s” provided that it begins with s and ends with a period (sentences don’t actually have periods in them). Assign equal probability to all infinite sequences of symbols. A sentence s is maximally specific provided that for every sentence s’ in the language, s entails s’ or s entails ~s’.
    Then in the case of those propositions p that are entailed by some maximally specific sentence (for others, we will need an infinitary setting, perhaps using non-standard analysis), we can identify the probability of p with the probability of a sequence that inscribes a maximally specific sentence that entails p, conditionally on the sequence inscribing a maximally specific sentence.
    The probability of an infinite sequence expressing s equals (n+1)^(-length(s)-1), where n is the number of symbols in the language, not counting the period. Roughly speaking, more complex stuff is exponentially less probable.
    Given this, it is very plausible that:
    ExEy(divine(x) & divine(y) & (z)(divine(z) → (x=z or y=z)))
    will receive a probability that’s several orders of magnitude lower than the probability of:
    Ex(divine(x) & (y)(divine(y) → x=y)).
    I don’t find this way of doing prior probabilities too compelling, as the exponential drop-off with complexity seems too fast.

    May 17, 2012 — 14:43
  • Carneades of Ga.- Skeptic Griggsy

    The ignostic-Ockham notes that as God has convoluted, ad hoc assumptions, then , despite Swinburne, He is vastly more complex than naturalism. His putative simplicity then is an ignoratio elenchi.
    Perhaps, as Dawkins and others argue, He’d have to be complex; but, again that is a separate issue.
    I hope that my other posts here aren’t too prolix! Cognitive difficulties enter into the style at times.

    May 24, 2012 — 23:13
  • Ebbedee

    I tend to think that Draper, along with most everyone else on both sides, has a poor notion of both ‘natural’ and ‘supernatural’. It seems to me that Draper thinks that ‘supernatural entity’, such as God, is to be understood as the exact opposite of our own most familiar ‘concrete-ness’. The problem, as I see it, is partly that ‘the concrete’ is itself not quite so easy for any of us to define, either in terms of our senses or in terms of what we think of as ‘the concrete in itself’. For example, only a stooge-of-a-caveman would insist that ‘the concrete’ is, by definition, inert, clunky, boulder-like, dirt-clod-like. To such a cave man, fire is incomprehensibly non-inert, and so it must be magic. To Draper, the concreteness of fundamental physical forces, including, say, combustion and gravity, is, while obscure, mundane and mindlessly regular.
    If I’m right in what I think Draper thinks about this, then I would say that Draper is thinking about the distinction between natural and supernatural like Steven Pinker thinks about the distinction between language and music. Pinker theorizes that music is, or, at least, originated as, a cheap thrill:
    “Music could vanish from humanity and the rest of our lifestyle would continue [on as it always has].”
    But, sounds have visceral power over biology. Sounds have this power because sound is vibration—that prince of motions. Hearing is the brain’s most refined, most specialized, most sensitive sense of motion. And, unless you want to get carried off like so much loose mud by every wind and tide, then you have to vibrate coherently, and you have to identify with other coherent vibrations. In short, the living organism, by definition, is a musical instrument.
    Since language is the device of attaining someone’s undivided attention (as I hope you notice in reading this written language), language is concerned for the interplay of maximum effect and maximum efficiency. So, among doing other things, auditory language co-opts music. In fact, in order for speech not to co-opt music, the living speaker would have to make special effort to pre-edit the musicality out of her speech: she would have to mentally simulate the alarmingly anti-conclusive monotony which is the pedantic version of an ‘irrational admiration of the rational’. Not even an autistic person normally speaks like ‘Computer’ of the original Star Trek TV series.

    July 5, 2013 — 16:20
  • Leave a Reply to Alexander Pruss Cancel reply

    Your email address will not be published. Required fields are marked *