Varieties of Skeptical Theism
December 23, 2010 — 15:08

Author: Ted Poston  Category: Existence of God Problem of Evil  Tags: , , , ,   Comments: 53

The goal of this post is to layout various skeptical theistic theses. Skeptical theism is the position that we should be leery of our ability to limn the limits of God’s reasons for permitting some cases of horrendous evils. Bergmann casts skeptical theism as responding to Rowe’s noseeum inference: (P) No good we know of justifies God in permitting E1 and E2 (the bambi and sue cases) to (Q) No good at all justifies God in permitting E1 and E2. From (Q) one deduces that there’s no God (~G).
Here’s a list of various skeptical theistic theses in the order of strongest to weakest.
First group: evidential irrelevance
1. Necessarily, for any evil, P(G|e)=P(G).
(1) claims that necessarily evil is evidentially irrelevant to the existence of God. (1) is clearly subject to counterexample: let e be a trillion sentient creatures suffer endless torment. Skeptical theism need not be committed to denying that this would be evidence against theism.
2. For any evil, necessarily, P(G|e)=P(G).
(2) restricts the evils to evils that occur in the actual world and claims that they are such that necessarily they are evidentially irrelevant to the existence of God.
3. For any evil, P(G|e)=P(G).
(3) drops the embedded necessity operator. Depending on how one understands the nature of the P function and the nature of evil, 2 and 3 could be equivalent.
4. For E1 and E2 (and similar evils), P(G|E1&E2)=P(G).
This further restricts the evils in our world to those like the Bambi and Sue cases. One advantage of this restriction is that it allows skeptical theism to be viewed as a special case defense and not a general strategy defense. (Also, we can add back in the necessity operator to get further distinctions here).
Second group: Relevance but not significance (my gloss: “evil isn’t a game changer”)
An evil is a game-changer if it can tip the balance of evidence in favor of atheism or agnosticism. A no-game changer version of skeptical theism says that while evil can detract from the probability of God it can’t be the proverbial straw that brought the camel’s back. I shall represent the ‘no-game changer thesis’ by using ‘≈’. This represents that the probabilities are closely similar.
5. Necessary, for any evil, P(G|e)≈P(G).
6. For any evil, necessarily, P(G|e)≈P(G).
7. For any evil, P(G|e)≈P(G).
8. For E1 and E2 (and similar evils), P(G|e)≈P(G).
To undermine Rowe’s inference all Bergmann and company need is 8. Thus, the skeptical theist can easily recognize that there are many evils that could occur that would significantly distract from the probability of theism. Also, 8 is interesting because it allows skeptical theism to be viewed as a special case defense that can be run along with other defenses–the free will defense, the soul-making defense, the value of natural laws, etc.

Comments:
  • Maverick Philosopher touched on evidential irrelevance last month (link)

    December 23, 2010 — 18:16
  • Trent Dougherty

    I’m not saying you intended to imply this (or even that the semantic value of your sentences entail it), but these formulae can’t serve alone to characterize skeptical theism, because theodicy can offer most of these equations.
    But theodicy is totally non-skeptical. Indeed, it proceeds from our understanding about God’s intentions and what is good. It’s the epitome of non-skeptical theism. Swinburne is, of course, the best example of this.

    December 24, 2010 — 16:04
  • Trent Dougherty

    Maybe the difference could be charcterized by an X factor here:
    Pr(e/G&X) = Pr(e)
    X for the theodicist is a narrative of God’s purposes in allowing e (will work for defense, too with some lemmas).
    X for the skeptical theist is considerations about our ignorance of the nature of the good.
    If X = knowledge, then you’re a theodicist.
    If X = ignorance, then you’re a skeptical theist.
    I’m a theodicist. 🙂

    December 24, 2010 — 16:07
  • Ted Poston

    This raises some interesting issues. I originally wanted to get a simple representation of the skeptical theist theses out there. Let’s add a third and fourth category by conditioning not on e but on (P): No good we know of justifies God in permitting e. We then get this group of theses:
    9. Necessarily, for any evil, P(G|P)=P(G).
    10. For any evil, necessarily, P(G|P)=P(G).
    11. For any evil, P(G|P)=P(G).
    12. For E1 and E2 (and similar evils), P(G|P)=P(G).
    Now we get the no-game changer views:
    13. Necessary, for any evil, P(G|P)≈P(G).
    14. For any evil, necessarily, P(G|P)≈P(G).
    15. For any evil, P(G|P)≈P(G).
    16. For E1 and E2 (and similar evils), P(G|P)≈P(G).
    In the original post, I assumed that skeptical theism aimed for a thesis about the proper attitude to have open learning of some evil, viz, it is either evidentially irrelevant to theism or it is not a game changer. Given 9-16, there’s conceptual space for the skeptical theist to deny 1-8 those theses and affirm one of 9-16 in its place. Again, all B and company need is the weak 16. Why might one think this? Perhaps upon initial acquaintance with evil the proper response to the evidence is to think it decreases the probability of a God. But then one begins to think about the resources of a perfect being and while noting that no good one knows of justifies this evil, the epistemic disparity being God and us makes the proper attitude now to be that evil isn’t evidentially relevant (or significant) to God’s existence. (There’s a tricky issue here of how one models changes in states of evidence.)

    December 25, 2010 — 8:55
  • Ted Poston

    Also, you said that theodicy can offer most of these equations. I don’t that’s right. Theodicies hold that evil is a significant defeater to the existence of God and then once one adds the theodicies that makes for a defeater defeater. The theodicist will get evidential irrelevance only once one adds in the defeater defeater. Suppose no one ever thought of the free will defense. In that case, all these equations would be false by the free-will defender’s light, but for the skeptical theist these are all true.

    December 25, 2010 — 8:59
  • Trent Dougherty

    Is I say here (http://el-prod.baylor.edu/certain_doubts/?author=51), I don’t believe in defeaters. But in any event, I think the diachronic picture is the same–in fact I think the last paragraph of your penultimate paragraph implies just this.
    The bottom line is that there is some set of considerations X–whether pertaining to our ignorance of the nature of the good or our understanding of it–which, when added to E, leaves the probability of G much the same as it was before we thought about E.
    If were were thinking about the good the right way in the first place–maybe just doing some armchair reflection on the best goods and how they might be achieved–then when we had our “Buddha moment” and learned of evil, we wouldn’t even be inclined to lower G.
    In fact, unlike the skeptical theist, one might *increase* G. After all, we may suppose your reflections have revealed that virtues like forgiveness, magnanimity, empathy, and compassion are among the best goods possible–no Bergmannian skeptical theist could ever know this. Since reflection also reveals that God would want a world with such virtues and such virtues entail the exemplification of vices and natural disasters, if one failed to observe such things, it would be evidence against the existence of God.

    December 25, 2010 — 18:25
  • Ted Poston

    “After all, we may suppose your reflections have revealed that virtues like forgiveness, magnanimity, empathy, and compassion are among the best goods possible–no Bergmannian skeptical theist could ever know this.”
    A skeptical theist can say that such goods are better than any we know of and that the exemplification of such goods requires some evils. The problematic claim is that these are the best goods *possible*. Think of the logic of comparatives and superlatives. Tom is the best student in the class = for every x, if x is a student in the class and x isn’t tom then tom is better than x and tom is in the class. The way I read your claim is that you put a box out in front. That’s not compatible with ST1. But when you drop the box, it is compatible. Related to this, I don’t think skeptical theism need be a general strategy defense. It could just be a specific instance defense for some very problematic evils, evils of which (P) is true, viz., no good we know of justifies God in permitting these evils. In fact, skeptical theist need not endorse that (P) is true. All that need is the conditional claim that if (P) is true then (Q) doesn’t follow. So one can be a Swinburnean theodicist and a skeptical theist!
    This raises a point I didn’t see in the initial post. Bergmann and company can’t claim that 1 is true, i.e., Necessarily, for any evil, P(G|e)=P(G). Why? B/c Mike thinks that ST2 is plausible, i.e. we have no good reason for that the possible evils we know of are representative of possible evils there are. If this is true, how could we survey modal space and think that any evil isn’t evidence for God? I think that’s interesting. At most, the skeptical theist is entitled to 2.

    December 26, 2010 — 8:42
  • Mike Almeida

    A skeptical theist can say that such goods are better than any we know of and that the exemplification of such goods requires some evils.
    Ted,
    Worth observing that the exemplification of such goods does not require the existence of unjustified evils. At most, it requires the existence of apparently unjustified evils. That would be evidence for ST. But I’m not sure they require even that, since even evils that we know are justified (e.g. terrible pains that we know are necessary to greater goods or the avoidance of worse evils) evoke these valuable responses.

    December 26, 2010 — 10:53
  • Ted Poston

    Mike, Of course. I didn’t suggest otherwise. My point was just that skeptical theism is consistent with a Swinburnean theodicy (or defense). Skeptical theism just needs the conditional claim that if (P) is true (i.e., no good we know of justifies God in permitting E1 & E2) then (Q) does not follow (i.e., it doesn’t follow that “no good at all justifies God in permitting E1 & E2). For my money the best interpretation of skeptical theism is just that Rowe’s noseeum inference isn’t good. To secure that conclusion we don’t need any general view about the evidential role of evil; rather all we need is there’s enough reason to be leery of our ability to limn the limits of a perfect’s being reasons. Consequently, we shouldn’t take (P) a game-changer apropos theism.

    December 27, 2010 — 8:20
  • Mike Almeida

    Ted,
    I guess I misunderstood you. I was referring to the just previous comment where you say,
    . . . the exemplification of such goods requires some evils
    The natural way to read this, I think, is as the claim that some evils are justified by the goods alluded to (the allegedly super-valuable virtues). But maybe you meant that such evils are justified whether or not such virtues flourish as a result and, if the virtues do flourish, that’s icing on the cake. I think this latter position is one that ST’s are committed to, unless there is some principled reason why we do know the point of some evils and we can’t know the point of others. But this is minor, I think.
    For my money the best interpretation of skeptical theism is just that Rowe’s noseeum inference isn’t good. To secure that conclusion we don’t need any general view about the evidential role of evil; rather all we need is there’s enough reason to be leery of our ability to limn the limits of a perfect’s being reasons.
    This is what makes skeptical theism such a difficult position to assess. Rowe’s argument does not depend on our ability to “limn the limits of a perfect’s being reasons”, as far as I can tell. Rowe certainly does not assume we can, since he doesn’t think there is a perfect being with purposes to limn. The failure to observe justifying goods constitutes evidence against a (certain sort of) perfect being if it is true that we in fact should expect to observe such goods. It does not depend on what we would expect were God to exist. Rowe can admit that what we would expect were God to exist is to fail to observe any justifying good for evil whatsoever. He can concede that we would not limn the purposes of perfection. Yeah, yeah, all good. All of that is consistent with it being in fact true that we should expect to observe the goods that justify such evils. And the reasons we should expect to observe such goods is that we have no God-independent reason to think that the realm of value is epistemically opaque to us (pace Bergmann).

    December 27, 2010 — 9:15
  • Ted Poston

    Mike, very interesting comment. Rowe’s inference does depend on a certain probability being rather high, i.e., the probability that no good at all justifies God in permitting E1 & E2 given that no good we know of justifies God in permitting E1 & E2. It’s hard for me to think of probability in this context without thinking of what kind of modal knowledge we have of God’s reasons. If we have sufficient reason to be skeptical of our ability to exhaust God’s reasons then we have sufficient reason for thinking that the aforementioned probability is not very high. (I think it’s a probability we just can’t evaluate). This sort of skepticism doesn’t depend on whether or not there’s a God. Rather it flows from our consideration of the nature of a perfect being and what kind of epistemic access we’d have to a perfect beings reasons were there actually one.

    December 27, 2010 — 10:04
  • Mike Almeida

    This sort of skepticism doesn’t depend on whether or not there’s a God. Rather it flows from our consideration of the nature of a perfect being and what kind of epistemic access we’d have to a perfect beings reasons were there actually one.
    This is a subtle point. It does not depend on what access we would have were there a perfect being. It depends on what access we do have if there is a perfect being. This distinction is muddled all the time in discussions of skeptical theism, and it’s a distinction that Wyktsra confuses as he moves back and forth between conditional probabilities and probabilified subjunctive conditionals. The difference is crucial. It can be true that the observation of no goods is strong evidence against the existence of God even if it is true that we would not observe any such goods were there a God.
    Let me illustrate, replacing ‘perfect beings’ with ‘the Elect’. Here’s what we know about the Elect: all of the Elect lead godly lives and go to heaven, and they would go to heaven anyway were they not to lead godly lives. Suppose all of the Reprobates lead ungodly lives and go to hell and would go to hell anyway were they to lead godly lives. Suppose our population is roughly evenly divided between the Elect and the Reprobates. Suppose we pick someone S at random. Let H be the hypothesis that S led a godly life, and let P be the proposition that S went to heaven. P is evidence for H, raising the probability of H from about .5 to 1.
    P(H | P) !> P(H)
    But wait! Why would P be evidence for H? After all, given what we know about the Elect, it is true that P = S would go to heaven anyway, were it true that ~H = S did not lead a godly life. Compare the skeptical theist.
    The observation of no God-justifying goods for existing evils is not evidence that there are no goods since we would observe no God-justifying goods if there were such goods. The fact that we would not observe God-justifying goods were there such (supposing this is something we know from the nature of perfect beings and their reasons) does not entail (it’s not even close to entailing) that the observation of no such goods is not evidence that there aren’t any. The fact that S goes to heaven is decisive evidence that S led a godly life even though it’s true that S would go to heaven anyway had S not led a godly life.
    The epistemic interplay between conditional probablities and probabilified subjunctive conditionals is something ST’s ought be much more careful about. On the other hand, I think much of the appeal to ST trades on subtle mistakes just here.

    December 27, 2010 — 11:54
  • Ted Poston

    I like your example, but one worry here is where the relevant information concerning probabilities and counterfactuals come from. In your example, you stipulate that we have two groups, E and R. And for the E’s it’s true that E iff G&H and for the R’s its true that R iff ~G&~H. Then we’re given the additional counterfactual information: G is modally irrelevant to H. For your example, we assume this information to illustrate how learning that S leads a godly life raises the probability that S went to heaven, if though it’s true that were S not to lead a godly life, S would still go to heaven. (I’ve flipped the propositions in your example; where you have P(H|P) !> P(H), I have P(P|H) !> P(P). I don’t think it matters).
    I have two concerns about the aptness of this model for understanding skeptical theism. First, the model assumes that H is evidence for P only given the assumption that x is a H iff x is a P. I don’t think this is an adequate condition for evidence; it seems like they’ll be lots of uninteresting cases of true biconditionals where the oneside is evidentially irrelevant to the other (grue example). Second, related to this, it looks like the notion of probability you are using in the example is an objective notion, tracking actualities rather that possibilities. I’m interested in epistemic probability which needs to do more than track actualities. I’m thinking of things from an internalist perspective trying to narrow modal space on the basis of our evidence. Am I in a God world or a ~God world? What considerations can I bring to bear on this question? If the considerations are apriori, I’ll have to use some modal knowledge, no?

    December 27, 2010 — 13:35
  • Mike Almeida

    First, the model assumes that H is evidence for P only given the assumption that x is a H iff x is a P. I don’t think this is an adequate condition for evidence; it seems like they’ll be lots of uninteresting cases of true biconditionals where the oneside is evidentially irrelevant to the other (grue example).
    I’m not sure I follow you here. In the example, we know that all and only the godly go to heaven. If you learn that someone goes to heaven, then your probability for their being godly is 1. The moral of the story is that what matters to conditional probabilities are actual stochastic relations between propositions. Counterfactual stochastic relations are irrelevant. So, for instance, the following two conditionals, (1) and (2), will not in general have the same probability, though they are based on the same evidence. They ‘say’ different things.
    1. P(~O | G)
    The probability that I observe no God-justifying goods given that God does exist.
    2. P(G []-> ~O)
    The probability that I would observe no God-justifying goods were God to exist.
    The value of (1) is very low. The fact is that I do observe no God-justifying goods. If we add the evidence that God exists it does not change the fact that I observe no God-justifying goods. But the value of (2) might be quite high. At least, that’s what the STers tell us.
    What is vital is not to read (1) as though it were (2). What we conditionalize in (1) are suppositions of actual fact, such as the supposition that God exists. That supposition does not bring with it all of the counterfactual changes that obtain in the closest non-actual worlds in which it is true. The only ‘changes’ it brings are those determined by the stochastic relations that actually obtain between various propositions/states of affairs. These stochastic relations depend in part on what we already know. To put the point crudely, the conditional probability that there exists the amount and depth of evil we observe in the actual world given that God does exist is certainty or 1. We know there is vast amount of evil, and learning that God exists won’t make that knowledge disappear. But it could well be true that in the closest worlds where God exists the depth and extent of evil is minimal.
    Is this useful?

    December 27, 2010 — 15:19
  • Trent Dougherty

    ” The problematic claim is that these are the best goods *possible*. ”
    Problematic for whom? For Mike? That’s what I said: He can’t say that. Not for me: I claim that we can know by reflection that there’s nothing better than mercy, etc. And since we can know such things, we can know that all the best worlds contain suffering.

    December 27, 2010 — 21:36
  • Ted Poston

    Problematic pour vous, mon ami. “Show-me-the-argument” for the claim that ‘Necessarily, mercy is the best good.’ 😉 There are other goods that are at least as good as mercy–the beatific vision; solidarity; faithfulness; trust. Perhaps, that are other goods that we don’t know about that are at least as good.

    December 28, 2010 — 11:22
  • David Warwick

    “At most, it requires the existence of apparently unjustified evils.”
    ‘Justified’ is an interesting word. Is God justifying all his actions? Where and to whom? In advance of the action, or after the event? If the answer’s yes, skeptical theism is false – each of God’s actions can be justified to a being that’s not God. If the answer’s no, or ‘He’s God, what he says goes’ or ‘he doesn’t, but he could’ or ‘he doesn’t have to, but he does’ then there are unjustified good and evil actions in the universe.

    December 28, 2010 — 12:00
  • Ted Poston

    Mike, This is useful. But help me understand why (1) is low. By Bayes’ Theorem P(~O|G)=P(G|~O)*P(~O)/P(G). Expanding the denominator gives us this ratio: P(G|~O)*P(~O)+P(G|O)*P(O). Assuming P(~O)=1, we get this equation: P(~O|G)=P(G|~O)/P(G|~O)=1. I have some additional thoughts about your post, but I’ll await some clarification on this first.

    December 28, 2010 — 12:06
  • Mike Almeida

    The comments seemed to have been closed on Trent’s post. I’ll provide a link to Almeida and Oppy AJP here http://www.informaworld.com/smpp/ftinterface~content=a713659758~fulltext=713240930

    December 28, 2010 — 13:20
  • Mike Almeida

    Ted,
    The point in (1) was that P(~O|G) = P(~O). I was reading ~O as it is not the case that I fail to observe such goods (that value would be zero). But you’re right, this is not what’s on paper, and it doesn’t conflict much with (2). The better example is the second one, with vast amounts of evil.

    To put the point crudely, the conditional probability that there exists the amount and depth of evil we observe in the actual world given that God does exist is certainty or 1. We know there is vast amount of evil, and learning that God exists won’t make that knowledge disappear. But it could well be true that in the closest worlds where God exists the depth and extent of evil is minimal.

    Here we get a serious inconsistency in value.
    1. P(E | G) = P(E) = 1
    2. P(G []-> E) much less than 1

    December 28, 2010 — 13:36
  • Mike Almeida

    Is God justifying all his actions? Where and to whom? In advance of the action, or after the event? If the answer’s yes, skeptical theism is false
    The skeptical theist does not deny that God is justified in permitting evils he might have prevented. He denies that we could know what justifies God in permitting them. So the idea, as Alston puts it, is this. When we try hard to determine what justifies God in permitting the evils we observe we are in a position analogous to someone trying hard to observe, from far across the room, whether the phone bill is on his desk. We can’t see the bill on the distant desk, but it is a mistake to conclude from that failure that the bill isn’t there. Similarly, we can’t determine what might justify God in allowing certain terrible evils, but it’s a mistake to conclude that there is no justification. We are just not in a position to know.

    December 28, 2010 — 13:47
  • Trent Dougherty

    How on Earth is any of that a problem for me, those other things are part of the “etc.” I don’t think a human could have the beatific vision w/o mercy, solidarity, etc. That’s part of the point of Purgatory. It’s all part of a package, that’s why we can’t have what we think we want: pleasure w/o pain.

    December 28, 2010 — 16:21
  • Mike Almeida

    Ted,
    One other quick way to make the distinction using counterparts of (1) and (2) and trying not to blog-rush.
    1. P(~O | G)
    The probability that I observe no God-justifying goods given that God does exist.
    2. P(G []-> ~O)
    The probability that I would observe no God-justifying goods were God to exist.
    The fact is that I do observe no God-justifying goods (i.e. P(~O) = 1), so the value of (1) just is the value of P(~O) = 1. If we add the evidence that God exists it does not change the fact that I observe no God-justifying goods. But the value of (2) might be quite low. At least, that’s what the non-skeptics tell us (though this might require non-centering). Again, just another case where the values of the two conditionals pull apart.

    December 28, 2010 — 16:30
  • Ted Poston

    I’m fine with all that. But I took you to be expressing the thought that skeptical theism was inconsistent with a Swinburean theodicy. I claim that the two are consistent since a higher-order good response doesn’t require the claim that “Necessarily, those higher goods are the best.”

    December 28, 2010 — 17:28
  • Ted Poston

    I think the difference you want between (1) & (2) does require non-centering for (2). Arguably, what’s true to make the probability of (1) high should be held constant when considering (2).

    December 28, 2010 — 17:33
  • Ted:
    A minor technical point.
    In (5)-(8), you want something slightly stronger than approximate equality. An atheist who is convinced by some argument for materialism will agree that P(G|e)≈P(G) because she will think both quantities are approximately equal to 0. Likewise, we can imagine a theist who thinks that P(G) is extremely close to 1 may hold that although e increases the probability of ~G by several orders of magnitude, nonetheless P(G|e) is still approximately 1, and hence P(G|e)≈P(G). Neither of these people should count as a sceptical theist.
    You need some other way way of comparing probabilities than approximate equality. Here’s how I like to do it: Say that a~b iff a/b is approximately 1 and (1-a)/(1-b) is also approximately equal to 1. Then in (5) and on, say something like that P(G|e)~P(G).
    That still doesn’t handle the following problem: a theist or an atheist who thinks that P(G) is 1 or 0, respectively, will agree with all of (1)-(8), but should not automatically count as a sceptical theist. Perhaps we can still define some ~ relation to take account of infinitesimal differences, and then use my ~-modified (5)-(8).

    December 28, 2010 — 18:33
  • Mike Almeida

    Arguably, what’s true to make the probability of (1) high should be held constant when considering (2).
    That’s not obvious. That P(~O) = P(~O|G)= 1, might entail that P(~O|G) = P(G|~O) = 1. I’d have to think about it, but quickly qiven the assumptions, P(~O|G)=
    P(G|~O)/P(G). But then P(~O|G)= P(~O|G)/P(G) iff. P(G) = 1. But then P(G|~O) = P(G) P(~O|G)/P(~O). So, P(G|~O) = P(~O|G) = 1. If that’s right, then I don’t think (2) is true without non-centering. That inference quickly came to mind, but I’m hoping that I missed something. It’s too quick, I must have overlooked something.

    December 28, 2010 — 19:18
  • David Warwick

    “He denies that we could know what justifies God in permitting them.”
    ‘We’ has to be ‘anyone that isn’t God’, though, I think, rather than ‘human beings alive now’ or whatever. Obviously intellectual, knowledge and linguistic factors limit us in ways God isn’t limited, but that would apply to the most advanced aliens (or angels), too.
    The point I’m trying to make is that ‘justification’ is a process in which one person/being explains what their reasons were to another. If God *can’t* explain his reasons, because of the limits of those who are not God, these actions are, by definition, unjustifiable, unjustified and unreasonable.

    December 29, 2010 — 6:54
  • Trent Dougherty

    Plus, we have revealed to us things like “mercy triumphs over judgement” and “the greatest of these is love” and “these are the fruits of the spirit” and “set your mind on these things” and the beatitudes and the examples from Israel. All I need to make ST1 false is that these things *give us a reason to think that* the goods we are aware of are *sufficiently similar* to any *relevant* goods there are. This is based on a reasonable reading of “representative”. I don’t think the strict reading of it is going to get anywhere.

    December 29, 2010 — 10:17
  • Ted Poston

    Thanks Alex. In effect, I should use a standard Bayesian measure of confirmation. Your example is basically the ratio measure.

    December 29, 2010 — 11:37
  • Ted Poston

    Let’s just go back to the core intuition. I take it that on some occasions the fact we wouldn’t see a reason if there were one makes it likely that if this is a reason we don’t see it. Alston’s chess analogy supports this. Suppose a novice a novice chess player observing a match between Kasporov and Spassky. At a crucial point in the match Kasporov moves his bishop to E7 capturing Spassky’s E7 pawn. Why, thinks the novice, did Kasporov do this when it will result in the immediate capture of the bishop and also deploy Spassky’s Queen? The novice simply can’t see a reason. Should the novice conclude that there is no reason? Of course not. It would be silly to suppose that the novice is in a position to discern every reason Kasporov might have for making this move. The novice should withhold judgment on whether Kasparov has an adequate reason.

    December 29, 2010 — 11:52
  • Mike Almeida

    The point I’m trying to make is that ‘justification’ is a process in which one person/being explains what their reasons were to another.
    This is not how ‘justification’ is used in this context. My being justified in doing x entails nothing about my explaining (or being able to explain) why I’m permitted to x to anyone.
    If God *can’t* explain his reasons, because of the limits of those who are not God, these actions are, by definition, unjustifiable, unjustified and unreasonable.
    Maybe there is some notion of justification on which something like this is true (but I suspect there isn’t, since it would have sufficently disabled persons unjustified in doing anything) but it’s not the notion in play. Being unable to explain your reasons for being permitted to x does not entail that you are not justified in doing x. Indeed, having no clue what the reasons are that permit you to do x does not entail that you’re not justified in doing x.

    December 29, 2010 — 11:53
  • Ted Poston

    That’s cool, but even though ST1 is false (on this reading of ‘representativeness’) there’s a close cousin to ST1 that’s true and sufficient to undermine Rowe’s noseeum inference. I.e., for all we know there are goods that are sufficiently similar to mercy, etc. in the respect of justifying some evils. I think the importance of this lies in the possibility of recognizing that *if* there are problematic evils, evils for which nothing we know of justifies God in permitting these, it doesn’t follow there is nothing that justifies God in permitting these evils. It seems to me like the view you’re opting for denies the antecedent. That’s fine with me since I just want to point out the consistency of skeptical theism with some other defenses.

    December 29, 2010 — 12:06
  • Mike Almeida

    Let’s just go back to the core intuition. I take it that on some occasions the fact we wouldn’t see a reason if there were one makes it likely that if this is a reason we don’t see it. Alston’s chess analogy supports this.
    Undoubtedly, on some occasions this is true. But it is not true in the skeptical theist’s case. Suppose for a moment that we live in an atheistic world W. Maybe God is contingent and inhabits other worlds. Maybe God necessarily does not exist. Whatever the case, the skeptical theist’s argument should not depend on assuming that God exists or that he likely exists. Some stipulated facts about W: (1) there are things that are morally right and morally wrong in W, and (2) there exists only the natural world, so the origin of morality is not supernatural. Morality finds its origin, lets suppose, in collective self-interest. Morality has some contractual basis.
    Now in our world W, I think an atheist can agree with the following counterfactual.
    1. P(if there were a God-given reason for the evils that exist, then I would not observe it) ≅ 1.
    But (1) does not entail (2).
    2. P(I do not observe the point of evils/ there is a point to evils) ≅ 1.
    Yes, I wouldn’t observe a God-given reason to evils, if there were one, since in God-worlds such reasons are opaque (I’m conceding). But we do not know that we are in a theistic world. We might be in a social contract world. If so, I’d very likely know the point of evils if there is one.
    So how does this fit with the Alston case? The Alston case begs the question about the source of moral opacity (or, analogously, chess opacity). My failure to see the point of the move is not evidence that there is no point, on the assumption that the players are chess gods. But our entire dispute concerns whether they are chess gods. In worlds where they are chess hacks, my failure to see the point is excellent evidence that there is no point. Though I can concede (1′),
    1′. P(if there were a chess-god reason for the apparently terrible move, then I would not know about it) ≅ 1.
    But if there is an apparently bad move in this chess hack world, then I do know about it.
    [Incidentally, fwiw, P(~O|G) = P(G|~O) = 1 in the last post.]

    December 29, 2010 — 12:34
  • Ted Poston

    What do you think about this case? You’ll observe one of two chess games but you don’t know which. And let’s, say, you’re a novice. One game is played between hacks and the other between grandmasters. You enter mid-game and observe a move for which you can’t see any justifying reason and you can see some reasons not to. Does that move give you reason to think you’re observing the game between hacks?

    December 29, 2010 — 13:00
  • Mike Almeida

    But now we’re addressing a different and much more complicated question. The problem is that God’s existence determines what evidence I have (for or against his existence) and the evidence I have (for and against his existence) determines whether God exists. That is, the evidence that I have for my priors depends on whether God exists and God’s existence depends on the evidence I have for my priors. It is not rational for the atheist to allow the theist to set priors in this argument that don’t include a long, long history of horrendous evil (it is as irrational as allowing the atheist to include in the priors the evidence from that long history and not allowing in facts about the goodness in the world). Really, the argument is over what the priors should be. Rowe puts the priors at .5. But the evidence used to set the priors depends entirely on whether God exists. So I have no idea how he arrives at .5. You seem to assume .5 as well, in the story above.

    December 29, 2010 — 13:59
  • Mike Almeida

    I’m wondering if the skeptical theist’s argument can be generated without reasoning with priors. I’ve asked Branden Fitelson for some leads on (quasi) bayesianism without prior probabilities.

    December 29, 2010 — 16:51
  • David Warwick

    “It is not rational for the atheist to allow the theist to set priors in this argument that don’t include a long, long history of horrendous evil”
    It’s perhaps not wise, but it could be rational.
    Atheism is falsifiable, at least at an individual level, in that if an atheist concedes one example of a god doing something anywhere, he’s conceded his atheism (on the simplest definition of atheism, ‘the understanding that there are and have never been any gods’).
    In discussions here, a number of theists have offered similar ‘magic bullet’ positions – ‘if an infallible Papal statement is false, Catholicism is false, if Catholicism is false, theism is false’. Biblical literalists would say the same about the Bible. Other positions take the accuracy of prophecy in the Bible as proof.
    The problem of evil represents an *apparent* ‘magic bullet’ position. What ought to be possible is a clear statement from a theist. Something like:
    ‘I do not believe that God would create something solely dedicated to inflicting disproportionate suffering and selfishness. If such a thing was found, I would understand there to be no God’.
    In which case, you’d be able to cite the larval stage of the Australian paralysis tick, a creature that permanently paralyses its large host (including koalas, dogs and even humans) to draw only a few drops of blood, even though – like other ticks – it doesn’t need to, because it has strong jaws that allow it to hang on long enough to feed. It then drops to the ground and lays 3000 eggs.
    In practice, as with the many Papal declarations, Biblical statements and prophecies that have been unambiguously demonstrated to be false, it’s always possible to find wriggle room.
    So the question is whether we can derive a ‘magic bullet’ position about the existence of God from the problem of evil.
    I don’t think we can. The theist position that ‘good’ is defined by what God does means that even if every other being in the universe felt it was evil, God would be right that it was good (which I think is an absurd moral position to take, but a perfectly sound logical position). The theist assertion that we can only assess such things eternally, not locally, also makes it possible to perpetually defer any assessment.
    What most theistic positions seem to assume is that ‘good’ can be measured in the same way ‘temperature’ can be. I don’t think this is the case, and I certainly don’t see any evidence for it. I don’t think that if I’m wrong, it necessarily has any implications for theism (it would have plenty for morality, of course, it would make ethics more like accountancy).
    If we could demonstrate that ‘good’ *can’t* be objective measured, I think that would represent a magic bullet against monotheism, though.

    December 30, 2010 — 7:11
  • Ted Poston

    Mike, But the scenario I gave gets at the same question we’ve been investigating. Think about it the way van Inwagen sets up the debate over the problem of evil. Should the problem of evil move an ideal agnostic to atheism? If the problem of evil can’t do that then it’s a philosophical failure. If you accept the intuition that I’m going for in the second chess case then skeptical theism shows that the problem of ‘problematic’ evils is a philosophical failure.
    When you say that “the evidence that I have for my priors depends on whether God exists and God’s existence depends on the evidence I have for my priors” I really don’t know what you’re getting it. On one characterization of priors it’s all about assigning probabilities given complete ignorance of the character of the world. (Contra orthodox Bayesian, I don’t think this requires assigning 1 to every necessary truth) On another characterization of priors, it’s all about degrees of belief. You seem to be saying that your prior on God should be n, if God exists and m, not equal to n, if God doesn’t exist. But I think there’s a characterization of a subject’s evidence on which the probability of God’s existence is n whether or not God exists.
    Re “it’s all about the priors.” I think this is a huge concession to the theist. If it’s really all about the priors then the problem of ‘problematic’ evils is a philosophical failure. It shouldn’t move the ideal agnostic one bit.

    December 30, 2010 — 8:12
  • Mike Almeida

    The theist position that ‘good’ is defined by what God does means that even if every other being in the universe felt it was evil, God would be right that it was good (which I think is an absurd moral position to take, but a perfectly sound logical position
    I don’t think that’s any theist’s position. What God does certainly does not ‘define’ what’s good.

    December 30, 2010 — 8:40
  • Mike Almeida

    When you say that “the evidence that I have for my priors depends on whether God exists and God’s existence depends on the evidence I have for my priors” I really don’t know what you’re getting it.
    Ted,
    What we were talking about initially (what I was talking about, anyway) was the relation between values for probabilified subjunctives and values for conditional probabilities. I spent some time worrying about these since they’re conflated all of the time in discussions of skeptical theism (compare the gibberish in CORNEA discussions, for instance, which just runs together all sorts of conditionals). So this is what was under investigation initially.
    The case you provided is, I agree, the typically way to set up the problem. Rowe sets it up that way and so does van Inwagen. But I don’t think either one puts a lot of thought into this way of setting it up. The entire discussion depends on how the priors are determined. You write.
    On one characterization of priors it’s all about assigning probabilities given complete ignorance of the character of the world.
    I have no idea how one could set the prior probability for, say, God existing, if one were assigning values from ignorance. That’s a near certain way to generate Bertrand’s Paradox.
    On another characterization of priors, it’s all about degrees of belief.
    Surely this is a caricature. One’s degree of belief is itself open to rational criticism, so its more about getting the right degree of belief. And I think this is the problem.
    Let me try to underscore the problem that I’m focusing on. Some focus on whether we should be skeptical at all about value, whether or not God exists. That’s a nice thing to worry about, but it is not what I’m worried about. I want to concede (for the moment) the skeptical theist’s skeptical claim. When you make this concession you are conceding that God’s existence makes an epistemological difference. Let O be an observation of no good justifying known evil. The evidential weight of O against God’s existence depends on the probability that God exists. The more likely it is that God exists, the more likely it is that I’m epistemically limited in this respect.
    So how do I assess the degree of belief in God in my priors? My priors too are conditional probabilities. They are conditional on all of the evidence I have prior to the current evidence (say, Rowe’s e1 and e2). My current evidence is just going to update my beliefs via conditionalization on it. But the extent of my evidence for my priors depends on what my epistemic position was. And my epistemic position depends on whether God exists. I see no way to escape this circle. This is the problem I’m worried about.

    December 30, 2010 — 9:12
  • David Warwick

    “I don’t think that’s any theist’s position. What God does certainly does not ‘define’ what’s good.”
    Define is too precise a word, and I admit that the whole areas of divine purposes and being able to precisely quantify ‘good’ and ‘evil’ are alien territory to me, bearing, as far as I can see, no relationship whatsoever to the observable world.
    It is a widespread belief that God’s ‘divine purpose’ and ‘good’ are synonymous, to the point that any apparently ‘evil’ divine actions must be for the greater good, and any apparently ‘good’ non-divine actions that go against the divine purpose can not be good.

    December 30, 2010 — 10:02
  • Ted Poston

    “What we were talking about initially (what I was talking about, anyway) was the relation between values for probabilified subjunctives and values for conditional probabilities. I spent some time worrying about these since they’re conflated all of the time in discussions of skeptical theism (compare the gibberish in CORNEA discussions, for instance, which just runs together all sorts of conditionals). So this is what was under investigation initially.”
    Yes, I agree. But my point is that this difference doesn’t amount to anything in the specific evidential situation I was interested in (where it’s not known that God exists and not known that God doesn’t exist).
    I have considerable sympathy with the rest of what you say. I’m attracted to coherentism, after all. But I think it’s worth the price of the debate to get this conclusion: the problem of ‘problematic’ evils is a philosophical failure. It seems we agree about that, no?

    December 30, 2010 — 10:15
  • Mike Almeida

    But I think it’s worth the price of the debate to get this conclusion: the problem of ‘problematic’ evils is a philosophical failure. It seems we agree about that, no?
    I think I might put it this way, and this may not be to disagree. I just don’t know what the problem is supposed to be. I’m looking for the problem. It feels like there might be one. But I find it hard to say what it’s supposed to be. So whereas you say that’s because there isn’t any problem. I say, maybe there’s one, but I don’t know what it is.
    They used to say about Goodman that he was like a bulldog when he got his teeth into a problem. He would not let the thing go until he shook out a solution or new problem. So it is definitely worth banging the problem around.

    December 30, 2010 — 10:46
  • Mike Almeida

    I’m prepared not to nitpick about it. What interesting thing follows from this?

    December 30, 2010 — 11:35
  • Mike Almeida

    It is a widespread belief that God’s ‘divine purpose’ and ‘good’ are synonymous, to the point that any apparently ‘evil’ divine actions must be for the greater good, and any apparently ‘good’ non-divine actions that go against the divine purpose can not be good.
    I’m prepared not to nitpick about it. What interesting thing follows from this?

    December 30, 2010 — 11:37
  • Ted:
    The ratio measure is P(H|E)/P(H). That becomes unhelpful when P(H) is very close to 1. What I was suggesting as more helpful here is max(P(H|E)/P(H),P(~H)/P(~H|E)). I don’t know if that’s among the standard measures of confirmation, since I know next to nothing about standard measures of confirmation.

    December 30, 2010 — 13:56
  • Ted Poston

    Alex,
    Thanks for the clarification. How, though, does that avoid problem of extreme values to P(H). If P(H)=1 then P(H|E)=1 and max(P(H|E)/P(H),P(~H)/P(~H|E))=1, no?

    December 31, 2010 — 7:59
  • Mike Almeida

    As P(H) becomes ‘very close to 1’, E (the very same evidence) makes less and less difference to H until it makes no difference at all (where P(H) = 1). There is effectively a continuum of ‘problems of old evidence’ reflecting the priors for H. I see why you would not want to use the typical measure of confirmation as H gets closer to 1. But I’m not sure why you would not want to use the ratio measure.

    December 31, 2010 — 8:40
  • Mike:
    The problem of old evidence is, I take it, where P(E) is close to 1, not where P(H) is close to 1. Why not ratio measure? Because if P(H) is close to 1, the ratio measure says that nothing is significant evidence. But it is sometimes very important to keep track of evidence for stuff near 1. I don’t want to work out the numbers, but a change of probability from 0.99 to, say, 0.999 can be quite significant, because it makes the resulting belief more immune to counter-evidence.
    Ted:
    I’d want to use infinitesimals to model degrees of confidence for things one is certain of. So, instead of P(H)=1, I’d make it be P(H)=1-e, where e is a positive infinitesimal. I am certain that 1+1=2 and that 2+2=4, but I am more certain that 1+1=2.

    December 31, 2010 — 10:43
  • Mike Almeida

    The problem of old evidence is, I take it, where P(E) is close to 1, not where P(H) is close to 1.
    It’s the same problem (well, more or less). If T is your theory and E is your evidence, then since P(E) = 1 (old evidence), P(E|T) = 1, but then P(T|E) = P(E). E confirms T only if P(E|T) is greater than P(E). But if P(T) = 1 we have the same problem. This time T is old evidence for E, so P(E|T) = P(E) and E again does not confirm T.

    December 31, 2010 — 14:01
  • David Warwick

    “What interesting thing follows from this?”
    I think you, Ted and I are violently agreeing. The problem with the problem of evil is that, as formulated, it says nothing about the existence of God. But, intuitively, it ought to.
    Given an omnibenevolent God, there’s no problem – while it might not look like it to us, ‘everything happens for the best’. Given atheism, there’s no problem explaining why God allows evil.
    Ted,
    “Should the problem of evil move an ideal agnostic to
    atheism?”
    I don’t know about an ‘ideal’ agnostic. We are all ‘ideal agnostics’ to some extent. The atheist can’t rule out the existence of some god somewhere (however many specific claims for specific gods he can demonstrate to be false); no theist can ultimately rule out their god being the product of delusion (even – especially? – those who claim direct divine revelation).
    The universe is not agnostic. There was, is or will be at least one god involved in it at some point (World T) or not (World A). Stretch the definition of ‘god’ and ‘involved’ as you will, that holds.
    I offer this: an atheistic universe (A) would operate differently to a theistic one (T). It makes a difference if there’s a God or not.
    We live in either A or T. A and T are different. The issue becomes one of what that difference is, and whether we can infer which of the alternatives we live in.
    An honest agnostic, I think, could have a working idea of what the difference between A and T might be – and what others believe and claim – and make empirical observations. An individual can reach a decision, even if it ends up as an aesthetic rather than scientific one.
    Theistic models of how the universe was formed and how it operates generally tend to stress ‘good and evil’ and ‘purpose’ more than atheistic models. That does seem like it ought to be a useful place to look when we’re assessing whether we’re in A or T.

    January 1, 2011 — 8:36
  • For the record, I now prefer the standard likelihood ratio to my simple measure. (But asymptotically, they behave the same way.)

    January 1, 2011 — 13:15