Abstract
Skeptical theists argue that no seemingly unjustified evil (SUE) could ever lower the probability of God's existence at all. Why? Because God might have justifying reasons for allowing such evils (JuffREs) that are undetectable. However, skeptical theists are unclear regarding whether or not God's existence is relevant to the existence of JuffREs, and whether or not God's existence is relevant to their detectability. But I will argue that, no matter how the skeptical theist answers these questions, it is undeniable that the skeptical theist is wrong; SUEs lower the probability of God's existence. To establish this, I will consider the four scenarios regarding the relevance of God's existence to the existence and detectability of JuffREs, and show that in each—after we establish our initial probabilities, and then update them given the evidence of a SUE—the probability of God's existence drops.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
Rowe seems to accept a traditional view of God’s knowledge and providence, one akin to Molinism. God has full knowledge of the future and what every free person would do in every circumstance. This, I believe, is the view most generous to skeptical theism. If, for example, one accepts Open Theism, and thus maintains that God does not know the future and allows libertarian freedom to have full rein, it would be much harder for God to have reasons to allow evil that are beyond human comprehension. Without full knowledge of the consequences of an evil, or how humans will react to that evil, it would be much more difficult to identify greater goods that might make horrendous evils like E1 and E2 justified. If God cannot know what kind of consequences such evils will have, it seems that he should just determine whether to allow them based on their own merits—and clearly, by their own merits, they should not be allowed. Open Theism does make a free will theodicy for the evidential problem more easily defensible, but the free will theodicy does not help the skeptical theist defend their position (and it cannot deal with natural evil, such as E1). So that we are considering the strongest case for skeptical theism we can, I will be assuming a Molinistic view of God’s providence and knowledge. It should be noted, however, that granting Molinism does not allow the theist to solve the problem of evil without skeptical theism. It does make available more possible reasons to allow evil—God may have allowed Hitler to invade Poland because he knew, given how all possible free creates would react, that it would lead to a greater good. (This is how Plantinga uses Molinism in his free will solution to the logical problem of moral evil.) But it does not effectively deal with the evidential problem of evil, or either version of the problem of natural evil. Skeptical theists admit that, even if Molinism is true, some evils seem unjustified; they simply claim that, because God may have reasons to allow evil that are beyond our ken, such evil cannot reduce the probability of his existence. Molinism gives God more knowledge and thus perhaps raises the probability that he has reasons we can’t comprehend, but that does not solve the evidential problem, nor does it affect the logic of the argument I will present about why skeptical theism fails—why the fact that God may have reasons we can’t understand does not prevent seemingly unjustified evil from reducing the probability of God’s existence.
Peter van Inwagen has argued that God could allow evil that, he suggests, does not have such a justifying good. For example, in his 1998 article ‘The Place of Chance in a World Sustained by God,’ he argues that God may allow evils that are ‘chancy’ and not a part of anyone’s plan. According to van Inwagen, such evils are compatible with God’s existence because God might permit them in the name of upholding some morally adequate ‘general policy,’ such as not ever interfering with human free will. Although I am hesitant to consider the general policy of not interfering with human free will morally adequate, following the lead of William Alston (1996, p. 101), I would consider upholding morally adequate general policies to be in the category of greater goods specified above [that is, I would consider them to have satisfied conditions (a) and (b)]. I would thus, unlike van Inwagen, not consider evils allowed in their name gratuitous. However, I think it is worth pointing out that a deity that universally upholds general policies, such as ‘never interfere with free will,’ ‘never interfere with course of human events’ or ‘never interfere with the natural order,’ cannot be the Christian God; the Christian God is believed by most Christians, and the Biblical authors, to do all three of these things on a fairly regular basis. Such a deity is most akin to the deist God. Thus, I would not expect the skeptical theist to endorse such goods as JuffREs. Van Inwagen (2000) defends a similar point with this ‘no minimum’ argument. For more on the problems with that argument, and his chance argument, see Trakakis (2007, Chap. 12).
So, for example, an eternity of bliss in heaven cannot justify the rape and murder of an innocent five year old. Although that might be a good that outweighs the evil in question, that evil is not necessary for that good to exist. There are certainly other less evil ways for a child to die and enter into the eternal bliss of heaven.
It’s important to make clear that a JuffRE may or may not be detectable by humans.
Rowe did so in the same issue of Nous in which Plantinga’s argument was published. See Rowe (1998). It’s also worth noting that Plantinga’s argument was criticizing Rowe’s direct inference from (P) to God’s non-existence, not his defense of the inference from (P) to (Q) that I am about to introduce and about which I am primarily concerned in this article.
See, for example, Peterson and van Arragon (2004).
Trakakis (2007, p. 341).
Nick Trakakis (2007) quotes two philosophers, Alvin Plantinga and Jeff Jordan, to show that they share this misconception on page 68.
Ibid., pp. 68-69.
Ibid., p 69.
See, for example, Rowe’s defenses offered in 1979 and 1986.
After all, the theist could admit that SUEs reduce the probability of God’s existence, but still argue that there is other outweighing evidence for God’s existence that makes it more likely. But to make that argument, one would have to present the outweighing evidence, and that is outside the scope of skeptical theism and thus outside the scope of this paper.
It is important to note that setting belief in God aside, thus beginning with a prior probability for God’s existence at 50%, is something that both sides of the debate agree should be done when dealing with this issue. Both sides have done it, and neither side has objected. See Rowe (2001a) and Howard-Snyder & Bergmann (2004). Since the issue is merely whether evil will reduce the probability of God’s existence at all, it actually doesn’t matter what prior probability you assign to God’s existence. As long as it goes down, after considering the evidence, then the evidence does reduce the probability of God's existence.
Although this is not necessarily a tenet of skeptical theism—skeptical theism is, after all, simply a denial of the inference from (P) to (Q) —as theists, skeptical theists agree that God would not allow evil without a justifying reason for doing so. In fact, it is this assumption that motivates skeptical theism. If one accepted that God could allow evil that has no justifying reason (i.e., gratuitous evil), then one could grant the inference from (P) to (Q) but still argue that E1 and E2 do not decrease the probability of God’s existence. Interestingly enough, some theists [such as John Hick (1977), Michael Peterson (1982), Keith Yandell (1989), William Hasker (1992) and Peter van Inwagen (1998)] have defended this position—although it is quite controversial. Given what I will show, the skeptical theist cannot maintain that E1 and E2 do not reduce the probability of God’s existence without also embracing that God could allow gratuitous evil; but since embracing the latter is quite controversial and would make defending skeptical theism unnecessary, I will assume that the skeptical theist does not embrace it. For a refutation of the most popular arguments that God can allow gratuitous evil, and criticism of their ability to deal with Rowe’s argument, I encourage the reader to see Trakakis (2007, Chap. 12).
My argument may also establish that Rowe’s inference directly from (P) to God’s non-existence is also strong, but I am not interested in defending that conclusion specifically. I am only interested in defeating the skeptical theist’s claim that the inference from (P) to (Q) is not strong and that SUEs don’t reduce the probability of God’s existence at all.
Of course, one does not need to be a theist to endorse the skeptical theist’s argument—to deny the inference from (P) to (Q). But I would hesitate to call an atheist who denied that inference a ‘skeptical theist.’ Nevertheless, such a philosopher would not be committed to the existence of JuffREs. However, even if JuffREs don’t exist, since skeptical theism is suggesting that they may be beyond our ken, one still wonders whether God’s existence would be relevant to their existence or detectability. This still raises the questions I am about to mention, which is my only goal here.
This example is only meant to clarify the difference between a good that exists naturally and one that God himself brings about. Obviously, this is not a greater good in Rowe’s sense; if God can directly control the leaders of Israel, he can directly control anyone and establish the Nation of Israel without WWII.
It’s important to clarify what I mean by ‘detectable.’ Something is detectable if it is the kind of thing we would see (i.e., be aware of) upon close examination or careful consideration. I will set aside here the possibility of a justifying reason being detectable but going undetected. [Alston (1996, p. 109) points to this possibility.] Klaas Kraay (2007) explores the consequences of this possibility in his paper ‘Absence of Evidence and Evidence of Absence’ (it includes a nicely done chart), but exploring the issue further here would take us too far off track. After all, skeptical theists already agree that there are evils with no detectable justifying reasons; that is what motivates their argument.
One might also make some further distinctions. If our limits make them undetectable, is it because we cannot conceive of them, or because we cannot see if a conceivable outweighing good actually came to fruition? And, in either case, is our difficulty a result of the fact that we have limited information—or, is it the case that even if we had full information, we still couldn’t detect the outweighing goods. It would be interesting to see a full exploration of these issues, but I believe they are ultimately irrelevant to my conclusion, so I will ignore them for simplicity’s sake.
Although, I admit, there are some difficulties here. Recall, in order for a good to be a justifying reason to allow an evil, not only does that good have to outweigh that evil, but allowing that evil has to be the only way to accomplish that good. And it’s hard to imagine a situation in which the only way an all powerful being could bring about a particular good was by allowing some particular evil. But I will grant the possibility for argument’s sake; showing that it is possible only presents a difficulty for the skeptical theist.
I will consider the possibility that God’s existence only makes JuffREs unlikely below.
Wykstra (1984 p. 155)
In fact, if it is possible at all that he will not wear the perfect costume, our conclusion will not change. I’ll deal with the possibility of a 0% chance that he doesn’t wear the perfect costume, if he attends the party, below.
When I say ‘the relevant JuffRE,’ I am talking about the JuffRE of the SUE about to be considered. Of course, I cannot name what that JuffRE is—it is undetectable. However, even though we have not yet considered the evil in question, we can still know the probability that it will have a JuffRE given the assumptions laid down by the option in question.
In fact, as long as we assume that it is at least possible that the JuffRE is detectable, the same conclusions will follow. I will consider the impossibility of JuffRE detection below.
We could divide this hypothesis further. Even though God does not exist, and thus neither does the relevant JuffRE, it could still be true that the relevant JuffRE would be detectable if it existed. So, we could say that:
(F’) God does not exist, neither does the relevant JuffRE, but the JuffRE would be detectable = 25%
(F”) God does not exist, neither does the relevant JuffRE, but the JuffRE would be undetectable = 25%
But our inability to detect the relevant JuffRE, when it doesn’t exist, tells us nothing about whether or not it is detectable. Perhaps we don’t see it because it is undetectable, but perhaps we don’t see it because it is not there. Thus, the evidence we are considering does not falsify either hypothesis, so separating them out does not affect our conclusion. For simplicity then, and so it matches up with the example, now and later I will collapse into one hypothesis hypotheses that both (a) suggest that JuffREs don’t exist and (b) also match in regard to what they say about God’s existence.
Even if the initial probabilities are (D) = 1, (E) = 49, and (F) = 50, after updating, theism will only be 49.5% likely, and atheism will be 50.5% likely. Not a significant difference, but it still shows that an evil, and our inability to detect a JuffRE, does raise the probability that there is no JuffRE (thus supporting the inference from (P) to (Q)) and does count as evidence against theism.
In addition, these same conclusions follow if you reject our previous assumption that God’s existence, and nonexistence, guarantees the existence, and nonexistence, of JuffREs—and instead assume that God’s existence, and nonexistence, merely makes the existence of JuffREs more or less likely. This gets a bit complicated, but if God’s existence makes both the existence of a JuffRE and its detectability three times more likely, the initial probabilities would break down like this:
-
(d)
God exists, but the relevant JuffRE does not = 12.5%
-
(e)
God exists, so does the relevant JuffRE, and it is detectable = 9.375%
-
(f)
God exists, so does the relevant JuffRE, but it is undetectable = 28.125%
-
(g)
God does not exist, the relevant JuffRE does, but it is undetectable =3.125%
-
(h)
God does not exist, the relevant JuffRE does, and it is detectable = 9.375%
-
(i)
God does not exist and neither does the relevant JuffRE = 37.5%
For brevity, I will forgo the details; the reasoning is similar to what we will see in options 4. The end result is the falsification (d), (e), and (h), leaving us with the following end result:
-
(f)
God exists, so does the relevant JuffRE, but it is undetectable = 38.541666…%
-
(g)
God does not exist, the relevant JuffRE does, but it is undetectable =13.542666…%
-
(i)
God does not exist and neither does the relevant JuffRE = 47.908333…%
For the same reasons as it does in option 4, the updating shows us that (P) does provide inductive support for (Q) and that SUEs make God’s existence less likely (it drops from 50% to 39%). And, again, this holds true, no matter how much more detectable God’s existence makes JuffREs.
-
(d)
Although, in light of observations I make at the end of the next section, this is not necessarily the case. Although this move might allow the skeptical theist to criticize Rowe’s inference from (P) to (Q), it cannot successfully defend theistic belief.
Wykstra (1984 p.155)
I charitably interpret Wykstra as not meaning this.
In other words, if one part of God is comprehensible in this world, wouldn’t there be a possible world in which any other part of God, like his reasons for allowing an evil, is comprehensible, even if it is not comprehensible in this world? It seems to me that only if all aspects of God are incomprehensible in all possible worlds could one maintain that a certain part of god is not comprehensible in any possible world.
In fact, as we will see in the next sub-section, even this may not allow the skeptical theist to rescue their conclusion.
Perhaps an easier way to think about this last point is this: Our ability to detect the JuffREs of SUEs being impossible is the only way that evil doesn’t count as evidence against God’s existence. But if we haven’t a clue about how probable such an ability is, we should assign each probability an equal chance. It’s just as likely that it is impossible as it is that it is 99% possible, as it is that it is 98%…on down to 1%. So, given what we know, there is only a 1/100 chance that our ability to detect the JuffRE of a SUEs has the probability it needs to justify the skeptical theist’s conclusion. Thus, we should conclude that it does not have the probability it needs, and thus that SUEs do lower the probability of God’s existence.
Unless, of course, unlike this case, the probability of one of the conjuncts is 1.
For the skeptical theist to demand that their new God with guaranteed undetectable JuffREs be granted a probability of 50% would beg the question against their critics. This is certainly not the value that it would be assigned given the background knowledge both atheists and theists share.
This point was best made by Richard Carrier in his lecture ‘Bayes’ Theorem: Key to the Universe’ at Skepticon 4 (19 − 20 November 2011). This talk can be found at http://www.youtube.com/watch?v=HHIz-gR4xHo. Richard Carrier is an expert on Bayes, and has authored quite a few books and articles about the relevance of Bayes’ theorem to religious matters. See Carrier (2011, 2012).
The reasoning involved here is the same as the analogous point from option 1.
One might wonder here why (L) was originally included as an option. Doesn’t the impossibility of God allowing gratuitous evil rule this out before we even assign the probabilities? No. Ruling out gratuitous evil rules out the possibility of God co-existing with an evil that has no JuffRE; it does not rule out the mere possibility of God co-existing with no relevant JuffRE. There is a possible world in which God exists but the justifying reason to allow an evil does not: the possible world in which God does not allow that evil to occur. Recall, these are the probabilities that we assign before we observe the evil and realize that there is no detectable justifying reason to allow it. Before we know the evil in question has occurred, it is perfectly possible that God exists but no justifying reason to allow that evil does. In fact, we would never observe that evil if that were true. Thus, (L) must be included in our initial calculations, and the fact that we observe the evil shows us that (L) is false.
This however does not falsify Rowe’s inference from (P) to (Q). Recall, that is an inference from the fact that an evil seems to be unjustified to the fact that it is; we have not yet considered what the fact that the evil seems to be unjustified does to the probability that it is unjustified. That is the next step.
Given this option, if God’s existence made JuffRE detection impossible, we would get the following probability distributions:
-
(k)
God exists, so does the relevant JuffRE, but it is undetectable = 25%
-
(l)
God exists, but the relevant JuffRE does not = 25%
-
(m)
God does not exist, but the relevant JuffRE does, and it is detectable = 18.75%
-
(n)
God does not exist, but the relevant JuffRE does, yet it is not detectable = 6.25%
-
(o)
God does not exist and neither does the relevant JuffRE = 25%
The observation of the evil would falsify (l) leaving us with:
-
(k)
God exists, so does the relevant JuffRE, but it is undetectable = 31.25%
-
(m)
God does not exist, but the relevant JuffRE does, and it is detectable = 25%
-
(n)
God does not exist, but the relevant JuffRE does, yet it is undetectable = 12.5%
-
(o)
God does not exist and neither does the relevant JuffRE = 31.25%
Like before, the probability of there being a JuffRE rose (from 50% to 68.75), yet the probability of God’s existence dropped (from 50% to 31.25). Then, the fact that we cannot detect a JuffRE for the evil in question falsifies (m) and we get:
-
(k)
God exists, so does the relevant JuffRE, but it is undetectable = 39.58333…%
-
(n)
God does not exist, but the relevant JuffRE does, yet it is undetectable = 20.83333…%
-
(o)
God does not exist and neither does the relevant JuffRE = 39.58333…%
Again, although the mere existence of the evil initially raised it, the fact that it seems that the evil in question is unjustified lowers the probability that it is (from 68.75 to 60.4166…); thus, Rowe’s inference from (P) to (Q) is still supported. Since God’s existence makes the JuffRE impossible to detect, but God’s non-existence does not, the probability of God’s existence does rise (to 39.58333…) when we realize there is no detectable JuffRE; but it does not rise above the initial starting point of 50%. Thus, even if God’s existence made JuffRE detection impossible, the inference from (P) to (Q) would be strong, and the existence of a SUE would still lower the probability of God’s existence.
-
(k)
It’s worth noting that this option presents another problem. If our inability to detect JuffREs is merely the result of our limits, and whether there are JuffREs has nothing to do with God’s existence, then we must admit that we actually cannot tell, objectively, whether anything that seems evil actually is evil. Take the Holocaust, for example. Given this option, for all we know, there was an undetectable JuffRE that arose from it, and it actually was a (all things considered) good thing. In short, this option renders our objective evaluations of good and evil vacuous.
I would like to thank the anonymous reviewers, and William Irwin and Greg Bassham, for their extremely helpful comments on earlier drafts of this essay.
These works of Rowe (2001a and 2001b) do first appear in this format in the 2001c book. However, they also exist, as independent chapters, in the following book, on pages 193-206 and 249-254 respectively: William Rowe (2007) (edited by Nick Trakakis), William Rowe on Philosophy of Religion: Selected Writings (Hampshire, England. Ashgate Publishing).
References
Alston, W. (1996). The inductive argument from evil and the human cognitive condition. In Howard-Snyder (1996, pp. 97–125).
Carrier, R. (2011) Why I am not a Christian: Four conclusive reasons to reject the faith. CreateSpace.
Carrier, R. (2012). Proving history: Bayes's theorem and the quest for the historical Jesus. Prometheus Books.
Hasker, W. (1992). The necessity of gratuitous evil. Faith and Philosophy, 9(1), 23–44.
Hick, J. (1977). Evil and the God of love (2nd ed.). New York: HarperCollins.
Howard-Snyder, D. (Ed.). (1996). The evidential argument from evil. Bloomington and Indianapolis: Indiana University Press.
Howard-Snyder, D., & Bergmann, M. (2004). Evil does not make atheism more reasonable than theism. In M. Peterson & R. J. VanArragon (Eds.), Contemporary debates in philosophy of religion. Blackwell Publishing.
Kraay, K. J. (2007). Absence of evidence and evidence of absence. Faith and Philosophy, 24, 203–228.
Peterson, M. L. (1982). Evil and the Christian God. Grand Rapids: Baker Book House.
Peterson, M., & VanArragon, R. J. (Eds.). (2004). Contemporary debates in philosophy of religion. Blackwell Publishing.
Plantinga, A. (1998). Degenerate evidence and Rowe’s new evidential argument from evil. Nous, 32(4), 531–544.
Rowe, W. L. (1979). The problem of evil and some varieties of atheism. American Philosophical Quarterly, 16, 335–41.
Rowe, W. L. (1986). The empirical argument from evil. In R. Audi & W. Wainwright (Eds.), Rationality, religious belief, and moral commitment. Ithaca: Cornell University Press, pp. 227–47.
Rowe, W. L. (1996). The evidential argument from evil: A second look. In D. Howard-Snyder (Ed.), The evidential argument from evil. Bloomington and Indianapolis: Indiana University Press (1996, pp. 262–85).
Rowe, W. L. (1998). Reply to plantinga. Nous, 32(4), 545–552.
Rowe, W. L. (2001a). Grounds for belief aside, does evil make atheism more reasonable than theism. In Rowe (2001c, pp.124-37).
Rowe, W. L. (2001b). Reply to Howard-Snyder and Bergmann. In Rowe (2001c, pp. 155–58).
Rowe, William (ed.), (2001c). God and the Problem of Evil. (Malden, MA: Blackwell).Footnote
These works of Rowe (2001a and 2001b) do first appear in this format in the 2001c book. However, they also exist, as independent chapters, in the following book, on pages 193-206 and 249-254 respectively: William Rowe (2007) (edited by Nick Trakakis), William Rowe on Philosophy of Religion: Selected Writings (Hampshire, England. Ashgate Publishing).
Swinburne, R. (1998). Providence and the problem of evil. Oxford: Clarendon Press.
Tooley, M. (2010). The problem of evil. In E. N. Zalta (Ed), The stanford encyclopedia of philosophy (Spring 2010 Edition). http://plato.stanford.edu/archives/spr2010/entries/evil/.
Trakakis, N. (2005). The evidential problem of evil. Internet encyclopedia of philosophy, http://www.iep.utm.edu/evil-evi/.
Trakakis, N. (2007). The God beyond belief: In defence of William Rowe's evidential argument from evil. Dordrecht: Springer Publishers.
Van Inwagen, P. (1988). The place of chance in a world sustained by God. In T. V. Morris (1988). Divine and human action. Ithaca: Cornell University Press, pp. 211–35.
Van Inwagen, P. (2000). The Argument from particular horrendous evil. Proceedings of the American Catholic Philosophical Association, 74, 65–80.
Wykstra, S. J. (1984). The humean obstacle to evidential arguments from suffering: On avoiding the evils of ‘appearance’. International Journal for Philosophy of Religion, 16, 73–93.
Wykstra, S. J. (1996). Rowe’s noseeum arguments from evil. In D. Howard-Snyder (Ed.), The evidential argument from evil. Bloomington and Indianapolis: Indiana University Press, pp. 126–50.
Yandell, K. (1989). Gratuitous evil and divine existence. Religious Studies, 25(1), 15–30.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Johnson, D.K. A Refutation of Skeptical Theism. SOPHIA 52, 425–445 (2013). https://doi.org/10.1007/s11841-012-0326-0
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11841-012-0326-0