Pain and Pleasure:An Evidential for Theists Problem
PAUL DRAPER FLORIDA INTERNATIONAL UNIVERSITY

I.

THE NATURE

OF THE PROBLEM

I will argue in this paper that our knowledge about pain and pleasure creates an epistemic problem for theists. The problem is not that some proposition about pain and pleasure can be shown to be both true and logically inconsistent with theism. Rather, the problem is evidential. A statement reporting the observations and testimony upon which our knowledge about pain and pleasure is based bears a certain significant negative evidential relation to theism.' And because of this, we have a prima facie good epistemic reason to reject theism-that is, a reason that is sufficient for rejecting theism unless overridden by other reasons for not rejecting theism. By "theism" I mean the following statement: There exists an omnipotent, omniscient, and morally perfect person who created the Universe. I will use the word "God" as a title rather than as a proper name, and I will stipulate that necessary and sufficient conditions for bearing this title are that one be an omnipotent, omniscient, and morally perfect person who created the Universe. Given this (probably technical) use of the term "God," theism is the statement that God exists. Some philosophers believe that the evils we find in the world create an evidential problem for theists because theism fails to explain these evils (or most of what we know about them). (See, for example, (Hare 1968).) This position is attractive. It seems to reflect
NOUS 23 (1989) 331-350 ? 1989 by Nouis Publications 331

332

NOUS

the intuitions of a great many people who have regarded evil as an epistemic problem for theists. After all, the most common way of stating the problem of evil is to ask a why-question like "if God exists, then why is there so much evil in the world?" And such questions are either genuine or rhetorical requests for explanation. Moreover, the relevance of theodicies to this alleged problem of evil is quite clear, since a theodicy can very naturally be understood as an attempt to explain certain evils or facts about evil in terms of theism. But other philosophers who agree that theism fails to explain most of the evils we find in the world deny that this creates an epistemic problem for theists-that is, they deny that this explanatory failure is a prinz facie good reason to reject theism. This disagreement has led to a debate over how much evil, if any, theism needs to explain to avoid disconfirmation. (See, for example, (Yandell 1969a and 1969b), (Kane 1970), (Mavrodes 1970, pp. 90-111), (Ahern 1971), (Hare 1972), and (Yandell 1972).) What the members of both sides of this debate have failed to recognize is that one cannot determine what facts about evil theism needs to explain or how well it needs to explain them without considering alternatives to theism. The important question, a question that David Hume asked (1980, Part XI, pp. 74-75) but that most contemporary philosophers of religion have ignored, is whether or not any serious hypothesis that is logically inconsistent with theism explains some significant set of facts about evil or about good and evil much better than theism does. I will argue for an affirmative answer to this question. Specifically, I will compare theism to the following alternative, which I will call "the Hypothesis of Indifference" ("HI" for short): HI: neither the nature nor the condition of sentient beings on earth is the result of benevolent or malevolent actions performed by nonhuman persons. Unlike theism, HI does not entail, that supernatural beings exist and so is consistent with naturalism. But HI is also consistent with the existence of supernatural beings. What makes HI inconsistent with theism is that it entails that, if supernatural beings do exist, then no action performed by them is motivated by a direct concern for our well-being. Now let "O" stand for a statement reporting both the observations one has made of humans and animals experiencing pain or pleasure and the testimony one has encountered concerning the observations others have made of sentient beings experiencing pain or pleasure. By "pain" I mean physical or mental

PAIN AND PLEASURE

333

suffering of any sort. I will argue that the pain and pleasure in our world create an epistemic problem for theists by arguing that: C: HI explains the facts 0 reports much better than theism does.

One problem with this formulation of C is that the verb "to explain" has a number of distinct but easily confused meanings. For my purposes here, it will suffice to point out that in some instances the claim that one hypothesis explains some observation report much better than another is equivalent in meaning, or at least bears a close conceptual connection, to the claim that the truth of that observation report is much less surprising on the first hypothesis than it is on the second. Since I suspect that it is only in these instances that comparisons of explanatory power support comparisons of probability, I will reformulate C as the claim that the facts 0 reports are much more surprising on theism than they are on HI, or, more precisely, that the antecedent probability of 0 is much greater on the assumption that HI is true than on the assumption that theism is true. By the "antecedent" probability of 0, I mean O's probability, independent of (rather than temporally prior to) the observations and testimony it reports. So my reformulation of C is best expressed as follows: C: Independent of the observations and testimony 0 reports, 0 is much more probable on the assumption that HI is true than on the assumption that theism is true. For the sake of brevity, I will use P(x/y) to represent the probability of the statement x, independent the observations testimony0 reports, and of on the assumption that the statement y is true. Using this notation, I can abbreviate C in the following way: C: P(O/HI) is much greater than P(O/theism).

One last elucidatory remark about C. The probabilities employed in C are epistemic ones rather than, for example, statistical, physical, or logical probabilities.2 Thus, they can vary from person to person and from time to time, since different persons can be in different epistemic situations at the same time and the same person can be in different epistemic situations at different times. For example, suppose that six hands of poker are dealt. Then the epistemic probability that one hand includes four aces will be different for those players who inspect their hands and find no aces and those players who inspect their hands and discover one or more aces. And the epistemic probability for any of the six players that one hand includes four aces will be different before inspecting his or her hand than after inspecting it.

334

NOUS

Now suppose that I succeed in showing that C is true (relative to my own and my readers' epistemic situations). Then the truth of C is (for us) a primafacie good (epistemic) reason to believe that theism is less probable than HI. Thus, since the denial of theism is obviously entailed by HI and so is at least as probable as HI, the truth of C is a primafacie good reason to believe that theism is less probable than not. And since it is epistemically irrational to believe both that theism is true and that it is less probable than not, the truth of C is also a prima facie good reason to reject (i.e., to cease or refrain from believing) theism. In Section II, I will argue that C is true. However, my argument will depend on the assumption that theodicies do not significantly raise P(O/theism). In Section III, I will defend this assumption. And in Section IV, I will discuss the significance of C's truth.
II. THE BIOLOGICAL UTILITY OF PAIN AND PLEASURE

The claim that P(O/HI) is much greater than P(O/theism) is by no means obviously true. The fact that 0 reports observations and testimony about pleasure as well as pain should make this clear. So an argument for this claim is needed. I will argue that it is the biological role played by both pain and pleasure in goal-directed organic systems that renders this claim true. In order to explain precisely why this is so, I will need to introduce a concept of "biological usefulness." Though no one doubts that organic systems are goal-directed in some objective sense, it is by no means easy to provide a precise analysis of this kind of goal-directedness. As a first approximation, we may say that a system S is "goal-directed" just in case for some property G that S has exhibited or will exhibit, a broad range of potential environmental changes are such that: (i) if they occurred at a time when S is exhibiting G and no compensating changes took place in the parts of S, then S would cease to exhibit G and never exhibit G again, and (ii) if they occurred at a time when S is exhibiting G, then compensating changes would take place in the parts of S, resulting in either S's continuing to exhibit G or in S's exhibiting G once again. (Cf. (Boorse 1976) and (Ruse 1973).) Notice that to be goal-directed in this sense does not entail direction to the conscious end of some intelligent being. Notice also that the organic world is made up of complex and interdependent goaldirected systems, including ecosystems, populations of organisms, organisms, parts of organisms, parts of parts of organisms, and so on. I will call the goals to which organic systems are directed in this sense their "biological goals." And I will say that a part of

PAIN AND PLEASURE

335

some goal-directed organic system S is "biologically useful" just in case (i) it causally contributes to one of S's biological goals (or to one of the biological goals of some other goal-directed organic system of which it is a part), and (ii) its doing so is not biologically accidental. (It is in virtue of clause (ii) that, for example, a nonfatal heart attack that prevents a person from committing suicide cannot be called biologically useful.) Notice that much of the pain and pleasure in the world is biologically useful in this sense. Consider, for example, the pain my cat Hector felt when he jumped on top of a hot oven door. Hector's quick response to this pain enabled him to avoid serious injury, and he now flees whenever an oven door is opened. Hector's pain in this case, like much of the pain reported by 0, was biologically useful. For it causally contributed to two central biological goals of individual organisms, namely, survival and reproduction, and its doing so was plainly not accidental from a biological point of view. Of course, there is also much pain and pleasure in our world that is not biologically useful: for instance, masochistic pleasure and pain resulting from burns that ultimately prove fatal. (I will sometimes call this kind of pain and pleasure "biologically gratuitous.") This notion of biological utility enables me to introduce a statement logically equivalent to 0 that will help me show that C is true. Let "01," "02," and "03" stand for statements respectively reporting the facts 0 reports about: (1) moral agents experiencing pain or pleasure that we know to be biologically useful, (2) sentient beings that are not moral agents experiencing pain or pleasure that we know to be biologically useful, and (3) sentient beings experiencing pain or pleasure that we do not know to be biologically useful. Since 0 is obviously logically equivalent to the conjunction of 01, 02, and 03, it follows that, for any hypothesis h: P(0/h) = P(01 & 02 & 03/h).

But the following theorem of the mathematical calculus of probability holds for epistemic probability: P(01 & 02 & 03/h) 01 & 02).3 Thus, C is true-P(0/HI) in case: A: P(01/HI) x P(02/HI = P(01/h) x P(02/h & 01) x P(03/h &

is much greater than P(0/theism)-just & 01) x P(03/HI & 01 & 02)

336

NOUS

is much greater than B: P(01/theism) 02). x P(02/theism & 01) x P(03/theism & 01 &

I will argue that A is much greater than B by arguing that each of the multiplicands of A is either greater or much greater than the corresponding multiplicand of B. As I will explain in Section III, my arguments will assume that theodicies do not significantly raise P(O/theism). Let us begin with 01, which reports those facts reported by 0 about humans (who are moral agents) experiencing pain or pleasure that we know to be biologically useful. We know antecedently-that is, we know independent of the observations and testimony 0 reports-that humans are goal-directed organic systems, composed of parts that systematically contribute to the biological goals of these systems. This seems to give us reason to expect that human pain and pleasure, if they exist, will also systematically contribute to these goals. (And this is, of course, precisely what 01 reports.) But notice that pain and pleasure are in one respect strikingly dissimilar to other parts of organic systems: they have intrinsic moral value. Pain is intrinsically bad, and pleasure is intrinsically good. Does this difference substantially decrease the amount of support that our antecedent knowledge about humans gives to the "prediction" that pain and pleasure, if they exist, will systematically contribute to biological goals? I submit that it does if we assume that theism is true, but does not if we assume that HI is true. It is this difference between HI and theism that makes P(O1/HI) much greater than P(Ol/theism). Allow me to explain. HI entails that, if pain and pleasure exist, then they are not the result of malevolent or benevolent actions performed by nonhuman persons. So on HI, the moral difference between pain and pleasure and other parts of organic systems gives us no antecedent reason to believe that pain and pleasure will not play the same biological role that other parts of organic systems play. Indeed, a biological explanation of pain and pleasure is just the sort of explanation that one would expect on HI. But theism entails that God is responsible for the existence of any pain and pleasure in the world. Since God is morally perfect, He would have good moral reasons for producing pleasure even if it is never biologically useful, and He would not permit pain unless He had, not just a biological reason, but also a morally sufficient reason to do so. And since God is omnipotent and omniscient, He could create goal-directed organic systems (including humans) without biologically useful pain and pleasure. So theism entails both that God does not

PAIN AND PLEASURE

337

need biologically useful pain and pleasure to produce human goaldirected organic systems and that, if human pain and pleasure exist, then God had good moral reasons for producing them, reasons that, for all we know antecedently, might very well be inconsistent with pain and pleasure systematically contributing to the biological goals of human organisms. Therefore, we would have much less reason on theism than on HI to be surprised if it turned out that human pain and pleasure differed from other parts of organic systems by not systematically contributing to the biological goals of those systems. Hence, since 01 reports that the pain and pleasure experienced by humans (who are moral agents) do contribute in this way, P(O1/HI) is much greater than P(Ol/theism). One might object that from theism and our antecedent knowledge that goal-directed organic systems exist we can infer that the biological functions of the parts of those systems are themselves morally worthwhile, which gives us reason on theism that we do not have on HI to expect pain and pleasure to have biological functions. It might be thought that this counterbalances the reasons offered above for concluding that 01 is antecedently much more likely given HI than it is given theism.4 Now we obviously cannot infer from theism and our antecedent knowledge that, the greater the number of functioning parts in an organic system, the more valuable the system. We might be able to infer that organic systems are valuable and that the parts of these systems that have biological functions are valuable because the systems could not exist without functioning parts. But this does not imply that we have as much or even close to as much reason on theism as on HI to expect pain and pleasure to have biological functions. For an omnipotent and omniscient being could produce such systems without biologically useful pain and pleasure. Thus, since a morally perfect being would try to accomplish its goals with as little pain as possible, the value of organic systems gives us no reason on theism to expect pain to have biological functions. And since pleasure has intrinsic value and so is worth producing whether or not it furthers some other goal, the value of organic systems gives us very little reason on theism to expect pleasure to have biological functions. 02 reports the observations and testimony reported by 0 about sentient beings that are not moral agents (e.g., young human children and nonhuman animals) experiencing pain or pleasure that we know to be biologically useful. Independent of the observations and testimony 0 reports, we know that some sentient beings that are not moral agents are biologically very similar to moral agents. Since 01 implies that moral agents experience biologically useful pain and pleasure, this knowledge makes it antecedently likely on HI & 01

338

NOUS

that some sentient beings that are not moral agents will also experience biologically useful pain and pleasure. Now at first glance, one might think that this knowledge makes the existence of such pain and pleasure just as likely on theism & 01. After all, from the assumption that theism and 01 are both true it follows that God has good moral reasons for permitting biologically useful pain. But there is an important difference between the biologically useful pain that 01 reports and the biologically useful pain that 02 reports. Given theism & 01, we have reason to believe that God permits the pain 01 reports because it plays some sort of (presently indiscernible) moral role in the lives of the humans that experience it. But the pain 02 reports cannot play such a role, since the subjects of it are not moral agents. This difference is plainly not relevant on HI & 01, but it gives us some reason on theism & 01 to expect that the good moral reasons God has for permitting moral agents to experience pain do not apply to animals that are not moral agents, and hence some reason to believe that God will not permit such beings to experience pain. So P(02/HI & 01) is somewhat greater than P(02/theism & 01). 03 reports facts about sentient beings experiencing pain or pleasure that we do not know to be biologically useful. This includes much pain and pleasure that we know to be biologically gratuitous, as well as some that is not known to be useful and is also not known to be gratuitous. I will give a two-part argument for the conclusion that P(03/HI & 01 & 02) is much greater than P(03/theism & 01 & 02). First, we obviously have much more reason on theism & 01 & 02 than we have on HI & 01 & 02 to expect sentient beings (especially nonhuman animals) to be happy-in any case much more happy than they would be if their pleasure were limited to that reported by 01 and 02. Instead, when the facts 03 reports are added to those reported by 01 and 02, we find that many humans and animals experience prolonged and intense suffering and a much greater number are far from happy. In addition, we have more reason on theism & 01 & 02 than on HI & 01 & 02 to expect to discover a close connection between certain moral goods (e.g., justice and virtue) and biologically gratuitous pain and pleasure, but we discover no such connection. Second, we have, antecedently, much more reason on HI & 01 & 02 than on theism & 01 & 02 to believe that the fundamental role of pain and pleasure in our world is a biological one and that the presence of biologically gratuitous pain and pleasure is epiphenomenal, a biological accident resulting from nature's or an indif-

PAIN AND PLEASURE

339

ferent creator's failure to "fine tune" organic systems. And this is undeniably supported (though not entailed) by what 03 reports. To demonstrate this, a couple definitions are needed. First, by "pathological" pain or pleasure, I mean pain or pleasure that results from the failure of some organic system to function properly. For example, pain caused by terminal cancer and sadistic pleasure are pathological in this sense. And second, by "biologically appropriate" pain or pleasure, I mean pain or pleasure that occurs in a situation which is such that, it is biologically useful that pain or pleasure is felt in situations of this sort. For instance, the pain felt by a person killed in a fire is not biologically useful, but it is biologically appropriate because it is biologically useful that humans feel pain when they come in contact with fire. Clearly much of the pain and pleasure reported by 03 is either pathological or biologically appropriate, and very little is known to be both non-pathological and biologically inappropriate.5 And this is exactly what one would expect if pain and pleasure are fundamentally biological rather than moral phenomena, and so is much more to be expected on HI & 01 & 02 than on theism & 01 & 02. Therefore, assuming that theodicies do not significantly raise P(0/theism), the first and third multiplicands of A are much greater than the first and third multiplicands of B, and the second multiplicand of A is greater than the second multiplicand of B. And this implies that P(0/HI) is much greater than P(0/theism).
III. THE MORAL VALUE OF PAIN AND PLEASURE

In addition to their biological roles, pain and pleasure also play various moral roles in our world. By appealing to these roles, the theist might hope to explain some of the facts 0 reports in terms of theism, and thereby render 0 less surprising on theism than it is initially. This would seem to be the theist's most promising strategy for undermining the argument for C given above. Theodicies can be treated as attempts to carry out such a strategy.6 While few would deny that most theodicies are rather obvious failures, it is widely thought that plausible theistic explanations of suffering can be constructed by appealing to the intrinsic or instrumental moral value of free will. So it is necessary to determine what effect such theodicies have on P(0/theism). Additionally, it is important to evaluate the increasingly popular position that evidential arguments from evil against theism fail because the disproportion between omniscience and human knowledge makes it quite likely, on the assumption that God exists, that humans would not understand why God permits evil. (For a defense of this position, see (Wykstra 1984).)

340 A. EVALUATING THEODICIES

NOUS

Explaining some phenomenon in terms of a statement usually involves adding other statements to that statement. This is certainly true in the case of theodicies, which typically add to the claim that God exists the claims that God has a certain goal, that even God must produce or permit certain evils in order to accomplish that goal, and that accomplishing the goal is, from a moral point of view, worth the evils. I will say that a statement h* is an "expansion" of a statement h just in case h* is known to entail h. (Notice that h* can be an expansion of h even if it is logically equivalent to h.) The effect of a theodicy on P(O/theism) can be assessed by identifying an appropriate expansion Tn of theism that the theodicy employs and then using the following principle to evaluate P(O/theism) (cf. (Adams 1985, appendix, p. 252)): P(O/theism) = (P(Tn/theism) x P(O/T )) + (P(- Tn/theism) x P(O/theism & -TJ)).7 I will call this principle the "Weighted Average Principle" ("WAP" for short) because it identifies one probability with a probability weighted average of two others. Roughly, WAP tells us that P(O/theism) is the average of P(O/Tn) and P(O/theism & -TJ). This average, however, is a probability weighted average, the weights The higher of which are P(T n/theism) and P(- Tn/theism). the closer P(O/theism) will be to P(O/T ). And the P(Tn/theism), lower P(Tn/theism), the closer P(O/theism) will be to P(O/theism TJ) & WAP clarifies the relationship between theodicies and the argument for C I gave in Section II. For example, suppose that, for some expansion Tn of theism that a certain theodicy employs, is high. My argument for C in Section II ignores P(Tn/theism) this theodicy and so in effect equates P(O/theism) with P(O/theism is high, WAP tells us that & '-TJ ). Since P(Tn/theism) P(O/theism) is actually closer to P(O/T ) than to P(O/theism & -T ) (assuming that these are not' the same). To successfully defend my assumption in Section II that this theodicy does not significantly raise P(O/theism), I would need to show that P(O/T ) is In other words, not significantly greater than P(O/theism & -T I would need to show that, independent of the observations and testimony 0 reports, we have little or no more reason on Tn than we have on theism & -Tn to believe that 0 is true.
B. FREE WILL AND THE ADVANCEMENT OF MORALITY

Most free will theodicies appeal to a certain sort of moral freedom,

PAIN AND PLEASURE

341

which I will call "freedom*." An action is free* only if (i) it is free in an incompatibilist sense-that is, in a sense incompatible with its being determined by antecedent conditions outside the agent's control-and (ii) if it is morally right, then at least one alternative action that is open in an incompatibilist sense to the agent is such that it would be morally wrong for the agent to perform that alternative action. This concept of freedom is used to give the following theistic explanation of immorality. Freedom* has great value (either because morally right actions that are freely* performed are more valuable than right actions that are not freely* performed or because, following Hick (1966), moral virtue that is acquired by freely* performing right actions is more valuable than moral virtue that is not freely* acquired). For this reason, God endows humans with freedom*. However, since it is logically impossible to force a person to freely* perform a right action instead of a wrong one, God cannot give humans freedom* and ensure that humans will never perform morally wrong actions. Unfortunately, humans sometimes abuse their freedom* by performing wrong actions. Nevertheless, God is justified in giving humans freedom* because a world in which humans freely* perform both right and wrong actions is (provided that the balance of right over wrong actions or of morally good humans over morally bad humans is sufficiently favorable) better than a world in which immorality is prevented by withholding freedom* from humans. Notice that, so far, we have no explanation of the existence of pain. For there are morally right actions and morally wrong actions that do not entail the existence of pain. Wrong actions of this sort include some instances of breaking promises, killing, attempting to cause pain, and depriving someone of pleasure. So God could have given humans freedom* without permitting pain. The first version of the free will theodicy that I will evaluate adds to the above explanation of immorality the proposal that God permits pain in order to advance morality. This proposal can be spelled out in the following way. God wants humans to freely* perform right actions instead of wrong ones. Of course, as mentioned above, He cannot force humans to freely* perform only morally right actions, but He would have some control over the balance of right over wrong actions because even free* choices can be influenced and because God would know what free* choices humans would make (or would be likely to make8) in various situations. In particular, God might use pain to influence humans to freely* perform right actions instead of wrong ones. Also, some right actions entail the existence of pain, and God might know prior to creating humans that some or all humans would perform (or would be likely to perform) these right actions if given the chance. Therefore, God might use pain to obtain a more favor-

342

NOUS

able balance of freely* performed right actions over wrong actions. This version of the free will theodicy employs the following expansion of theism: Ti: God exists, and one of His final ends is a favorable balance of freely* performed right actions over wrong actions.9 I doubt that a consensus could ever be reached about P(Ti/theism). For T1 presupposes several very controversial metaphysical and ethical positions. For example, it presupposes that the concept of "~freedom*" is coherent, that humans have freedom*, and that freedom* is of great value. Since I obviously do not have the space here to discuss how plausible these claims are, I will assume for the sake of argument that P(Ti/theism) is high. I will argue, however, that P(O/Ti) is not significantly higher than P(O/theism & -Ti). This implies that, even if P(Ti/theism) is high, our first version of the free will theodicy does not significantly increase P(O/theism). If, as I will assume, it is morally permissible for God to use pain to advance morality, then we have reason on T1 that we do not have on theism & -T1 to expect that the world will contain both pain that influences humans to perform morally right actions and pain that is logically necessary for some of the right actions humans perform. Since 0 reports the existence of pain of both these sorts, we have a predictive success for the theodicy. But 0 also reports both that pain often influences humans to perform morally wrong actions and that pain is logically necessary for many of the wrong actions humans perform. And we have reason on T1 that we do not have on theism & -T1 to be surprised by these facts. Furthermore, the observations and testimony 0 reports provide strong evidence that the world does not presently contain a very impressive balance of right over wrong actions performed by humans and that this is due in part both to a variety of demoralizing conditions like illness, poverty and ignorance, and to the absence of conditions that tend to promote morality. All of this is even more surprising on T1 than on theism & -T1. (Cf. (Adams 1985, pp. 250-251).) So T1 's "predictive" advantages are counterbalanced by several serious "predictive" disadvantages, and for this reason P(O/Ti) is not significantly greater than P(O/theism & -T1).
C. FREE WILL AND RESPONSIBILITY

Some free will theodicists claim that God gives humans the freedom* to bring about suffering (either by producing it or by failing to prevent it) in order to increase the responsibility humans have for their own well-being and the well-being of others and thereby increase

PAIN AND PLEASURE

343

the importance of the moral decisions humans make. By an "important" moral decision, these theodicists mean a decision upon which the presence or absence of something of great positive or negative value depends. The key value judgement here is that, all else held equal, the more important the moral decisions we are free* to make, the more valuable our freedom* is. By not preventing us from freely* bringing about evils, including serious ones, God increases our control over how valuable the world is and thereby increases the value of our freedom*. This theodicy employs the following expansion of theism: T2: God exists, and one of His final ends is for humans to have the freedom* to make very important moral decisions. I will assume, once again, that P(T2/theism) is high, and I will argue that this second version of the free will theodicy does not significantly raise P(O/theism) by arguing that P(O/T2) is not significantly greater than P(O/theism & -T2). I will begin by arguing that Richard Swinburne (1979, Ch. 11) fails in his attempt to extend this theodicy so that it accounts for pain for which humans are not morally responsible. (I will call this sort of pain "amoral pain.") Swinburne believes that free will theodicies that employ T2 can account for such pain because (i) they explain why God gives humans the freedom* to bring about suffering and (ii) amoral pain is necessary if humans are to have genuine freedom* to bring about suffering. Swinburne defends (ii) in the following way. Freedom* to bring about suffering requires the knowledge of how to bring about suffering. And humans can obtain such knowledge in only one of two ways: either by God telling them how to bring about suffering or by experiencing how this is done. Unfortunately, if God told humans how to bring about suffering, then humans would know that God exists, and hence would have little temptation to do evil and so no genuine freedom* to bring about suffering. So for humans to have such freedom, they must learn by experience how to bring about suffering, and hence must learn this either by observing suffering for which no human is morally responsible or by observing suffering brought about by other humans. But for any particular kind of suffering, there must have been a first time that a human brought it about, and so a time when a human knew how to bring it about despite never having observed suffering of that kind brought about by a human. Hence, if humans are to learn by experience how to bring about suffering, then amoral pain must exist. Therefore, such suffering is necessary for humans to have the freedom* to bring about suffering. I will make three comments about this argument for (ii). First,

344

NOUS

even if it is sound, it obviously does not provide an adequate theistic account of amoral pain from which humans gain no new knowledge about how to produce or prevent suffering. Second, even if it is sound, it does not provide an adequate theistic explanation of most of the amoral pain that does give humans new knowledge of this sort. For an omnipotent and omniscient being could have greatly decreased the variety of ways in which humans know how to harm others, and so greatly decreased the amount of amoral pain needed for this knowledge, without decreasing the amount of harm humans can do to others and so without decreasing the amount of control that humans have over the well-being of others. Third, and most importantly, the argument is not sound. As Stump (1983) and Moser (1984) have observed, God could, without permitting amoral pain, give humans the knowledge of how to bring about suffering without revealing His existence and so without undermining human freedom*. For example, as Stump (pp. 52-53) has pointed out, humans might regularly have vivid, message-laden dreams and learn of their reliability, and yet not be compelled to believe in God. So if this second version of the free will theodicy raises P(O/theism) at all, it is because we have reason on T2 that we do not have on theism & - T2 to expect the existence of pain for which humans are morally responsible. Now giving humans the freedom* to bring about intense suffering is certainly one way (though not the only way) of giving humans the freedom* to make important moral decisions. So assuming that there is no better way,'0 we have some reason on T2 to expect humans to have such freedom, and so reason on T2 to expect the existence of pain for which humans are morally responsible. But even granting all this, it can be shown that P(O/T2) is not significantly greater than P(O/theism & -T2) by showing that other facts 0 reports are even more surprising oh T2 than they are on theism & -T2. An analogy between God and a good parent will be useful here. Ironically, such an analogy is often used to defend this sort of theodicy. For example, Swinburne (1979) responds to the objection that God should not give humans the freedom* to seriously harm others by asserting that the objector is asking that God "make a toy-world, a world where [our choices] matter, but not very much" (p. 219). Such a God "would be like the over-protective parent who will not let his child out of sight for a moment" (p. 220). But Swinburne neglects to ask whether or not humans are worthy of the freedom* to seriously harm others. A good parent gradually increases a child's responsibility as the child becomes capable of handling greater responsibility. Children who are unworthy of a certain responsibility are not benefitted by parents who give them

PAIN AND PLEASURE

345

that responsibility. On the assumption that T2 is.true, one would expect that God would behave like a good parent, giving humans great responsibility only when we are worthy of it. I am not claiming that on T2 one would expect God to impose a good moral character on humans before He gives them serious responsibilities. Nor am I claiming that creatures who are worthy of great responsibility would never abuse that responsibility. Rather, I am claiming that on T2 one would expect God to give all or some humans less responsibilityand in particular no ability to do serious evils-until they freely* developed the strength of character that would make them worthy of greater responsibility. And if at some point humans become worthy of and are given great responsibility, but nevertheless abuse this responsibility to such an extent that they are no longer worthy of it, then one would on T2 expect God, like a good parent, to decrease the amount of responsibility humans have until they are worthy of a second chance. But 0 conflicts with all of these expectations. Many humans are plainly not worthy of the freedom* to do serious evils. Nor is the human rate making any significant amount of moral progress. If God exists, then for centuries He has been allowing his children to torment, torture, and kill each other. Thus, even if they were once worthy of great responsibility, they no longer are, and hence are not benefitted by having such responsibility. So like T1, T2's predictive advantages are counterbalanced by several serious predictive disadvantages. Therefore, P(O/T2) is not significantly greater than P(O/theism & - T2), and hence this second version of the free will theodicy fails to significantly raise P(O/theism).
D. THE "INFINITE INTELLECT DEFENSE"

Some philosophers think that "evidential arguments from evil" can be refuted by pointing out that, since God's knowledge about good and evil is limitless, it is not all that surprising that He produces or permits evils for reasons that are unknown to humans. The expansion of theism suggested here is the following: T3: God exists and has a vast amount of knowledge about good and evil and how they are related that humans do not have. Since P(T3/theism) = one, P(O/theism) = P(O/T3). But this does not reveal any defect in my argument for C. For antecedentlythat is, independent of the observations and testimony 0 reports-we have no reason to think that God's additional knowledge concerning good and evil is such that He would permit any of the facts 0 reports to obtain. Of course, an omnipotent and omniscient being might, for all we know antecedently, have moral reasons unknown to us

346

NOUS

to permit the evil reported by 0. But it is also the case that such a being might, for all we know antecedently, have moral reasons unknown to us to prevent this evil. Indeed, we have no more reason antecedently to believe that such a being would know of some great good unknown to us whose existence entails the existence of the pain 0 reports than we have reason to believe that such a being would know of some great good unknown to us whose existence entails the nonexistence of the pain or the pleasure that 0 reports. And an omnipotent and omniscient being might very well know of means, far too complicated for humans to understand, by which He could obtain certain goods without the evil 0 reports. Of course, given thefacts 0 reports, we have some reason on T3 to expect that humans will be unable to produce a plausible theistic explanation of those facts. But HI gives us even more reason to expect this. So human ignorance does not solve the theist's evidential problems. Hence, none of the theodicies we have considered significantly raises P(O/theism). Therefore, relative to the epistemic situations of those of us who are unable to think of some other much more successful theodicy (i.e., all of us, I suspect), C is true: P(O/HI) is much greater than P(O/theism).
IV. THE SIGNIFICANCE OF THE PROBLEM

In The Origin of Species, Charles Darwin argued that his theory of the evolution of species by means of natural selection explains numerous facts (e.g., the geographical distribution of species and the existence of atrophied organs in animals) much better than the alternative hypothesis that each species of plant and animal was independently created by God. (Let us call this latter hypothesis "special creationism.") Darwin's results were significant partly because special creationists at Darwin's time did not have nor were they able to obtain any evidence favoring special creationism over evolution theory that outweighed or at least offset Darwin's evidence favoring evolution theory over special creationism. For this reason, many theists, while continuing to believe in creationism, which is consistent with Darwin's theory, rejected special creationism. And those theists who were familiar with Darwin's arguments and yet remained special creationists did so at a cost: their belief in special creationism was no longer an epistemically rational one. Similarly, how significant my results are depends, in part, on how many theists have or could obtain propositional or nonpropositional evidence favoring theism over HI that offsets the propositional evidence, provided by my argument for C, favoring HI over theism." Any theist confronted with my argument for C that lacks

PAIN AND PLEASURE

347

such evidence and is unable to obtain it cannot rationally continue to believe that theism is true. It is beyond the scope of this paper to determine how many theists would be in such a position. But I will make four sets of comments that I hope indicate how difficult a theist's search for the needed evidence might be. First, I do not see how it could be shown that HI is an ad hoc hypothesis or that theism is intrinsicallymore probable than HI. For HI is consistent with a wide variety of both naturalistic and supernaturalistic hypotheses, and it has no positive ontological commitments. Theism, on the other hand, is a very specific supernaturalistic claim with a very strong ontological commitment. Indeed, such differences between theism and HI might very well provide additional evidence favoring HI over theism. Second, traditional and contemporary arguments for theism are far from compelling-that is, they are far from being so persuasive as to coerce the acceptance of all or even most rational theists. Thus, even if some such argument were sound, most theists, including many philosophically sophisticated ones, would not recognize this, and hence the argument would not provide them with evidence favoring theism over HI. (The evidence would exist, but they would not have it.) Third, many traditional and contemporary arguments for theism, including many versions of the cosmological argument, the teleological argument, and the argument from consciousness, may not solve the theist's problem even if they are sound and recognized by the theist to be so. For they at most purport to show that an omnipotent and omniscient being exists - not that that being is morally perfect. Suppose then that some such argument is sound. My argument for C would work just as well if HI were replaced with the following hypothesis, which I will call "the Indifferent Deity Hypothesis": There exists an omnipotent and omniscient person who created the Universe and who has no intrinsic concern about the pain or pleasure of other beings. Like theism, this hypothesis entails that an omnipotent and omniscient being exists. So establishing that such a being exists would help the theist only if the theist also has strong evidence favoring theism over the Indifferent Deity Hypothesis.'2 Finally, religious experiences of the kind appealed to by "Reformed Epistemologists" like Alvin Plantinga (1983) are ambiguous with respect to the moral attributes of the creator. While Plantinga is correct in claiming that theists typically do feel inclined in certain circumstances (e.g., "when life is sweet and satisfying") to think that the creator is morally good, sensitive theists also feel

348

NOUs

inclined in other circumstances-namely, when they experience poignant evil-to believe that the creator is indifferent to their good or to the good of others. And many atheists have very powerful experiences in which they seem to be aware of the ultimate indifference of nature. These experiences are very common and are very similar phenomenologically to the experiences Plantinga mentions. Moreover, C implies that these "experiences of indifference" are better corroborated than the "theistic experiences" to which Plantinga appeals. Thus, even if Plantinga is correct in thinking that theistic experiences confer prima facie justification on the theist's belief in God, experiences of indifference defeat this justification. Therefore, theistic experiences do not provide non-propositional evidence that favors theism over HI, or at least none that outweighs the propositional evidence favoring HI over theism provided by my argument for C.13
REFERENCES Adams, Robert M.
1977 14, Middle Knowledge and the Problem of Evil, AmericanPhilosophicalQuarterly 109-117.

James E. Tomberlin and Peter Plantinga on the Problem of Evil, in Alvin Plantinga, van Inwagen, eds., Profiles,Dordrecht: D. Reidel, pp. 225-255. Ahern, M.B. of 1971 The Problem Evil, London: Routledge and Kegan Paul. Boorse, Christopher 1985
1976 1968 Wright on Functions, The Philosophical Review 85, 70-86. Charles C. Thomas.

Hare, Peter H. and Madden, Edward H.
Evil and the Concept of God, Springfield:

Evil and Inconclusiveness, Sophia 11, 8-12. 1972 Hick, John
1966 1980 Evil and the God of Love, New York: Harper and Row. Dialogues Concerning Natural Religion, Richard Popkin, ed., Indianapolis: Hackett

Hume, David Publishing Co. Kane, Stanley Theism and Evil, Sophia 9, 14-21. 1970 Mavrodes, George I.
1970 1984 Belief in God: A Study in the Epistemology of Religion, New York: Random House.

Moser, Paul K.
Natural Evil and the Free Will Defense, InternationalJournalfor the Philosophyof Religion

15, 49-56. Pargetter, Robert Evil as Evidence Against the Existence of God, Mind 85, 242-245. 1976 Plantinga, Alvin Studies35, 1-53. The Probabilistic Argument From Evil, Philosophical 1979
1983 Reason and Belief in God, in Faith and Rationality: Reason and Belief in God, Alvin

Plantinga and Nicholas Wolterstorff, eds., Notre Dame: University of Notre Dame Press, pp. 16-93.

PAIN AND PLEASURE

349

Reichenbach, Bruce 1980 The Inductive Argument From Evil, American Philosophical Quarterly 17, 221-227. Ruse, Michael of 1973 The Philosophy Biology, London: Hutchingson & Co. Sircello, Guy A New Theory of Beauty, Princeton: Princeton University Press. 1975 Stump, Eleonore for Journal thePhilosophy 1983 Knowledge, Freedom and the Problem of Evil, International of Religion 14, 49-58. Swinburne, Richard 1979 The Existence God, Oxford: Clarendon Press. of Wykstra, Steven 1984 The Humean Obstacle to Evidential Arguments From Suffering: On Avoiding the of the InternationalJournalfor Philosophy Religion 16, 73-93. Evils of "Appearance," Yandell, Keith E. 1969a Ethics, Evils, and Theism, Sophia 8, 18-28. 1969b A Premature Farewell to Theism, Religious Studies 5, 251-255. Theism and Evil: A Reply, Sophia 11, 1-7. 1972
NOTES

'I agree with most philosophers of religion that theists face no serious logical problem of evil. This paper challenges the increasingly popular view (defended recently by Pargetter, 1976, Plantinga, 1979, and Reichenbach, 1980) that theists face no serious evidential problem of evil. 2The concept of epistemic probability is an ordinary concept of probability for which no adequate philosophical analysis has, in my opinion, been proposed. As a first approximation, however, perhaps the following analysis will do: Relative to K, p is epistemically more probable than q, where K is an epistemic situation and p and q are propositions, just in case any fully rational person in K would have a higher degree of belief in p than in q. 30ne difficulty with the claim that this theorem of the probability calculus is true for epistemic probability is that, since multiplication and addition can only be performed on numbers, it follows that the theorem presupposes that probabilities have numerical values. But most epistemic probabilities have only comparative values. This difficulty can be overcome by intepreting the claim that this theorem is true for epistemic probability as the claim that (i) if each of the probabilities in the theorem have numerical values, then the theorem states the numerical relationshipswhich hold between them, and (ii) if at least one probability in the theorem does not have a numerical value, then all statementsof comparativeprobability entailed by that theorem are true. My reason for believing that this theorem is true for epistemic probability in this sense is that I can find no counterexample to it. I do not place a lot of emphasis on the mere fact that it is a theorem of the probability calculus. For I do not believe that all theorems of the probability calculus are true for epistemic probability. 4I am grateful to a Nodsreferee for this objection. 5Even the enjoyment of perceiving beauty may be biologically appropriate. For our enjoyment of clear perception is plausibly thought to be biolgically useful, and Guy Sircello (1975, pp. 129-134) gives a very interesting argument for the conclusion that perceiving beauty is a special case of clear perception. 6The term "theodicy" is often defined as "an attempt to state what God's actual reason for permitting evil is." This definition implies that, in order to show that some theodicy is successful, one must show that God exists. I prefer a definition of "theodicy" that avoids this implication. By a "theodicy" I mean an attempt to give a plausible theistic explanation of some fact about evil.

350

NOUS

= 7More generally, it follows from the probability calculus that P(O/theism) WAP x P(O/theism & -T.)). (P(T /theism) x P(O/theism & T.)) + (P(-T T/theism) replaces P(O/theism & T ) with P(O/T ) because T is an expansion of theism and hence is known to be logically equivalent to theism & T 8Robert Adams (1977) argues that God, despite being omniscient, would not know what free* choice a particular human would make in a certain situation prior to deciding both to place that human in that situation and to allow him to make that choice. Adams also argues, however, that God would have prior knowledge of what free* choices humans would be likely to make in various situations. 9A slightly different version of this theodicy employs the following expansion of theism:

T,*: God exists, and one of His final ends is a favorable balance of morally good humans whose moral goodness was freely* acquired over morally bad humans. I suspect that P(T1 */theism) is greater than P(T1/theism) because God would be more likely to be concerned about persons than about actions. However, I need not evaluate T, * separately because I will assume that P(T1/theism) is high and my arguments concerning P(O/T1) would work just as well if T, were replaced with T,*. 100ne might challenge this assumption and thereby attack theodicies that employ T2 in the following way. Choosing whether or not to produce a large amount of pleasure is, all else held equal, a more important moral decision than choosing whether or not to produce a small amount of pain. Hence, it would seem that by increasing our capacity to produce or prevent pleasure, God could give us the power to make moral decisions about pleasure that are as important as any that we now make concerning pain. But it is antecedently likely that such a world would be a better world than one in which humans have the ability to cause others to suffer. Therefore, it is antecedently unlikely that God would use pain to accomplish His goal of giving humans important moral choices. "1One way of attempting to show that such evidence exists would be to (i) identify an appropriate body of evidence (call it O*) that is broader than 0 (e.g., a statement reporting the relevant observations and testimony, not just about pain and pleasure, but about all intrinsic goods and evils) and then (ii) attempt to show that, independent of the observations and testimony O* reports, O* is at least as likely on theism as it is on HI. 12Swinburne (1979, Ch. 5) argues that quasi-theistic hypotheses like the Indifferent Deity Hypothesis are intrinsically much less probable than theism. I do not believe his argument is sound, but if it were, then strong evidence favoring theism over the Indifferent Deity Hypothesis would be available. 13For criticisms of previous versions of this paper, I am grateful to Gary Gutting, C. Stephen Layman, Nelson Pike, Alvin Plantinga, Philip L. Quinn, and an anonymous Nou's referee.