and (if so) to what extent the existence of evil (or certain
instances, kinds, quantities, or distributions of evil) constitutes
evidence against the existence of God, that is to say, a being perfect
in power, knowledge and goodness. Evidential arguments from evil
attempt to show that, once we put aside any evidence there might be in
support of the existence of God, it becomes unlikely, if not highly
unlikely, that the world was created and is governed by an omnipotent,
omniscient, and wholly good being. Such arguments are not to be
confused with logical arguments from evil, which have the more
ambitious aim of showing that, in a world in which there is evil, it
is logically impossible – and not just unlikely – that God exists.
This entry begins by clarifying some important concepts and
distinctions associated with the problem of evil, before providing an
outline of one of the more forceful and influential evidential
arguments developed in contemporary times, namely, the evidential
argument advanced by William Rowe. Rowe's argument has occasioned a
range of responses from theists, including the so-called "skeptical
theist" critique (according to which God's ways are too mysterious for
us to comprehend) and the construction of various theodicies, that is,
explanations as to why God permits evil. These and other responses to
the evidential problem of evil are here surveyed and assessed.
1. Background to the Problem of Evil
Before delving into the deep and often murky waters of the problem of
evil, it will be helpful to provide some philosophical background to
this venerable subject. The first and perhaps most important step of
this stage-setting process will be to identify and clarify the
conception of God that is normally presupposed in contemporary debates
(at least within the Anglo-American analytic tradition) on the problem
of evil. The next step will involve providing an outline of some
important concepts and distinctions, in particular the age-old
distinction between "good" and "evil," and the more recent distinction
between the logical problem of evil and the evidential problem of
evil.
a. Orthodox Theism
The predominant conception of God within the western world, and hence
the kind of deity that is normally the subject of debate in
discussions on the problem of evil in most western philosophical
circles, is the God of 'orthodox theism'. According to orthodox
theism, there exists just one God, this God being a person or
person-like. The operative notion, however, behind this form of theism
is that God is perfect, where to be perfect is to be the greatest
being possible or, to borrow Anselm's well-known phrase, the being
than which none greater can be conceived. (Such a conception of God
forms the starting-point in what has come to be known as 'perfect
being theology'; see Morris 1987, 1991, and Rogers 2000). On this
view, God, as an absolutely perfect being, must possess the following
perfections or great-making qualities:
1. omnipotence: This refers to God's ability to bring about any
state of affairs that is logically possible in itself as well as
logically consistent with his other essential attributes.
2. omniscience: God is omniscient in that he knows all truths or
knows all that is logically possible to know.
3. perfect goodness: God is the source of moral norms (as in divine
command ethics) or always acts in complete accordance with moral
norms.
4. aseity: God has aseity (literally, being from oneself, a se
esse) – that is to say, he is self-existent or ontologically
independent, for he does not depend either for his existence or for
his characteristics on anything outside himself.
5. incorporeality: God has no body; he is a non-physical spirit but
is capable of affecting physical things.
6. eternity: Traditionally, God is thought to be eternal in an
atemporal sense – i.e., God is timeless or exists outside of time (a
view upheld by Augustine, Boethius, and Aquinas). On an alternative
view, God's eternality is held to be temporal in nature, so that God
is everlasting or exists in time, having infinite temporal duration in
both of the two temporal directions.
7. omnipresence: God is wholly present in all space and time. This
is often interpreted metaphorically to mean that God can bring about
an event immediately at any place and time, and knows what is
happening at every place and time in the same immediate manner.
8. perfectly free: God is absolutely free either in the sense that
nothing outside him can determine him to perform a particular action,
or in the sense that it is always within his power not to do what he
does.
9. alone worthy of worship and unconditional commitment: God, being
the greatest being possible, is the only being fit to be worshipped
and the only being to whom one may commit one's life without
reservation.
The God of traditional theism is also typically accorded a further
attribute, one that he is thought to possess only contingently:
10. creator and sustainer of the world: God brought the (physical
and non-physical) world into existence, and also keeps the world and
every object within it in existence. Thus, no created thing could
exist at a given moment unless it were at that moment held in
existence by God. Further, no created thing could have the causal
powers and liabilities it has at a given moment unless it were at that
moment supplied with those powers and liabilities by God.
According to orthodox theism, God was free not to create a world. In
other words, there is at least one possible world in which God creates
nothing at all. But then God is a creator only contingently, not
necessarily. (For a more comprehensive account of the properties of
the God of orthodox theism, see Swinburne 1977, Quinn & Taliaferro
1997: 223-319, and Hoffman & Rosenkrantz 2002.)
b. Good and Evil
Clarifying the underlying conception of God is but the first step in
clarifying the nature of the problem of evil. To arrive at a more
complete understanding of this vexing problem, it is necessary to
unpack further some of its philosophical baggage. I turn, therefore,
to some important concepts and distinctions associated with the
problem of evil, beginning with the ideas of 'good' and 'evil'.
The terms 'good' and 'evil' are, if nothing else, notoriously
difficult to define. Some account, however, can be given of these
terms as they are employed in discussions of the problem of evil.
Beginning with the notion of evil, this is normally given a very wide
extension so as to cover everything that is negative and destructive
in life. The ambit of evil will therefore include such categories as
the bad, the unjust, the immoral, and the painful. An analysis of evil
in this broad sense may proceed as follows:
An event may be categorized as evil if it involves any of the following:
1. some harm (whether it be minor or great) being done to the
physical and/or psychological well-being of a sentient creature;
2. the unjust treatment of some sentient creature;
3. loss of opportunity resulting from premature death;
4. anything that prevents an individual from leading a fulfilling
and virtuous life;
5. a person doing that which is morally wrong;
6. the 'privation of good'.
Condition (a) captures what normally falls under the rubric of pain as
a physical state (e.g., the sensation you feel when you have a
toothache or broken jaw) and suffering as a mental state in which we
wish that our situation were otherwise (e.g., the experience of
anxiety or despair). Condition (b) introduces the notion of injustice,
so that the prosperity of the wicked, the demise of the virtuous, and
the denial of voting rights or employment opportunities to women and
blacks would count as evils. The third condition is intended to cover
cases of untimely death, that is to say, death not brought about by
the ageing process alone. Death of this kind may result in loss of
opportunity either in the sense that one is unable to fulfill one's
potential, dreams or goals, or merely in the sense that one is
prevented from living out the full term of their natural life. This is
partly why we consider it a great evil if an infant were killed after
impacting with a train at full speed, even if the infant experienced
no pain or suffering in the process. Condition (d) classifies as evil
anything that inhibits one from leading a life that is both fulfilling
and virtuous – poverty and prostitution would be cases in point.
Condition (e) relates evil to immoral choices or acts. And the final
condition expresses the idea, prominent in Augustine and Aquinas, that
evil is not a substance or entity in its own right, but a privatio
boni: the absence or lack of some good power or quality which a thing
by its nature ought to possess.
Paralleling the above analysis of evil, the following account of
'good' may be offered:
An event may be categorized as good if it involves any of the following:
1. some improvement (whether it be minor or great) in the physical
and/or psychological well-being of a sentient creature;
2. the just treatment of some sentient creature;
3. anything that advances the degree of fulfillment and virtue in
an individual's life;
4. a person doing that which is morally right;
5. the optimal functioning of some person or thing, so that it does
not lack the full measure of being and goodness that ought to belong
to it.
Turning to the many varieties of evil, the following have become
standard in the literature:
Moral evil. This is evil that results from the misuse of free will on
the part of some moral agent in such a way that the agent thereby
becomes morally blameworthy for the resultant evil. Moral evil
therefore includes specific acts of intentional wrongdoing such as
lying and murdering, as well as defects in character such as
dishonesty and greed.
Natural evil. In contrast to moral evil, natural evil is evil that
results from the operation of natural processes, in which case no
human being can be held morally accountable for the resultant evil.
Classic examples of natural evil are natural disasters such as
cyclones and earthquakes that result in enormous suffering and loss of
life, illnesses such as leukemia and Alzheimer's, and disabilities
such as blindness and deafness.
An important qualification, however, must be made at this point. A
great deal of what normally passes as natural evil is brought about by
human wrongdoing or negligence. For example, lung cancer may be caused
by heavy smoking; the loss of life occasioned by some earthquakes may
be largely due to irresponsible city planners locating their creations
on faults that will ultimately heave and split; and some droughts and
floods may have been prevented if not for the careless way we have
treated our planet. As it is the misuse of free will that has caused
these evils or contributed to their occurrence, it seems best to
regard them as moral evils and not natural evils. In the present work,
therefore, a natural evil will be defined as an evil resulting solely
or chiefly from the operation of the laws of nature. Alternatively,
and perhaps more precisely, an evil will be deemed a natural evil only
if no non-divine agent can be held morally responsible for its
occurrence. Thus, a flood caused by human pollution of the environment
will be categorized a natural evil as long as the agents involved
could not be held morally responsible for the resultant evil, which
would be the case if, for instance, they could not reasonably be
expected to have foreseen the consequences of their behavior.
A further category of evil that has recently played an important role
in discussions on the problem of evil is horrendous evil. This may be
defined, following Marilyn Adams (1999: 26), as evil "the
participation in which (that is, the doing or suffering of which)
constitutes prima facie reason to doubt whether the participant's life
could (given their inclusion in it) be a great good to him/her on the
whole". As examples of such evil, Adams lists "the rape of a woman and
axing off of her arms, psycho-physical torture whose ultimate goal is
the disintegration of personality, betrayal of one's deepest
loyalties, child abuse of the sort described by Ivan Karamazov, child
pornography, parental incest, slow death by starvation, the explosion
of nuclear bombs over populated areas" (p.26).
A horrendous evil, it may be noted, may be either a moral evil (e.g.,
the Holocaust of 1939-45) or a natural evil (e.g., the Lisbon
earthquake of 1755). It is also important to note that it is the
notion of a 'horrendous moral evil' that comports with the current,
everyday use of 'evil' by English speakers. When we ordinarily employ
the word 'evil' today we do not intend to pick out something that is
merely bad or very wrong (e.g., a burglary), nor do we intend to refer
to the death and destruction brought about by purely natural processes
(we do not, for example, think of the 2004 Asian tsunami disaster as
something that was 'evil'). Instead, the word 'evil' is reserved in
common usage for events and people that have an especially horrific
moral quality or character.
Clearly, the problem of evil is at its most difficult when stated in
terms of horrendous evil (whether of the moral or natural variety),
and as will be seen in Section II below, this is precisely how William
Rowe's statement of the evidential problem of evil is formulated.
Finally, these notions of good and evil indicate that the problem of
evil is intimately tied to ethics. One's underlying ethical theory may
have a bearing on one's approach to the problem of evil in at least
two ways.
Firstly, one who accepts either a divine command theory of ethics or
non-realism in ethics is in no position to raise the problem of evil,
that is, to offer the existence of evil as at least a prima facie good
reason for rejecting theism. This is because a divine command theory,
in taking morality to be dependent upon the will of God, already
assumes the truth of that which is in dispute, namely, the existence
of God (see Brown 1967). On the other hand, non-realist ethical
theories, such as moral subjectivism and error-theories of ethics,
hold that there are no objectively true moral judgments. But then a
non-theist who also happens to be a non-realist in ethics cannot help
herself to some of the central premises found in evidential arguments
from evil (such as 'If there were a perfectly good God, he would want
a world with no horrific evil in it'), as these purport to be
objectively true moral judgments (see Nelson 1991). This is not to
say, however, that atheologians such as David Hume, Bertrand Russell
and J.L. Mackie, each of whom supported non-realism in ethics, were
contradicting their own meta-ethics when raising arguments from evil –
at least if their aim was only to show up a contradiction in the
theist's set of beliefs.
Secondly, the particular normative ethical theory one adopts (e.g.,
consequentialism, deontology, virtue ethics) may influence the way in
which one formulates or responds to an argument from evil. Indeed,
some have gone so far as to claim that evidential arguments from evil
usually presuppose the truth of consequentialism (see, for example,
Reitan 2000). Even if this is not so, it seems that the adoption of a
particular theory in normative ethics may render the problem of evil
easier or harder, or at least delimit the range of solutions
available. (For an excellent account of the difficulties faced by
theists in relation to the problem of evil when the ethical framework
is restricted to deontology, see McNaughton 1994.)
c. Versions of the Problem of Evil
The problem of evil may be described as the problem of reconciling
belief in God with the existence of evil. But the problem of evil,
like evil itself, has many faces. It may, for example, be expressed
either as an experiential problem or as a theoretical problem. In the
former case, the problem is the difficulty of adopting or maintaining
an attitude of love and trust toward God when confronted by evil that
is deeply perplexing and disturbing. Alvin Plantinga (1977: 63-64)
provides an eloquent account of this problem:
The theist may find a religious problem in evil; in the presence
of his own suffering or that of someone near to him he may find it
difficult to maintain what he takes to be the proper attitude towards
God. Faced with great personal suffering or misfortune, he may be
tempted to rebel against God, to shake his fist in God's face, or even
to give up belief in God altogether… Such a problem calls, not for
philosophical enlightenment, but for pastoral care. (emphasis in the
original)
By contrast, the theoretical problem of evil is the purely
'intellectual' matter of determining what impact, if any, the
existence of evil has on the truth-value or the epistemic status of
theistic belief. To be sure, these two problems are interconnected –
theoretical considerations, for example, may colour one's actual
experience of evil, as happens when suffering that is better
comprehended becomes easier to bear. In this article, however, the
focus will be exclusively on the theoretical dimension. This aspect of
the problem of evil comes in two broad varieties: the logical problem
and the evidential problem.
The logical version of the problem of evil (also known as the a priori
version and the deductive version) is the problem of removing an
alleged logical inconsistency between certain claims about God and
certain claims about evil. J.L. Mackie (1955: 200) provides a succinct
statement of this problem:
In its simplest form the problem is this: God is omnipotent; God
is wholly good; and yet evil exists. There seems to be some
contradiction between these three propositions, so that if any two of
them were true the third would be false. But at the same time all
three are essential parts of most theological positions: the
theologian, it seems, at once must adhere and cannot consistently
adhere to all three. (emphases in the original)
In a similar vein, H.J. McCloskey (1960: 97) frames the problem of
evil as follows:
Evil is a problem for the theist in that a contradiction is
involved in the fact of evil, on the one hand, and the belief in the
omnipotence and perfection of God on the other. (emphasis mine)
Atheologians like Mackie and McCloskey, in maintaining that the
logical problem of evil provides conclusive evidence against theism,
are claiming that theists are committed to an internally inconsistent
set of beliefs and hence that theism is necessarily false. More
precisely, it is claimed that theists commonly accept the following
propositions:
11. God exists
12. God is omnipotent
13. God is omniscient
14. God is perfectly good
15. Evil exists.
Propositions (11)-(14) form an essential part of the orthodox
conception of God, as this has been explicated in Section 1 above. But
theists typically believe that the world contains evil. The charge,
then, is that this commitment to (15) is somehow incompatible with the
theist's commitment to (11)-(14). Of course, (15) can be specified in
a number of ways – for example, (15) may refer to the existence of any
evil at all, or a certain amount of evil, or particular kinds of evil,
or some perplexing distributions of evil. In each case, a different
version of the logical problem of evil, and hence a distinct charge of
logical incompatibility, will be generated.
The alleged incompatibility, however, is not obvious or explicit.
Rather, the claim is that propositions (11)-(15) are implicitly
contradictory, where a set S of propositions is implicitly
contradictory if there is a necessary proposition p such that the
conjunction of p with S constitutes a formally contradictory set.
Those who advance logical arguments from evil must therefore add one
or more necessary truths to the above set of five propositions in
order to generate the fatal contradiction. By way of illustration,
consider the following additional propositions that may be offered:
16. A perfectly good being would want to prevent all evils.
17. An omniscient being knows every way in which evils can come into
existence.
18. An omnipotent being who knows every way in which an evil can
come into existence has the power to prevent that evil from coming
into existence.
19. A being who knows every way in which an evil can come into
existence, who is able to prevent that evil from coming into
existence, and who wants to do so, would prevent the existence of that
evil.
From this set of auxiliary propositions, it clearly follows that
20. If there exists an omnipotent, omniscient, and perfectly good
being, then no evil exists.
It is not difficult to see how the addition of (16)-(20) to (11)-(15)
will yield an explicit contradiction, namely,
21. Evil exists and evil does not exist.
If such an argument is sound, theism will not so much lack evidential
support, but would rather be, as Mackie (1955: 200) puts it,
"positively irrational." For more discussion, see the article The
Logical Problem of Evil.
The subject of this article, however, is the evidential version of the
problem of evil (also called the a posteriori version and the
inductive version), which seeks to show that the existence evil,
although logically consistent with the existence of God, counts
against the truth of theism. As with the logical problem, evidential
formulations may be based on the sheer existence of evil, or certain
instances, types, amounts, or distributions of evil. Evidential
arguments from evil may also be classified according to whether they
employ (i) a direct inductive approach, which aims at showing that
evil counts against theism, but without comparing theism to some
alternative hypothesis; or (ii) an indirect inductive approach, which
attempts to show that some significant set of facts about evil counts
against theism, and it does this by identifying an alternative
hypothesis that explains these facts far more adequately than the
theistic hypothesis. The former strategy, as will be seen in Section
II, is employed by William Rowe, while the latter strategy is
exemplified best in Paul Draper's 1989 paper, "Pain and Pleasure: An
Evidential Problem for Theists". (A useful taxonomy of evidential
arguments from evil can be found in Russell 1996: 194 and Peterson
1998: 23-27, 69-72.)
Evidential arguments purport to show that evil counts against theism
in the sense that the existence of evil lowers the probability that
God exists. The strategy here is to begin by putting aside any
positive evidence we might think there is in support of theism (e.g.,
the fine-tuning argument) as well as any negative evidence we might
think there is against theism (that is, any negative evidence other
than the evidence of evil). We therefore begin with a 'level playing
field' by setting the probability of God's existing at 0.5 and the
probability of God's not existing at 0.5 (cf. Rowe 1996: 265-66; it is
worth noting, however, that this 'level playing field' assumption is
not entirely uncontroversial: see, for example, the objections raised
by Jordan 2001 and Otte 2002: 167-68). The aim is to then determine
what happens to the probability value of 'God exists' once we consider
the evidence generated by our observations of the various evils in our
world. The central question, therefore, is: Grounds for belief in God
aside, does evil render the truth of atheism more likely than the
truth of theism? (A recent debate on the evidential problem of evil
was couched in such terms: see Rowe 2001a: 124-25.) Proponents of
evidential arguments are therefore not claiming that, even if we take
into account any positive reasons there are in support of theism, the
evidence of evil still manages to lower the probability of God's
existence. They are only making the weaker claim that, if we
temporarily set aside such positive reasons, then it can be shown that
the evils that occur in our world push the probability of God's
existence significantly downward.
But if evil counts against theism by driving down the probability
value of 'God exists' then evil constitutes evidence against the
existence of God. Evidential arguments, therefore, claim that there
are certain facts about evil that cannot be adequately explained on a
theistic account of the world. Theism is thus treated as a large-scale
hypothesis or explanatory theory which aims to make sense of some
pertinent facts, and to the extent that it fails to do so it is
disconfirmed.
In evidential arguments, however, the evidence only probabilifies its
conclusion, rather than conclusively verifying it. The probabilistic
nature of such arguments manifests itself in the form of a premise to
the effect that 'It is probably the case that some instance (or type,
or amount, or pattern) of evil E is gratuitous'. This probability
judgment usually rests on the claim that, even after careful
reflection, we can see no good reason for God's permission of E. The
inference from this claim to the judgment that there exists gratuitous
evil is inductive in nature, and it is this inductive step that sets
the evidential argument apart from the logical argument.
2. William Rowe's Evidential Argument from Evil
Evidential arguments from evil seek to show that the presence of evil
in the world inductively supports or makes likely the claim that God
(or, more precisely, the God of orthodox theism) does not exist. A
variety of evidential arguments have been formulated in recent years,
but here I will concentrate on one very influential formulation,
namely, that provided by William Rowe. Rowe's version of the
evidential argument has received much attention since its formal
inception in 1978, for it is often considered to be the most cogent
presentation of the evidential problem of evil. James Sennett (1993:
220), for example, views Rowe's argument as "the clearest, most easily
understood, and most intuitively appealing of those available." Terry
Christlieb (1992: 47), likewise, thinks of Rowe's argument as "the
strongest sort of evidential argument, the sort that has the best
chance of success." It is important to note, however, that Rowe's
thinking on the evidential problem of evil has developed in
significant ways since his earliest writings on the subject, and two
(if not three) distinct evidential arguments can be identified in his
work. Here I will only discuss that version of Rowe's argument that
received its first full-length formulation in Rowe (1978) and, most
famously, in Rowe (1979), and was successively refined in the light of
criticisms in Rowe (1986), (1988), (1991), and (1995), before being
abandoned in favour of a quite different evidential argument in Rowe
(1996).
a. An Outline of Rowe's Argument
In presenting his evidential argument from evil in his justly
celebrated 1979 paper, "The Problem of Evil and Some Varieties of
Atheism", Rowe thinks it best to focus on a particular kind of evil
that is found in our world in abundance. He therefore selects "intense
human and animal suffering" as this occurs on a daily basis, is in
great plenitude in our world, and is a clear case of evil. More
precisely, it is a case of intrinsic evil: it is bad in and of itself,
even though it sometimes is part of, or leads to, some good state of
affairs (Rowe 1979: 335). Rowe then proceeds to state his argument for
atheism as follows:
1. There exist instances of intense suffering which an omnipotent,
omniscient being could have prevented without thereby losing some
greater good or permitting some evil equally bad or worse.
2. An omniscient, wholly good being would prevent the occurrence of
any intense suffering it could, unless it could not do so without
thereby losing some greater good or permitting some evil equally bad
or worse.
3. (Therefore) There does not exist an omnipotent, omniscient,
wholly good being. (Rowe 1979: 336)
This argument, as Rowe points out, is clearly valid, and so if there
are rational grounds for accepting its premises, to that extent there
are rational grounds for accepting the conclusion, that is to say,
atheism.
b. The Theological Premise
The second premise is sometimes called 'the theological premise' as it
expresses a belief about what God as a perfectly good being would do
under certain circumstances. In particular, this premise states that
if such a being knew of some intense suffering that was about to take
place and was in a position to prevent its occurrence, then it would
prevent it unless it could not do so without thereby losing some
greater good or permitting some evil equally bad or worse. Put
otherwise, an omnipotent, omniscient, wholly good God would not permit
any gratuitous evil, evil that is (roughly speaking) avoidable,
pointless, or unnecessary with respect to the fulfillment of God's
purposes.
Rowe takes the theological premise to be the least controversial
aspect of his argument. And the consensus seems to be that Rowe is
right – the theological premise, or a version thereof that is immune
from some minor infelicities in the original formulation, is usually
thought to be indisputable, self-evident, necessarily true, or
something of that ilk. The intuition here, as the Howard-Snyders
(1999: 115) explain, is that "on the face of it, the idea that God may
well permit gratuitous evil is absurd. After all, if God can get what
He wants without permitting some particular horror (or anything
comparably bad), why on earth would He permit it?"
An increasing number of theists, however, are beginning to question
Rowe's theological premise. This way of responding to the evidential
problem of evil has been described by Rowe as "radical, if not
revolutionary" (1991: 79), but it is viewed by many theists as the
only way to remain faithful to the common human experience of evil,
according to which utterly gratuitous evil not only exists but is
abundant. In particular, some members of the currently popular
movement known as open theism have rallied behind the idea that the
theistic worldview is not only compatible with, but requires or
demands, the possibility that there is gratuitous evil (for the
movement's 'manifesto', see Pinnock et al. 1994; see also Sanders
1998, Boyd 2000, and Hasker 2004).
Although open theists accept the orthodox conception of God, as
delineated in Section I.1 above, they offer a distinct account of some
of the properties that are constitutive of the orthodox God. Most
importantly, open theists interpret God's omniscience in such a way
that it does not include either foreknowledge (or, more specifically,
knowledge of what free agents other than God will do) or middle
knowledge (that is, knowledge of what every possible free creature
would freely choose to do in any possible situation in which that
creature might find itself). This view is usually contrasted with two
other forms of orthodox theism: Molinism (named after the
sixteenth-century Jesuit theologian Luis de Molina, who developed the
theory of middle knowledge), according to which divine omniscience
encompasses both foreknowledge and middle knowledge; and Calvinism or
theological determinism, according to which God determines or
predestines all that happens, thus leaving us with either no morally
relevant free will at all (hard determinism) or free will of the
compatibilist sort only (soft determinism).
It is often thought that the Molinist and Calvinist grant God greater
providential control over the world than does the open theist. For
according to the latter but not the former, the future is to some
degree open-ended in that not even God can know exactly how it will
turn out, given that he has created a world in which there are agents
with libertarian free will and, perhaps, indeterminate natural
processes. God therefore runs the risk that his creation will come to
be infested with gratuitous evils, that is to say, evils he has not
intended, decreed, planned for, or even permitted for the sake of some
greater good. Open theists, however, argue that this risk is kept in
check by God's adoption of various general strategies by which he
governs the world. God may, for example, set out to create a world in
which there are creatures who have the opportunity to freely choose
their destiny, but he would then ensure that adequate recompense is
offered (perhaps in an afterlife) to those whose lives are ruined
(through no fault of their own) by the misuse of others' freedom
(e.g., a child that is raped and murdered). Nevertheless, in creating
creatures with (libertarian) free will and by infusing the natural
order with a degree of indeterminacy, God relinquishes exhaustive
knowledge and complete control of all history. The open theist
therefore encourages the rejection of what has been called 'meticulous
providence' (Peterson 1982: chs 4 & 5) or 'the blueprint worldview'
(Boyd 2003: ch.2), the view that the world was created according to a
detailed divine blueprint which assigns a specific divine reason for
every occurrence in history. In place of this view, the open theist
presents us with a God who is a risk-taker, a God who gives up
meticulous control of everything that happens, thus opening himself up
to the genuine possibility of failure and disappointment – that is to
say, to the possibility of gratuitous evil.
Open theism has sparked much heated debate and has been attacked from
many quarters. Considered, however, as a response to Rowe's
theological premise, open theism's prospects seem dim. The problem
here, as critics have frequently pointed out, is that the open view of
God tends to diminish one's confidence in God's ability to ensure that
his purposes for an individual's life, or for world history, will be
accomplished (see, e.g., Ware 2000, Ascol 2001: 176-80). The worry is
that if, as open theists claim, God does not exercise sovereign
control over the world and the direction of human history is
open-ended, then it seems that the world is left to the mercy of Tyche
or Fortuna, and we are therefore left with no assurance that God's
plan for the world and for us will succeed. Consider, for example,
Eleonore Stump's rhetorical questions, put in response to the idea of
a 'God of chance' advocated in van Inwagen (1988): "Could one trust
such a God with one's child, one's life? Could one say, as the
Psalmist does, 'I will both lay me down in peace and sleep, for thou,
Lord, only makest me dwell in safety'?" (1997: 466, quoting from Psalm
4:8). The answer may in large part depend on the degree to which the
world is thought to be imbued with indeterminacy or chance.
If, for example, the open theist view introduces a high level of
chance into God's creation, this would raise the suspicion that the
open view reflects an excessively deistic conception of God's relation
to the world. Deism is popularly thought of as the view that a supreme
being created the world but then, like an absentee landlord, left it
to run on its own accord. Deists, therefore, are often accused of
postulating a remote and indifferent God, one who does not exercise
providential care over his creation. Such a deity, it might be
objected, resembles the open theist's God of chance. The objection, in
other words, is that open theists postulate a dark and risky universe
subject to the forces of blind chance, and that it is difficult to
imagine a personal God – i.e., a God who seeks to be personally
related to us and hence wants us to develop attitudes of love and
trust towards him – providing us with such a habitat. To paraphrase
Einstein, God does not play dice with our lives.
This, however, need not mean that God does not play dice at all. It is
not impossible, in other words, to accommodate chance within a
theistic world-view. To see this, consider a particular instance of
moral evil: the rape and murder of a little girl. It seems plausible
that no explanation is available as to why God would permit this
specific evil (or, more precisely, why God would permit this girl to
suffer then and there and in that way), since any such explanation
that is offered will inevitably recapitulate the explanation offered
for at least one of the major evil-kinds that subsumes the particular
evil in question (e.g., the class of moral evils). It is therefore
unreasonable to request a reason (even a possible reason) for God's
permission of a particular event that is specific to this event and
that goes beyond some general policy or plan God might have for
permitting events of that kind. If this correct, then there is room
for theists to accept the view that at least some evils are chancy or
gratuitous in the sense that there is no specific reason as to why
these evils are permitted by God. However, this kind of commitment to
gratuitous evil is entirely innocuous for proponents of Rowe's
theological premise. For one can simply modify this premise so that it
ranges either over particular instances of evil or (to accommodate
cases where particular evils admit of no divine justification) over
broadly defined evils or evil-kinds under which the relevant
particular evils can be subsumed. And so a world created by God may be
replete with gratuitous evil, as open theists imagine, but that need
not present a problem for Rowe.
(For a different line of argument in support of the compatibility of
theism and gratuitous evil, see Hasker (2004: chs 4 & 5), who argues
that the consequences for morality would be disastrous if we took
Rowe's theological premise to be true. For criticisms of this view,
see Rowe (1991: 79-86), Chrzan (1994), O'Connor (1998: 53-70), and
Daniel and Frances Howard-Snyder (1999: 119-27).)
c. The Factual Premise
Criticisms of Rowe's argument tend to focus on its first premise,
sometimes dubbed 'the factual premise', as it purports to state a fact
about the world. Briefly put, the fact in question is that there exist
instances of intense suffering which are gratuitous or pointless. As
indicated above, an instance of suffering is gratuitous, according to
Rowe, if an omnipotent, omniscient being could have prevented it
without thereby losing some greater good or permitting some evil
equally bad or worse. A gratuitous evil, in this sense, is a state of
affairs that is not (logically) necessary to the attainment of a
greater good or to the prevention of an evil at least as bad.
i Rowe's Case in Support of the Factual Premise
Rowe builds his case in support of the factual premise by appealing to
particular instances of human and animal suffering, such as the
following:
E1: the case of Bambi
"In some distant forest lightning strikes a dead tree, resulting
in a forest fire. In the fire a fawn is trapped, horribly burned, and
lies in terrible agony for several days before death relieves its
suffering" (Rowe 1979: 337).
Although this is presented as a hypothetical event, Rowe takes it to
be "a familiar sort of tragedy, played not infrequently on the stage
of nature" (1988: 119).
E2: the case of Sue
This is an actual event in which a five-year-old girl in Flint,
Michigan was severely beaten, raped and then strangled to death early
on New Year's Day in 1986. The case was introduced by Bruce Russell
(1989: 123), whose account of it, drawn from a report in the Detroit
Free Press of January 3 1986, runs as follows:
The girl's mother was living with her boyfriend, another man who
was unemployed, her two children, and her 9-month old infant fathered
by the boyfriend. On New Year's Eve all three adults were drinking at
a bar near the woman's home. The boyfriend had been taking drugs and
drinking heavily. He was asked to leave the bar at 8:00 p.m. After
several reappearances he finally stayed away for good at about 9:30
p.m. The woman and the unemployed man remained at the bar until 2:00
a.m. at which time the woman went home and the man to a party at a
neighbor's home. Perhaps out of jealousy, the boyfriend attacked the
woman when she walked into the house. Her brother was there and broke
up the fight by hitting the boyfriend who was passed out and slumped
over a table when the brother left. Later the boyfriend attacked the
woman again, and this time she knocked him unconscious. After checking
the children, she went to bed. Later the woman's 5-year old girl went
downstairs to go to the bathroom. The unemployed man returned from the
party at 3:45 a.m. and found the 5-year old dead. She had been raped,
severely beaten over most of her body and strangled to death by the
boyfriend.
Following Rowe (1988: 120), the case of the fawn will be referred to
as 'E1', and the case of the little girl as 'E2'. Further, following
William Alston's (1991: 32) practice, the fawn will be named 'Bambi'
and the little girl 'Sue'.
Rowe (1996: 264) states that, in choosing to focus on E1 and E2, he is
"trying to pose a serious difficulty for the theist by picking a
difficult case of natural evil, E1 (Bambi), and a difficult case of
moral evil, E2 (Sue)." Rowe, then, is attempting to state the
evidential argument in the strongest possible terms. As one
commentator has put it, "if these cases of evil [E1 and E2] are not
evidence against theism, then none are" (Christlieb 1992: 47).
However, Rowe's almost exclusive preoccupation with these two
instances of suffering must be placed within the context of his belief
(as expressed in, e.g., 1979: 337-38) that even if we discovered that
God could not have eliminated E1 and E2 without thereby losing some
greater good or permitting some evil equally bad or worse, it would
still be unreasonable to believe this of all cases of horrendous evil
occurring daily in our world. E1 and E2 are thus best viewed as
representative of a particular class of evil which poses a specific
problem for theistic belief. This problem is expressed by Rowe in the
following way:
(P) No good state of affairs we know of is such that an
omnipotent, omniscient being's obtaining it would morally justify that
being's permitting E1 or E2. Therefore,
(Q) It is likely that no good state of affairs is such that an
omnipotent, omniscient being's obtaining it would morally justify that
being in permitting E1 or E2.
P states that no good we know of justifies God in permitting E1 and
E2. From this it is inferred that Q is likely to be true, or that
probably there are no goods which justify God in permitting E1 and E2.
Q, of course, corresponds to the factual premise of Rowe's argument.
Thus, Rowe attempts to establish the truth of the factual premise by
appealing to P.
ii. The Inference from P to Q
At least one question to be addressed when considering this inference
is: What exactly do P and Q assert? Beginning with P, the central
notion here is 'a good state of affairs we know of'. But what is it to
know of a good state of affairs? According to Rowe (1988: 123), to
know of a good state of affairs is to (a) conceive of that state of
affairs, and (b) recognize that it is intrinsically good (examples of
states that are intrinsically good include pleasure, happiness, love,
and the exercise of virtue). Rowe (1996: 264) therefore instructs us
to not limit the set of goods we know of to goods that we know have
occurred in the past or to goods that we know will occur in the
future. The set of goods we know of must also include goods that we
have some grasp of, even if we do not know whether they have occurred
or ever will occur. For example, such a good, in the case of Sue, may
consist of the experience of eternal bliss in the hereafter. Even
though we lack a clear grasp of what this good involves, and even
though we cannot be sure that such a good will ever obtain, we do well
to include this good amongst the goods we know of. A good that we know
of, however, cannot justify God in permitting E1 or E2 unless that
good is actualized at some time.
On what grounds does Rowe think that P is true? Rowe (1988: 120)
states that "we have good reason to believe that no good state of
affairs we know of would justify an omnipotent, omniscient being in
permitting either E1 or E2" (emphasis his). The good reason in
question consists of the fact that the good states of affairs we know
of, when reflecting on them, meet one or both of the following
conditions: either an omnipotent being could obtain them without
having to permit E1 or E2, or obtaining them would not morally justify
that being in permitting E1 or E2 (Rowe 1988: 121, 123; 1991: 72).
This brings us, finally, to Rowe's inference from P to Q. This is, of
course, an inductive inference. Rowe does not claim to know or be able
to prove that cases of intense suffering such as the fawn's are indeed
pointless. For as he acknowledges, it is quite possible that there is
some familiar good outweighing the fawn's suffering and which is
connected to that suffering in a way unbeknown to us. Or there may be
goods we are not aware of, to which the fawn's suffering is intimately
connected. But although we do not know or cannot establish the truth
of Q, we do possess rational grounds for accepting Q, and these
grounds consist of the considerations adumbrated in P. Thus, the truth
of P is taken to provide strong evidence for the truth of Q (Rowe
1979: 337).
3. The Skeptical Theist Response
Theism, particularly as expressed within the Judeo-Christian and
Islamic religions, has always emphasized the inscrutability of the
ways of God. In Romans 11:33-34, for example, the apostle Paul
exclaims: "Oh, the depth of the riches of the wisdom and knowledge of
God! How unsearchable his judgments, and his paths beyond tracing out!
Who has known the mind of the Lord?" (NIV). This emphasis on mystery
and the epistemic distance between God and human persons is a
characteristic tenet of traditional forms of theism. It is in the
context of this tradition that Stephen Wykstra developed his
well-known CORNEA critique of Rowe's evidential argument. The heart of
Wykstra's critique is that, given our cognitive limitations, we are in
no position to judge as improbable the statement that there are goods
beyond our ken secured by God's permission of many of the evils we
find in the world. This position – sometimes labelled 'skeptical
theism' or 'defensive skepticism' – has generated a great deal of
discussion, leading some to conclude that "the inductive argument from
evil is in no better shape than its late lamented deductive cousin"
(Alston 1991: 61). In this Section, I will review the challenge posed
by this theistic form of skepticism, beginning with the critique
advanced by Wykstra.
a. Wykstra's CORNEA Critique
In an influential paper entitled, "The Humean Obstacle to Evidential
Arguments from Evil," Stephen Wykstra raised a formidable objection to
Rowe's inference from P to Q. Wykstra's first step was to draw
attention to the following epistemic principle, which he dubbed
'CORNEA' (short for 'Condition Of ReasoNable Epistemic Access'):
(C) On the basis of cognized situation s, human H is entitled to
claim 'It appears that p' only if it is reasonable for H to believe
that, given her cognitive faculties and the use she has made of them,
if p were not the case, s would likely be different than it is in some
way discernible by her. (Wykstra 1984: 85)
The point behind CORNEA may be easier to grasp if (C) is simplified
along the following lines:
(C*) H is entitled to infer 'There is no x' from 'So far as I can
tell, there is no x' only if:
It is reasonable for H to believe that if there were an x, it is
likely that she would perceive (or find, grasp, comprehend, conceive)
it.
Adopting terminology introduced by Wykstra (1996), the inference from
'So far as I can tell, there is no x' to 'There is no x' may be called
a 'noseeum inference': we no see 'um, so they ain't there! Further,
the italicized portion in (C*) may be called 'the noseeum assumption',
as anyone who employs a noseeum inference and is justified in doing so
would be committed to this assumption.
C*, or at least something quite like it, appears unobjectionable. If,
for instance, I am looking through the window of my twentieth-floor
office to the garden below and I fail to see any caterpillars on the
flowers, that would hardly entitle me to infer that there are in fact
no caterpillars there. Likewise, if a beginner were watching Kasparov
play Deep Blue, it would be unreasonable for her to infer 'I can't see
any way for Deep Blue to get out of check; so, there is none'. Both
inferences are illegitimate for the same reason: the person making the
inference does not have what it takes to discern the sorts of things
in question. It is this point that C* intends to capture by insisting
that a noseeum inference is permissible only if it is likely that one
would detect or discern the item in question if it existed.
But how does the foregoing relate to Rowe's evidential argument?
Notice, to begin with, that Rowe's inference from P to Q is a noseeum
inference. Rowe claims in P that, so far as we can see, no goods
justify God's permission of E1 and E2, and from this he infers that no
goods whatever justify God's permission of these evils. According to
Wykstra, however, Rowe is entitled to make this noseeum inference only
if he is entitled to make the following noseeum assumption:
If there are goods justifying God's permission of horrendous evil,
it is likely that we would discern or be cognizant of such goods.
Call this Rowe's Noseeum Assumption, or RNA for short. The key issue,
then, is whether we should accept RNA. Many theists, led by Stephen
Wykstra, have claimed that RNA is false (or that we ought to suspend
judgement about its truth). They argue that the great gulf between our
limited cognitive abilities and the infinite wisdom of God prevents us
(at least in many cases) from discerning God's reasons for permitting
evil. On this view, even if there are goods secured by God's
permission of evil, it is likely that these goods would be beyond our
ken. Alvin Plantinga (1974: 10) sums up this position well with his
rhetorical question: "Why suppose that if God does have a reason for
permitting evil, the theist would be the first to know?" (emphasis
his). Since theists such as Wykstra and Plantinga challenge Rowe's
argument (and evidential arguments in general) by focusing on the
limits of human knowledge, they have become known as skeptical
theists.
I will now turn to some considerations that have been offered by
skeptical theists against RNA.
b. Wykstra's Parent Analogy
Skeptical theists have drawn various analogies in an attempt to
highlight the implausibility of RNA. The most common analogy, and the
one favoured by Wykstra, involves a comparison between the vision and
wisdom of an omniscient being such as God and the cognitive capacities
of members of the human species. Clearly, the gap between God's
intellect and ours is immense, and Wykstra (1984: 87-91) compares it
to the gap between the cognitive abilities of a parent and her
one-month-old infant. But if this is the case, then even if there were
outweighing goods connected in the requisite way to the instances of
suffering appealed to by Rowe, that we should discern most of these
goods is just as likely as that a one-month-old infant should discern
most of her parents' purposes for those pains they allow her to suffer
– that is to say, it is not likely at all. Assuming that CORNEA is
correct, Rowe would not then be entitled to claim, for any given
instance of apparently pointless suffering, that it is indeed
pointless. For as the above comparison between God's intellect and the
human mind indicates, even if there were outweighing goods served by
certain instances of suffering, such goods would be beyond our ken.
What Rowe has failed to see, according to Wykstra, is that "if we
think carefully about the sort of being theism proposes for our
belief, it is entirely expectable – given what we know of our
cognitive limits – that the goods by virtue of which this Being allows
known suffering should very often be beyond our ken" (1984: 91).
c. Alston's Analogies
Rowe, like many others, has responded to Wykstra'a Parent Analogy by
identifying a number of relevant disanalogies between a one-month-old
infant and our predicament as adult human beings (see Rowe 1996: 275).
There are, however, various other analogies that skeptical theists
have employed in order to cast doubt on RNA. Here I will briefly
consider a series of analogies that were first formulated by Alston
(1996).
Like Wykstra, Alston (1996: 317) aims to highlight "the absurdity of
the claim" that the fact that we cannot see what justifying reason an
omniscient, omnipotent being might have for doing something provides
strong support for the supposition that no such reason is available to
that being. Alston, however, chooses to steer clear of the
parent-child analogy employed by Wykstra, for he concedes that this
contains loopholes that can be exploited in the ways suggested by
Rowe.
Alston's analogies fall into two groups, the first of which attempt to
show that the insights attainable by finite, fallible human beings are
not an adequate indication of what is available by way of reasons to
an omniscient, omnipotent being. Suppose I am a first-year university
physics student and I am faced with a theory of quantum phenomena, but
I struggle to see why the author of the theory draws the conclusions
she draws. Does that entitle me to suppose that she has no sufficient
reason for her conclusions? Clearly not, for my inability to discern
her reasons is only to be expected given my lack of expertise in the
subject. Similarly, given my lack of training in painting, I fail to
see why Picasso arranged the figures in Guernica as he did. But that
does not entitle me to infer that he had no sufficient reason for
doing so. Again, being a beginner in chess, I fail to see any reason
why Kasparov made the move he did, but I would be foolish to conclude
that he had no good reason to do so.
Alston applies the foregoing to the noseeum inference from 'We cannot
see any sufficient reason for God to permit E1 and E2' to 'God has no
sufficient reason to do so'. In this case, as in the above examples,
we are in no position to draw such a conclusion for we lack any reason
to suppose that we have a sufficient grasp of the range of possible
reasons open to the other party. Our grasp of the reasons God might
have for his actions is thus comparable to the grasp of the neophyte
in the other cases. Indeed, Alston holds that "the extent to which God
can envisage reasons for permitting a given state of affairs exceeds
our ability to do so by at least as much as Einstein's ability to
discern the reason for a physical theory exceeds the ability of one
ignorant of physics" (1996: 318, emphasis his).
Alston's second group of analogies seek to show that, in looking for
the reasons God might have for certain acts or omissions, we are in
effect trying to determine whether there is a so-and-so in a territory
the extent and composition of which is largely unknown to us (or, at
least, it is a territory such that we have no way of knowing the
extent to which its constituents are unknown to us). Alston thus
states that Rowe's noseeum inference
…is like going from 'We haven't found any signs of life elsewhere
in the universe' to 'There isn't life elsewhere in the universe.' It
is like someone who is culturally and geographically isolated going
from 'As far as I have been able to tell, there is nothing on earth
beyond this forest' to 'There is nothing on earth beyond this forest.'
Or, to get a bit more sophisticated, it is like someone who reasons
'We are unable to discern anything beyond the temporal bounds of our
universe,' where those bounds are the big bang and the final collapse,
to 'There is nothing beyond the temporal bounds of our universe.'
(1996: 318)
Just as we lack a map of the relevant 'territory' in these cases, we
also lack a reliable internal map of the diversity of considerations
that are available to an omniscient being in permitting instances of
suffering. But given our ignorance of the extent, variety, or
constitution of the terra incognita, it is surely the better part of
wisdom to refrain from drawing any hasty conclusions regarding the
nature of this territory.
Although such analogies may not be open to the same criticisms
levelled against the analogies put forward by Wykstra, they are in the
end no more successful than Wykstra's analogies. Beginning with
Alston's first group of analogies, where a noseeum inference is
unwarranted due to a lack of expertise, there is typically no
expectation on the part of the neophyte that the reasons held by the
other party (e.g., the physicist's reasons for drawing conclusion x,
Kasparov's reasons for making move x in a chess game) would be
discernible to her. If you have just begun to study physics, you would
not expect to understand Einstein's reasons for advancing the special
theory of relativity. However, if your five-year-old daughter suffered
the fate of Sue as depicted in E2, would you not expect a perfectly
loving being to reveal his reasons to you for allowing this to happen,
or at least to comfort you by providing you with special assurances
that that there is a reason why this terrible evil could not have been
prevented? Rowe makes this point quite well:
Being finite beings we can't expect to know all the goods God
would know, any more than an amateur at chess should expect to know
all the reasons for a particular move that Kasparov makes in a game.
But, unlike Kasparov who in a chess match has a good reason not to
tell us how a particular move fits into his plan to win the game, God,
if he exists, isn't playing chess with our lives. In fact, since
understanding the goods for the sake of which he permits terrible
evils to befall us would itself enable us to better bear our
suffering, God has a strong reason to help us understand those goods
and how they require his permission of the terrible evils that befall
us. (2001b: 157)
There appears, then, to be an obligation on the part of a perfect
being to not keep his intentions entirely hidden from us. Such an
obligation, however, does not attach to a gifted chess player or
physicist – Kasparov cannot be expected to reveal his game plan, while
a physics professor cannot be expected to make her mathematical
demonstration in support of quantum theory comprehensible to a high
school physics student.
Similarly with Alston's second set of analogies, where our inability
to map the territory within which to look for x is taken to preclude
us from inferring from our inability to find x that there is no x.
This may be applicable to cases like the isolated tribesman's search
for life outside his forest or our search for extraterrestrial life,
for in such scenarios there is no prior expectation that the objects
of our search are of such a nature that, if they exist, they would
make themselves manifest to us. However, in our search for God's
reasons we are toiling in a unique territory, one inhabited by a
perfectly loving being who, as such, would be expected to make at
least his presence, if not also his reasons for permitting evil,
(more) transparent to us. This difference in prior expectations
uncovers an important disanalogy between the cases Alston considers
and cases involving our attempt to discern God's intentions. Alston's
analogies, therefore, not only fail to advance the case against RNA
but also suggest a line of thought in support of RNA. (For further
discussion on RNA and divine hiddenness, see Trakakis (2003); see also
Howard-Snyder & Moser (2002).)
4. Building a Theodicy, or Casting Light on the Ways of God
Most critics of Rowe's evidential argument have thought that the
problem with the argument lies with its factual premise. But what,
exactly, is wrong with this premise? According to one popular line of
thought, the factual premise can be shown to be false by identifying
goods that we know of that would justify God in permitting evil. To do
this is to develop a theodicy.
a. What Is a Theodicy?
The primary aim of the project of theodicy may be characterized in
John Milton's celebrated words as the attempt to "justify the ways of
God to men." That is to say, a theodicy aims to vindicate the justice
or goodness of God in the face of the evil found in the world, and
this it attempts to do by offering a reasonable explanation as to why
God allows evil to abound in his creation.
A theodicy may be thought of as a story told by the theist explaining
why God permits evil. Such a story, however, must be plausible or
reasonable in the sense that it conforms to all of the following:
1. commonsensical views about the world (e.g., that there exist
other people, that there exists a mind-independent world, that much
evil exists);
2. widely accepted scientific and historical views (e.g.,
evolutionary theory), and
3. intuitively plausible moral principles (e.g., generally,
punishment should not be significantly disproportional to the offence
committed).
Judged by these criteria, the story of the Fall (understood in a
literalist fashion) could not be offered as a theodicy. For given the
doubtful historicity of Adam and Eve, and given the problem of
harmonizing the Fall with evolutionary theory, such an account of the
origin of evil cannot reasonably held to be plausible. A similar point
could be made about stories that attempt to explain evil as the work
of Satan and his cohorts.
b. Distinguishing a 'Theodicy' from a 'Defence'
An important distinction is often made between a defence and a
theodicy. A theodicy is intended to be a plausible or reasonable
explanation as to why God permits evil. A defence, by contrast, is
only intended as a possible explanation as to why God permits evil. A
theodicy, moreover, is offered as a solution to the evidential problem
of evil, whereas a defence is offered as a solution to the logical
problem of evil. Here is an example of a defence, which may clarify
this distinction:
It will be recalled that, according to Mackie, it is logically
impossible for the following two propositions to be jointly true:
1. God is omnipotent, omniscient, and perfectly good,
2. Evil exists.
Now, consider the following proposition:
3. Every person goes wrong in every possible world.
In other words, every free person created by God would misuse their
free will on at least one occasion, no matter which world (or what
circumstances) they were placed in. This may be highly implausible, or
even downright false – but it is, at least, logically possible. And if
(3) is possible, then so is the following proposition:
4. It was not within God's power to create a world containing moral
good but no moral evil.
In other words, it is possible that any world created by God that
contains some moral good will also contain some moral evil. Therefore,
it is possible for both (1) and (2) to be jointly true, at least when
(2) is said to refer to 'moral evil'. But what about 'natural evil'?
Well, consider the following proposition:
5. All so-called 'natural evil' is brought about by the devious
activities of Satan and his cohorts.
In other words, what we call 'natural evil' is actually 'moral evil'
since it results from the misuse of someone's free will (in this case,
the free will of some evil demon). Again, this may be highly
implausible, or even downright false – but it is, at least, possibly
true.
In sum, Mackie was wrong to think that it is logically impossible for
both (1) and (2) to be true. For if you conjoin (4) and (5) to (1) and
(2), it becomes clear that it is possible that any world created by
God would have some evil in it. (This, of course, is the famous free
will defence put forward in Plantinga 1974: ch.9). Notice that the
central claims of this defence – namely, (3), (4), and (5) – are only
held to be possibly true. That's what makes this a defence. One could
not get away with this in a theodicy, for a theodicy must be more than
merely possibly true.
c. Sketch of a Theodicy
What kind of theodicy, then, can be developed in response to Rowe's
evidential argument? Are there any goods we know of that would justify
God in permitting evils like E1 and E2? Here I will outline a proposal
consisting of three themes that have figured prominently in the recent
literature on the project of theodicy.
(1) Soul-making. Inspired by the thought of the early Church Father,
Irenaeus of Lyon (c.130-c.202 CE), John Hick has put forward in a
number of writings, but above all in his 1966 classic Evil and the God
of Love, a theodicy that appeals to the good of soul-making (see also
Hick 1968, 1977, 1981, 1990). According to Hick, the divine intention
in relation to humankind is to bring forth perfect finite personal
beings by means of a 'vale of soul-making' in which humans may
transcend their natural self-centredness by freely developing the most
desirable qualities of moral character and entering into a personal
relationship with their Maker. Any world, however, that makes possible
such personal growth cannot be a hedonistic paradise whose inhabitants
experience a maximum of pleasure and a minimum of pain. Rather, an
environment that is able to produce the finest characteristics of
human personality – particularly the capacity to love – must be one in
which "there are obstacles to be overcome, tasks to be performed,
goals to be achieved, setbacks to be endured, problems to be solved,
dangers to be met" (Hick 1966: 362). A soul-making environment must,
in other words, share a good deal in common with our world, for only a
world containing great dangers and risks, as well as the genuine
possibility of failure and tragedy, can provide opportunities for the
development of virtue and character. A necessary condition, however,
for this developmental process to take place is that humanity be
situated at an 'epistemic distance' from God. On Hick's view, in other
words, if we were initially created in the direct presence of God we
could not freely come to love and worship God. So as to preserve our
freedom in relation to God, the world must be created religiously
ambiguous or must appear, to some extent at least, as if there were no
God. And evil, of course, plays an important role in creating the
desired epistemic distance.
(2) Free will. The appeal to human freedom, in one guise or another,
constitutes an enduring theme in the history of theodicy. Typically,
the kind of freedom that is invoked by the theodicist is the
libertarian sort, according to which I am free with respect to a
particular action at time t only if the action is not determined by
all that happened or obtained before t and all the causal laws there
are in such a way that the conjunction of the two (the past and the
laws) logically entails that I perform the action in question. My
mowing the lawn, for instance, constitutes a voluntary action only if,
the state of the universe (including my beliefs and desires) and laws
of nature being just as they were immediately preceding my decision to
mow the lawn, I could have chosen or acted otherwise than I in fact
did. In this sense, the acts I perform freely are genuinely 'up to me'
– they are not determined by anything external to my will, whether
these be causal laws or even God. And so it is not open to God to
cause or determine just what actions I will perform, for if he does so
those actions could not be free. Freedom and determinism are
incompatible.
The theodicist, however, is not so much interested in libertarian
freedom as in libertarian freedom of the morally relevant kind, where
this consists of the freedom to choose between good and evil courses
of action. The theodicist's freedom, moreover, is intended to be
morally significant, not only providing one with the capacity to bring
about good and evil, but also making possible a range of actions that
vary enormously in moral worth, from great and noble deeds to horrific
evils.
Armed therefore with such a conception of freedom, the free will
theodicist proceeds to explain the existence of moral evil as a
consequence of the misuse of our freedom. This, however, means that
responsibility for the existence of moral evil lies with us, not with
God. Of course, God is responsible for creating the conditions under
which moral evil could come into existence. But it was not inevitable
that human beings, if placed in those conditions, would go wrong. It
was not necessary, in other words, that humans would misuse their free
will, although this always was a possibility and hence a risk inherent
in God's creation of free creatures. The free will theodicist adds,
however, that the value of free will (and the goods it makes possible)
is so great as to outweigh the risk that it may be misused in various
ways.
(3) Heavenly bliss. Theodicists sometimes draw on the notion of a
heavenly afterlife to show that evil, particularly horrendous evil,
only finds its ultimate justification or redemption in the life to
come. Accounts of heaven, even within the Christian tradition, vary
widely. But one common feature in these accounts that is relevant to
the theodicist's task is the experience of complete felicity for
eternity brought about by intimate and loving communion with God. This
good, as we saw, plays an important role in Hick's theodicy, and it
also finds a central place in Marilyn Adams' account of horrendous
evil.
Adams (1986: 262-63, 1999: 162-63) notes that, on the Christian
world-view, the direct experience of 'face-to-face' intimacy with God
is not only the highest good we can aspire to enjoy, but is also an
incommensurable good – more precisely, it is incommensurable with
respect to any merely temporal evils or goods. As the apostle Paul put
it, "our present sufferings are not worth comparing with the glory
that will be revealed in us" (Rom 8:18, NIV; cf. 2 Cor 4:17). This
glorification to be experienced in heaven, according to Adams,
vindicates God's justice and love toward his creatures. For the
experience of the beatific vision outweighs any evil, even evil of the
horrendous variety, that someone may suffer, thus ensuring a balance
of good over evil in the sufferer's life that is overwhelmingly
favourable. But as Adams points out, "strictly speaking, there will be
no balance to be struck" (1986: 263, emphasis hers), since the good of
the vision of God is incommensurable with respect to one's
participation in any temporal or created evils. And so an everlasting,
post-mortem beatific vision of God would provide anyone who
experienced it with good reason for considering their life – in spite
of any horrors it may have contained – as a great good, thus removing
any grounds of complaint against God.
Bringing these three themes together, a theodicy can be developed with
the aim of explaining and justifying God's permission of evil, even
evil of the horrendous variety. To illustrate how this may be done, I
will concentrate on Rowe's E2 and the Holocaust, two clear instances
of horrendous moral evil.
Notice that these two evils clearly involve a serious misuse of
free will on behalf of the perpetrators. We could, therefore, begin by
postulating God's endowment of humans with morally significant free
will as the first good that is served by these evils. That is to say,
God could not prevent the terrible suffering and death endured by Sue
and the millions of Holocaust victims while at the same time creating
us without morally significant freedom – the freedom to do both great
evil and great good. In addition, these evils may provide an
opportunity for soul-making – in many cases, however, the potential
for soul-making would not extend to the victim but only to those who
cause or witness the suffering. The phenomenon of 'jailhouse
conversions', for example, testifies to the fact that even horrendous
evil may occasion the moral transformation of the perpetrator.
Finally, to adequately compensate the victims of these evils we may
introduce the doctrine of heaven. Postmortem, the victims are ushered
into a relation of beatific intimacy with God, an incommensurable good
that 'redeems' their past participation in horrors. For the beatific
vision in the afterlife not only restores value and meaning to the
victim's life, but also provides them with the opportunity to endorse
their life (taken as a whole) as worthwhile.
Does this theodicy succeed in exonerating God? Various objections
could, of course, be raised against such a theodicy. One could, for
example, question the intelligibility or empirical adequacy of the
underlying libertarian notion of free will (see, e.g., Pereboom 2001:
38-88). Or one might follow Tooley (1980:373-75) and Rowe (1996:
279-81, 2001a: 135-36) in thinking that, just as we have a duty to
curtail another person's exercise of free will when we know that they
will use their free will to inflict considerable suffering on an
innocent (or undeserving) person, so too does God have a duty of this
sort. On this view, a perfectly good God would have intervened to
prevent us from misusing our freedom to the extent that moral evil,
particularly moral evil of the horrific kind, would either not occur
at all or occur on a much more infrequent basis. Finally, how can the
above theodicy be extended to account for natural evil? Various
proposals have been offered here, the most prominent of which are:
Hick's view that natural evil plays an essential part in the
'soul-making' process; Swinburne's 'free will theodicy for natural
evil' – the idea, roughly put, is that free will cannot be had without
the knowledge of how to bring about evil (or prevent its occurrence),
and since this knowledge of how to cause evil can only be had through
prior experience with natural evil, it follows that the existence of
natural evil is a logically necessary condition for the exercise of
free will (see Swinburne 1978, 1987: 149-67, 1991: 202-214, 1998:
176-92); and 'natural law theodicies', such as that developed by
Reichenbach (1976, 1982: 101-118), according to which the natural
evils that befall humans and animals are the unavoidable by-products
of the outworking of the natural laws governing God's creation.
5. Further Responses to the Evidential Problem of Evil
Let's suppose that Rowe's evidential argument from evil succeeds in
providing strong evidence in support of the claim that there does not
exist an omnipotent, omniscient, wholly good being. What follows from
this? In particular, would a theist who finds its impossible to fault
Rowe's argument be obliged to give up her theism? Not necessarily, for
at least two further options would be available to such a theist.
Firstly, the theist may agree that Rowe's argument provides some
evidence against theism, but she may go on to argue that there is
independent evidence in support of theism which outweighs the evidence
against theism. In fact, if the theist thinks that the evidence in
support of theism is quite strong, she may employ what Rowe (1979:
339) calls 'the G.E. Moore shift' (cf. Moore 1953: ch.6). This
involves turning the opponent's argument on its head, so that one
begins by denying the very conclusion of the opponent's argument. The
theist's counter-argument would then proceed as follows:
(not-3)
There exists an omnipotent, omniscient, wholly good being.
(2) An omniscient, wholly good being would prevent the occurrence
of any intense suffering it could, unless it could not do so without
thereby losing some greater good or permitting some evil equally bad
or worse.
(not-1) (Therefore) It is not the case that there exist instances
of horrendous evil which an omnipotent, omniscient being could have
prevented without thereby losing some greater good or permitting some
evil equally bad or worse.
Although this strategy has been welcomed by many theists as an
appropriate way of responding to evidential arguments from evil (e.g.,
Mavrodes 1970: 95-97, Evans 1982: 138-39, Davis 1987: 86-87, Basinger
1996: 100-103) – indeed, it is considered by Rowe to be "the theist's
best response" (1979: 339) – it is deeply problematic in a way that is
often overlooked. The G.E. Moore shift, when employed by the theist,
will be effective only if the grounds for accepting not-(3) [the
existence of the theistic God] are more compelling than the grounds
for accepting not-(1) [the existence of gratuitous evil]. The problem
here is that the kind of evidence that is typically invoked by theists
in order to substantiate the existence of God – e.g., the cosmological
and design arguments, appeals to religious experience – does not even
aim to establish the existence of a perfectly good being, or else, if
it does have such an aim, it faces formidable difficulties in
fulfilling it. But if this is so, then the theist may well be unable
to offer any evidence at all in support of not-(3), or at least any
evidence of a sufficiently strong or cogent nature in support of
not-(3). The G.E. Moore shift, therefore, is not as straightforward a
strategy as it initially seems.
Secondly, the theist who accepts Rowe's argument may claim that Rowe
has only shown that one particular version of theism – rather than
every version of theism – needs to be rejected. A process theist, for
example, may agree with Rowe that there is no omnipotent being, but
would add that God, properly understood, is not omnipotent, or that
God's power is not as unlimited as is usually thought (see, e.g.,
Griffin 1976, 1991). An even more radical approach would be to posit a
'dark side' in God and thus deny that God is perfectly good. Theists
who adopt this approach (e.g., Blumenthal 1993, Roth 2001) would also
have no qualms with the conclusion of Rowe's argument.
There are at least two problems with this second strategy. Firstly,
Rowe's argument is only concerned with the God of orthodox theism as
described in Section I.1 above, not the God of some other version of
theism. And so objections drawn from non-orthodox forms of theism fail
to engage with Rowe's argument (although such objections may be useful
in getting us to reconsider the traditional understanding of God). A
second problem concerns the worship-worthiness of the sort of deity
being proposed. For example, would someone who is not wholly good and
capable of evil be fit to be the object of our worship, total devotion
and unconditional commitment? Similarly, why place complete trust in a
God who is not all-powerful and hence not in full control of the
world? (To be sure, even orthodox theists will place limits on God's
power, and such limits on divine power may go some way towards
explaining the presence of evil in the world. But if God's power, or
lack thereof, is offered as the solution to the problem of evil – so
that the reason why God allows evil is because he doesn't have the
power to prevent it from coming into being – then we are faced with a
highly impotent God who, insofar as he is aware of the limitations in
his power, may be considered reckless for proceeding with creation.)
6. Conclusion
Evidential arguments from evil, such as those developed by William
Rowe, purport to show that, grounds for belief in God aside, the
existence of evil renders atheism more reasonable than theism. What
verdict, then, can be reached regarding such arguments? A brief answer
to this question may be provided by way of an overview of the
foregoing investigation.
Firstly, as was argued in Section II, the 'open theist' response to
Rowe's theological premise either runs the risk of diminishing
confidence in God or else is entirely compatible with the theological
premise. Secondly, the 'sceptical theist' objection to Rowe's
inference from inscrutable evil to pointless evil was examined in
Section III and was found to be inadequately supported. Thirdly,
various theodical options were canvassed in Section IV as a possible
way of refuting Rowe's factual premise, and it was found that a
theodicy that appeals to the goods of free will, soul-making, and a
heavenly afterlife may go some way in accounting for the existence of
moral evil. Such a theodicy, however, raises many further questions
relating to the existence of natural evil and the existence of so much
horrendous moral evil. And finally, as argued in Section V, the
strategy of resorting to the "G.E. Moore shift" faces the daunting
task of furnishing evidence in support of the existence of a perfect
being; while resorting to a non-orthodox conception of God dissolves
the problem of evil at the cost of corroding religiously significant
attitudes and practices such as the love and worship of God.
On the basis of these results it can be seen that Rowe's argument has
a strongly resilient character, successfully withstanding many of the
objections raised against it. Much more, of course, can be said both
in support of and against Rowe's case for atheism. Although it might
therefore be premature to declare any one side to the debate
victorious, it can be concluded that, at the very least, Rowe's
evidential argument is not as easy to refute as is often presumed.
7. References and Further Reading
Adams, Marilyn McCord. 1996. "Redemptive Suffering: A Christian
Solution to the Problem of Evil," in Robert Audi and William J.
Wainwright (eds), Rationality, Religious Belief, and Moral Commitment.
Ithaca, NY: Cornell University Press, pp.248-67.
Adams, Marilyn McCord. 1999. Horrendous Evils and the Goodness of God.
Melbourne: Melbourne University Press.
Alston, William P. 1991. "The Inductive Argument from Evil and the
Human Cognitive Condition," Philosophical Perspectives 5: 29-67.
Alston, William P. 1996. "Some (Temporarily) Final Thoughts on the
Evidential Arguments from Evil," in Daniel Howard-Snyder (ed.), The
Evidential Argument from Evil. Bloomington, IN: Indiana University
Press, pp.311-32.
Ascol, Thomas K. 2001. "Pastoral Implications of Open Theism," in
Douglas Wilson (ed.), Bound Only Once: The Failure of Open Theism.
Moscow, ID: Canon Press, pp.173-90.
Basinger, David. 1996. The Case for Freewill Theism: A Philosophical
Assessment. Downers Grove, IL: InterVarsity Press.
Blumenthal, David R. 1993. Facing the Abusing God: A Theology of
Protest. Louisville, KY: Westminster John Knox Press.
Boyd, Gregory A. 2000. God of the Possible: A Biblical Introduction to
the Open View of God. Grand Rapids, MI: Baker Books.
Boyd, Gregory A. 2003. Is God to Blame? Moving Beyond Pat Answers to
the Problem of Evil. Downers Grove, IL: InterVarsity Press.
Brown, Patterson. 1967. "God and the Good," Religious Studies 2: 269-76.
Christlieb, Terry. 1992. "Which Theisms Face an Evidential Problem of
Evil?" Faith and Philosophy 9: 45-64.
Chrzan, Keith. 1994. "Necessary Gratuitous Evil: An Oxymoron
Revisited," Faith and Philosophy 11: 134-37.
Davis, Stephen T. 1987. "What Good Are Theistic Proofs?" in Louis P.
Pojman (ed.), Philosophy of Religion: An Anthology. Belmont, CA:
Wadsworth, pp.80-88.
Draper, Paul. 1989. "Pain and Pleasure: An Evidential Problem for
Theists," Nous 23: 331-50.
Evans, C. Stephen. 1982. Philosophy of Religion: Thinking about Faith.
Downers Grove, IL: InterVarsity Press.
Griffin, David Ray. 1976. God, Power, and Evil: A Process Theodicy.
Philadelphia, PA: Westminster Press.
Griffin, David Ray. 1991. Evil Revisited: Responses and
Reconsiderations. Albany, NY: State University of New York Press.
Hasker, William. 2004. Providence, Evil and the Openness of God.
London: Routledge.
Hick, John. 1966. Evil and the God of Love, first edition. London: Macmillan.
Hick, John. 1968. "God, Evil and Mystery," Religious Studies 3: 539-46.
Hick, John. 1977. Evil and the God of Love, second edition. New York:
HarperCollins.
Hick, John. 1981. "An Irenaean Theodicy" and "Response to Critiques,"
in Stephen T. Davis (ed.), Encountering Evil: Live Options in
Theodicy, first edition. Edinburgh: T & T Clark, pp.39-52, 63-68.
Hick, John. 1990. Philosophy of Religion, fourth edition. Englewood
Cliffs, NJ: Prentice-Hall.
Hoffman, Joshua, and Gary S. Rosenkrantz. 2002. The Divine Attributes.
Oxford: Blackwell.
Howard-Snyder, Daniel, and Frances Howard-Snyder. 1999. "Is Theism
Compatible with Gratuitous Evil?" American Philosophical Quarterly 36:
115-29.
Howard-Snyder, Daniel, and Paul K. Moser (eds). 2002. Divine
Hiddenness: New Essays. Cambridge: Cambridge University Press.
Jordan, Jeff. 2001. "Blocking Rowe's New Evidential Argument from
Evil," Religious Studies 37: 435-49.
Mackie, J.L. 1955. "Evil and Omnipotence," Mind 64: 200-212.
Mavrodes, George I. 1970. Belief in God: A Study in the Epistemology
of Religion. New York: Random House.
McCloskey, H.J. 1960. "God and Evil," Philosophical Quarterly 10: 97-114.
McNaughton, David. 1994. "The Problem of Evil: A Deontological
Perspective," in Alan G. Padgett (ed.), Reason and the Christian
Religion: Essays in Honour of Richard Swinburne. Oxford: Clarendon
Press, pp.329-51.
Moore, G.E. 1953. Some Main Problems of Philosophy. London: George
Allen & Unwin.
Morris, Thomas V. 1987. Anselmian Explorations: Essays in
Philosophical Theology. Notre Dame, IN: University of Notre Dame
Press.
Morris, Thomas V. 1991. Our Idea of God: An Introduction to
Philosophical Theology. Downers Grove, IL: InterVarsity Press.
Nelson, Mark T. 1991. "Naturalistic Ethics and the Argument from
Evil," Faith and Philosophy 8: 368-79.
O'Connor, David. 1998. God and Inscrutable Evil: In Defense of Theism
and Atheism. Lanham, MD: Rowman & Littlefield.
Otte, Richard. 2002. "Rowe's Probabilistic Argument from Evil," Faith
and Philosophy 19: 147-71.
Pereboom, Derk. 2001. Living Without Free Will. Cambridge: Cambridge
University Press.
Peterson, Michael L. 1982. Evil and the Christian God. Grand Rapids,
MI: Baker Book House.
Peterson, Michael L. 1998. God and Evil: An Introduction to the
Issues. Boulder, CO: Westview Press.
Pinnock, Clark H., Richard Rice, John Sanders, William Hasker, and
David Basinger. 1994. The Openness of God: A Biblical Challenge to the
Traditional Understanding of God. Downers Grove, IL: InterVarsity
Press.
Plantinga, Alvin. 1974. The Nature of Necessity. Oxford: Clarendon Press.
Plantinga, Alvin. 1977. God, Freedom, and Evil. Grand Rapids, MI: Eerdmans.
Quinn, Philip L., and Charles Taliaferro (eds). 1997. A Companion to
Philosophy of Religion. Cambridge, MA: Blackwell.
Reichenbach, Bruce R. 1976. "Natural Evils and Natural Law: A Theodicy
for Natural Evils," International Philosophical Quarterly 16: 179-96.
Reichenbach, Bruce R. 1982. Evil and a Good God. New York: Fordham
University Press.
Reitan, Eric. 2000. "Does the Argument from Evil Assume a
Consequentialist Morality?" Faith and Philosophy 17: 306-19.
Rogers, Katherin A. 2000. Perfect Being Theology. Edinburgh: Edinburgh
University Press.
Roth, John K. 2001. "A Theodicy of Protest", in Stephen T. Davis
(ed.), Encountering Evil: Live Options in Theodicy, second edition.
Louisville, KY: Westminster John Knox Press, pp.1-20.
Rowe, William L. 1978. Philosophy of Religion: An Introduction, first
edition. Encino, CA: Dickenson Publishing Company..
Rowe, William L. 1979. "The Problem of Evil and Some Varieties of
Atheism," American Philosophical Quarterly 16: 335-41.
Rowe, William L. 1986. "The Empirical Argument from Evil," in Audi and
Wainwright (eds), Rationality, Religious Belief, and Moral Commitment,
pp.227-47.
Rowe, William L. 1988. "Evil and Theodicy," Philosophical Topics 16: 119-32.
Rowe, William L. 1991. "Ruminations about Evil," Philosophical
Perspectives 5: 69-88.
Rowe, William L. 1995. "William Alston on the Problem of Evil," in
Thomas D. Senor (ed.), The Rationality of Belief and the Plurality of
Faith: Essays in Honor of William P. Alston. Ithaca, NY: Cornell
University Press, pp.71-93.
Rowe, William L. 1996. "The Evidential Argument from Evil: A Second
Look," in Daniel Howard-Snyder (ed.), The Evidential Argument from
Evil, pp.262-85.
Rowe, William L. 2001a. "Grounds for Belief Aside, Does Evil Make
Atheism More Reasonable than Theism" in William Rowe (ed.), God and
the Problem of Evil. Malden, MA: Blackwell, pp.124-37.
Rowe, William L. 2001b. "Reply to Howard-Snyder and Bergmann," in Rowe
(ed.), God and the Problem of Evil, pp.155-58.
Russell, Bruce. 1989. "The Persistent Problem of Evil," Faith and
Philosophy 6: 121-39.
Russell, Bruce. 1996. "Defenseless," in Daniel Howard-Snyder (ed.),
The Evidential Argument from Evil, pp.193-205.
Sanders, John. 1998. The God Who Risks: A Theology of Providence.
Downers Grove, IL: InterVarsity Press.
Sennett, James F. 1993. "The Inscrutable Evil Defense Against the
Inductive Argument from Evil," Faith and Philosophy 10: 220-29.
Stump, Eleonore. 1997. "Review of Peter van Inwagen, God, Knowledge,
and Mystery," Philosophical Review 106: 464-67.
Swinburne, Richard. 1977. The Coherence of Theism. Oxford: Clarendon Press.
Swinburne, Richard. 1978. "Natural Evil," American Philosophical
Quarterly 15: 295-301.
Swinburne, Richard. 1987. "Knowledge from Experience, and the Problem
of Evil," in William J. Abraham and Steven W. Holtzer (eds), The
Rationality of Religious Belief: Essays in Honour of Basil Mitchell.
Oxford: Clarendon Press, pp.141-67.
Swinburne, Richard. 1991. The Existence of God, revised edition.
Oxford: Clarendon Press.
Swinburne, Richard. 1998. Providence and the Problem of Evil. Oxford:
Clarendon Press.
Tooley, Michael. 1980. "Alvin Plantinga and the Argument from Evil,"
Australasian Journal of Philosophy 58: 360-76.
Trakakis, Nick. 2003. "What No Eye Has Seen: The Skeptical Theist
Response to Rowe's Evidential Argument from Evil," Philo 6: 263-79.
Van Inwagen, Peter. 1988. "The Place of Chance in a World Sustained by
God," in Thomas V. Morris (ed.), Divine and Human Action. Ithaca, NY:
Cornell University Press, pp.211-35.
Ware, Bruce. 2000. God's Lesser Glory: The Diminished God of Open
Theism. Wheaton, IL: Crossways Books.
Wykstra, Stephen 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, Stephen J. 1986. "Rowe's Noseeum Arguments from Evil," in
Daniel Howard-Snyder (ed.), The Evidential Argument from Evil,
pp.126-50.
No comments:
Post a Comment