Foreknowledge and Freedom, Theological Problem of
FOREKNOWLEDGE AND FREEDOM, THEOLOGICAL PROBLEM OF
Divine foreknowledge, like the other classical theistic attributes, raises philosophical problems of at least three kinds. First, there are problems with understanding the attribute itself. How should it be construed (assuming that it is even coherent)? And how might God come by such knowledge? (Are future events all present in their causes? Does God arrive at foreknowledge by inference from "middle knowledge"? Does he see the future as through a "time telescope"? Or does he just know it?) Second, there are questions about how this attribute can be compatible with the other divine attributes. As the creator, sustainer, and providential overseer of the world, for example, God is supposed to be the supreme agent—but how can God approach the future as an active agent if his foreknowledge presents to him everything, including his own decisions and engagements with the world, as a fait accompli?
Finally, there are problems reconciling God's possession of this attribute with other things that appear undeniable. Of these, the most important is surely human freedom. If God knows before a person is even born exactly what that person will do throughout life, how could this person nevertheless retain the power to do otherwise, as free agency apparently requires? This is the classic foreknowledge problem; efforts to solve it are often what drive proposed solutions to the other two problems.
In De Interpretatione, Aristotle worried that accepting the truth of future contingents would result in a necessitarianism incompatible with human freedom; for if it is true either that there will be a sea battle tomorrow or that there will not be a sea battle tomorrow, the admiral on whose decision this event depends either cannot issue the requisite order (if there will not be a sea battle) or cannot refrain from issuing the order (if there will be a sea battle). A similar worry was later elaborated into the influential "Master Argument" of Diodorus Cronus, discussed by the Stoics. Because this threat to human freedom rests solely on logical principles, like the Law of Excluded Middle, it is often called "logical fatalism" in contrast to the "theological fatalism" generated by divine foreknowledge.
The subtheistic nature of the ancient divinities and the pluralism of pagan theology made the problem of theological fatalism avoidable, but this was to change with the advent of Christianity. Augustine provides a classic early exposition of the problem in On Free Choice of the Will (III.3):
How is it that these two propositions are not contradictory and inconsistent: (1) God has foreknowledge of everything in the future; and (2) We sin by the will, not by necessity? For … if God foreknows that someone is going to sin, then it is necessary that he sin. But if it is necessary, the will has no choice about whether to sin.… [So:] either we draw the heretical conclusion that God does not foreknow everything in the future; or … we must admit that sin happens by necessity and not by will.
Augustine went on to offer his own solution to this problem; his medieval successors added further solutions and contributed enormously to the understanding of the problem, especially its modal character. Recent interest in the problem, sparked by a 1965 article by Nelson Pike, is probably as strong as it has been since the problem's heyday in the Middle Ages.
Formulating the Problem
As Augustine notes, the argument for theological fatalism is designed to show that a certain assumption about God is incompatible with a certain assumption about free will, so that one of them must be rejected.
the god assumption
The theological assumptions that play an actual role in the argument concern God's existence and cognitive excellence. It is assumed in the first place that God knows all truths, or
(i) God is omniscient.
Moreover, God believes only truths; indeed, he not only does not but could not believe any falsehoods. So
(ii) God is essentially inerrant, that is, infallible.
The final assumption about God is
(iii) God exists "from eternity."
The phrase from eternity is purposely ambiguous, straddling the view of God as an everlasting temporal being existing at all points in time (sempiternity) and the view of God as an atemporal being whose existence transcends time altogether (eternity proper). If (iii) is read, "There is no time such that the proposition God exists, if asserted at that time, would be false," then both views are accommodated. This allows for disambiguation, if necessary, to occur in the argument itself.
the freedom assumption
The assumption with which the God Assumption is supposed to be incompatible is simply this:
Someone sometime does something freely.
Freely should be understood here in whatever sense is required for morally responsible agency, but otherwise pretheoretically—that way the theory of freedom under which it is allegedly incompatible with the God Assumption can emerge as a premise in the argument, and rejection of that premise can count as a solution to the problem.
Suppose someone X performs an action A at a time T3. Let T2 be a time prior to X's birth and T1 any time prior to T2. Then
(1) It is true at T1 that X will do A at T3.
The principle underwriting this claim, sometimes called the omnitemporality of truth, is that a statement true at any time is (suitably modified) true at every time. This does not imply, in the case of (1), that anyone can know at T1 what X will do at T3, let alone that there are conditions at T1 sufficient for X's future action; it only says that, since X's doing A at T3 is an assumption of the argument, it is, at T1, true that X will do A at T3.
According to clauses (i) and (iii) of the God Assumption, an omniscient God who exists "from eternity" must know at T1 whatever is true at T1. So
(2) God knows at T1 that X will do A at T3.
And if (2) is true, then so is
(3); God believes at T1 that X will do A at T3.
This follows from the standard analysis of knowledge, according to which knowledge entails belief.
Once God holds this belief, it becomes part of the fixed past that he held that belief. It is no longer possible for him not to have held this belief. This is an instance of the "necessity of the past," conveyed in such maxims as "what's done is done." This is not logical necessity, since there are logically possible worlds with a different past; but it is arguably stronger than natural or causal necessity. Aristotle notes that "this alone is lacking even to God, to make undone things that have once been done" (Nicomachean Ethics VI.2.1139b10–11), and Aquinas comments, "As such it is more impossible than the raising of the dead to life, which implies no contradiction, and is called impossible only according to natural power" (Summa Theologiae I.25.4). Because what is necessary when past might have been nonnecessary or accidental when future, it is often called accidental necessity. The next step in the argument can therefore be stated this way:
(4) It is accidentally necessary at T2 that God believed at T1 that X will do A at T3.
Since T1 is past relative to T2, (4) is true, given (3).
Though accidental necessity was introduced as a modality characteristic of the past, it is more general in scope. For a proposition p to be accidentally necessary at a time T is for p to be true no matter how the world continues after T. The past is then accidentally necessary by default; but the future can also qualify as accidentally necessary if entailed by accidentally necessary facts about the past. One such fact is the following:
(5) It is accidentally necessary at T2 that X will do A at T3.
This follows from (4) combined with clause (ii) of the God Assumption, according to which God's believing that X will do A at T3 entails that X will do A at T3.
Since T2 is a time prior to X's birth, X comes into existence with it already being the case that he must do A at T3. It is therefore too late for X to bring it about that he fails to do A at T3—that is,
(6) X cannot refrain from doing A at T3.
But if X cannot refrain from doing A at T3, then
(7) X does not do A at T3 freely.
This last inference is sanctioned by a "freedom version" of the so-called Principle of Alternate Possibilities, according to which a person is morally responsible for performing an action only if the person could have refrained from performing it. If a person is not morally responsible, owing to an inability to refrain, this person is not free in the sense required for moral responsibility. This is precisely the sense of "free" that is relevant to the Freedom Assumption.
The foregoing argument does not turn on any peculiar features of X, A, or T3; the same argument can be given for any agent, action, and time. So no one ever does anything freely. If the God Assumption is true, the Freedom Assumption is false.
Some Comments on the Argument
Before canvassing possible responses to this argument, some explanatory remarks are in order.
First, some versions of the argument bypass (4), inferring the necessity of X's future action from (3) and divine infallibility alone. Such versions might succeed if clause (ii) of the God Assumption could be parsed this way:
(iia) If God believes that p, then necessarily p.
Unfortunately, the correct analysis of divine infallibility is
(iib) Necessarily, if God believes that p, then p.
And all that follows from (iib), given simply that God believes that X will do A at T3, is that X will do A at T3 (and will do so in any world in which God holds this belief). For the action to be necessary, based on (iib), God's belief must be necessary. The illusion that (5) can be derived without reliance on (4) is produced by an equivocation between (iia) and (iib). Boethius, who called the necessity in (iia) "simple necessity" and the necessity in (iib) "conditional necessity," and Aquinas, who termed these "the necessity of the consequent" and "the necessity of the consequence" respectively, diagnosed the problem accurately and rightly insisted on the ineliminability of (4).
Second, step (4) does not rest on the simplistic principle that all true statements indexed to the past through tense or temporal references like "at T1" are accidentally necessary. This principle is in fact false. Confident of victory in tomorrow's election, the candidate proclaims, "My campaign for President began two years before its successful completion." Having just been fooled, I vow, "That was the last time I'm falling for that trick!" Suppose these declarations are in fact true. Though both assert something about the past, neither one is accidentally necessary, since either could (though ex hypothesi it won't) turn out false: The candidate might lose, I might get fooled again. Statements like these, which are not genuinely and strictly about the past, are called "soft facts" about the past as opposed to the "hard facts" to which the necessity of the past is applicable. What justifies (4), then, is that (3) looks like a hard fact about the past. (Certainly there is little question about the human analogue: If Joe believed yesterday that he will shave tomorrow, it is a hard fact, and therefore accidentally necessary, today that he held this belief yesterday.) This also explains the apparently trivial move from (2) to (3). To say of God's cognition in (2) that it constitutes knowledge is to say, in part, that it is true; but its truth depends on how things go at T3. So (2) is not strictly about the past; unlike (3), it is not a hard fact relative to T2.
Third, some critics point out that the future-truth argument for logical fatalism also begins with (1), but then moves directly to (4') It is accidentally necessary at T2 that it was true at T1 that X will do A at T3, and thence to (5). Their claim is that the argument for theological fatalism is just a needlessly complicated version of this argument, and is equally fallacious. The problem with this critique is that (1) is a paradigmatic soft fact relative to T2, undermining the inference to (4'), whereas routing the argument through the theological premises (2) and (3) allows (4) to follow from a prima facie hard fact about the past. This gives the argument for theological fatalism a clear logical advantage.
If the argument succeeds, either the God Assumption or the Freedom Assumption must be rejected. Those who deny the Freedom Assumption in response to the argument are "theological fatalists." There appear to be very few theological fatalists in this sense. Calvinists would qualify if anyone would. But most Calvinists are compatibilists and would therefore affirm the Freedom Assumption, while those who do reject the Freedom Assumption tend to do so on grounds other than the argument for theological fatalism.
Denying the God Assumption does not entail atheism unless the falsity of just one of the three clauses constituting the God Assumption is sufficient for there being no God. Some theists, indeed, deny that clause (i) is essential to theism when omniscience includes future contingents. If the argument succeeds, such truths are logically unknowable and should be excluded from divine omniscience, just as the logically impossible is excluded from divine omnipotence. "Open Theists" sometimes take this position, maintaining that God willingly limits his foreknowledge to make space for human freedom. There are, however, a number of reasons for thinking that the argument does not succeed.
the aristotelian solution
Step (1) has been rejected on the grounds that a statement about the future is not (yet) either true or false; it acquires a truth value only when what is now future becomes present. This seems to have been the position Aristotle adopted in response to the "future truth" argument for logical fatalism. It is also the favored position of Open Theists who prefer not to deny the God Assumption: If future contingents lack truth value, a deity who fails to foreknow them will not thereby lack anything necessary to omniscience. Critics, however, have pointed to serious logical costs associated with this move.
the boethian solution
Step (2) follows from (1) only if God exists at T1. But if God does not exist in time, a view famously associated with Boethius, (2) is false; what is true instead is:
(2*) God (timelessly) knows that X will do A at T3.
Two questions may be raised here. The first is whether this view of God is coherent: Though it is the classical view, it has come in for increasing criticism in recent years. The second question is whether a timeless deity might succumb to a modified version of the argument. It has been claimed, for example, that (2) can be replaced by:
(2#) It is true at T1 that God (timelessly) knows that X does A at T3.
(3) and (4) can be similarly modified, and (5) will then follow as before. It has also been claimed that what is fixed in eternity may be no less accidentally necessary than what is fixed in the past, so that (2*) leads to:
(4*) It is accidentally necessary at T2 that God (timelessly) believes that X will do A at T3.
and thence again to (5). But intuitions are a fragile guide here, and the viability of the Boethian solution remains open.
the ockhamist solution
The most popular solution in the contemporary debate is the denial of (4). A radical critique might challenge the very idea of accidental necessity as a modality characteristic of the past; but this extreme position runs counter to deep intuitions about the necessity of the past. The principal assault has come from those who accept the necessity of the past but argue, following William Ockham, that (3) is really a soft fact about the past.
In his treatise Predestination, God's Foreknowledge and Future Contingents, Ockham distinguishes hard and soft facts this way: "Some propositions are about the [past] as regards both their wording and their subject matter. … Other propositions are about the [past] as regards their wording only and are equivalently about the future, since their truth depends on the truth of propositions about the future." Ockham's modern followers have cited at least four grounds for placing (3) among the latter propositions.
First, God's belief that X will do A at T3 is counterfactually dependent on X's doing A at T3; if X were to do otherwise, God would have believed otherwise. Unfortunately, this counterfactual dependence can obtain even if X cannot do otherwise; hence it provides no reason to think that God can still believe otherwise.
Second, one might develop necessary and sufficient conditions for hard facthood and show that (3) does not qualify. If this is done in terms of an "entailment criterion," it appears that (3) is a soft fact after all, since it entails the future fact that X will do A at T3. But analyses of the hard/soft distinction, most employing entailment criteria, have grown mind-numbingly complex in response to counterexamples, and none has won consensus. This strategy has fallen into disfavor.
Third, one might approach the question from the side of the divine beliefs. How should divine cognition be construed so that (3) can be a soft fact relative to T2? Perhaps the "narrow content" of God's belief is a hard fact about the past, but its "wide content" is determined by the way the future actually unfolds—that the belief counts as the belief that p might then be a soft fact about the past. Or perhaps God's beliefs about future contingents are dispositional rather than occurrent in nature, and this makes a difference to their status as hard or soft; then God might be (dispositionally) omniscient at the same time as future contingents remain contingent. Or perhaps, as William Alston (1986) argues, God does not even have beliefs—a position which Linda Zagzebski (1991) terms Thomistic Ockhamism. Even if coherent, such proposals appear to make God's foreknowledge unavailable to him for action-guidance.
Fourth, one might finesse the difficulties of the above approaches with a direct demonstration that (3) is a soft fact, as suggested by Alvin Plantinga (1986) and (in another form) Ted Warfield (1997). If God exists necessarily, then (3) is true in all and only the worlds in which (1) is true, making (3) logically equivalent to (1). Since (1) is a paradigmatic soft fact relative to T2, (3) must be a soft fact as well, and (4) no longer follows. Critics, however, have charged this argument with question-begging.
the scotist solution
The inference from (4) to (5) has this form:
(4) It is accidentally necessary that M
(iib) Necessarily, if M, then N
(5) It is accidentally necessary that N
This is a so-called transfer principle, since it transfers necessity from one proposition to another. Whether the inference is valid depends on the logic of accidental necessity. The parallel inference for logical necessity is certainly valid; if accidental-necessity-at-T can be modeled as truth in all of some subset of logically possible worlds—for example, the set of all worlds that share the same past up to T—then the above inference should be valid as well. Nevertheless, some types of necessity appear not to work like this, and similar transfer principles, like Peter Van Inwagen's (1983) "rule β," have been disputed.
the edwardsian solution
A compatibilist about free will and causal determinism will not agree that (5) is a reason for endorsing (6). The case for (and against) compatibilism is too large a subject to be broached here and is best pursued in connection with the problem of freedom vs. determinism, where it has received its most sophisticated development. Among theists who come to compatibilism from theological rather than causal determinism, Jonathan Edwards is notable for rejecting step (6).
the augustinian solution
Augustine seems to have argued that the agent might remain free even if divine foreknowledge closes all alternatives, so long as the agent's action is self-initiated and God's foreknowledge does not cause, compel, or otherwise explain the action. (How this fits with what Augustine says about divine grace, sovereignty, and predestination is another question.) The moral Augustine draws from foreknowledge cases is arguably the same moral that Harry Frankfurt (1969) draws from cases in which a mechanism eliminates an agent's alternatives without interfering with the agent's actual course of action; indeed, when divine foreknowledge is the mechanism, the result appears to be a perfect "Frankfurt-type counterexample" to the Principle of Alternate Possibilities, on which step (7) rests. If, however, only a predetermined future can be foreknown, even by God, then this solution fails.
The Nature of the Problem
There are a number of philosophical problems in the neighborhood of this one that can be approached simply as thought experiments, without regard to whether the world is arranged as the problem presupposes. These include Newcomb's puzzle, the paradoxes of time travel and retrocausation, and perhaps even causal determinism itself. The problem of theological fatalism might be one of these; if it is, certain solutions become irrelevant. If someone reflecting on Zeno's paradoxes of motion is puzzled about how Achilles could fail to pass the tortoise, the puzzlement is not addressed by denying Achilles' existence or by reconceiving his attributes (for example, by making him a cripple or supposing he is in a coma). Likewise, someone reflecting on the argument for theological fatalism might be puzzled about how a paradigmatic candidate for free agency might be rendered unfree simply by adding infallible foreknowledge to the situation. Reconceiving God or denying God's existence outright simply removes God from complicity in this puzzle; it does not solve the puzzle. The purely theological solutions—the Boethian and the third and fourth of the Ockhamist responses—fail to address this deeper puzzle, assuming that it is genuine.
See also Alston, William; Aristotle; Augustine, St.; Boethius, Anicius Manlius Severinus; Diodorus Cronus; Duns Scotus, John; Edwards, Jonathan; Frankfurt, Harry; Freedom; Plantinga, Alvin; Precognition; Stoicism; Thomas Aquinas, St.; William of Ockham.
Adams, Marilyn McCord. "Is the Existence of God a 'Hard' Fact?" Philosophical Review 76 (1967): 492–503.
Alston, William P. "Does God Have Beliefs?" Religious Studies 22 (1986): 287–306.
Craig, William Lane. The Only Wise God. Grand Rapids, MI: Baker Book House, 1987.
Fischer, John Martin. "Freedom and Foreknowledge." Philosophical Review 92 (1983): 67–79.
Fischer, John Martin. "Scotism." Mind 94 (1985): 231–243.
Frankfurt, Harry. "Alternate Possibilities and Moral Responsibility." Journal of Philosophy 46 (1969): 828–839.
Freddoso, Alfred J. "Accidental Necessity and Logical Determinism." Journal of Philosophy 80 (1983): 257–278.
Haack, Susan. "On a Theological Argument for Fatalism." Philosophical Quarterly 24 (1974): 156–159.
Hasker, William. God, Time, and Knowledge. Ithaca, NY: Cornell University Press, 1989.
Hunt, David P. "The Compatibility of Divine Determinism and Human Freedom: A Modest Proposal." Faith and Philosophy 19 (2002): 485–502.
Hunt, David P. "Dispositional Omniscience." Philosophical Studies 80 (1995): 243–278.
Hunt, David P. "On Augustine's Way Out." Faith and Philosophy 16 (1999): 3–26.
Hunt, David P. "What Is the Problem of Theological Fatalism?" International Philosophical Quarterly 38 (1998): 17–30.
Kvanvig, Jonathan. The Possibility of an All–Knowing God. New York: St. Martin's Press, 1986.
Pike, Nelson. "Divine Omniscience and Voluntary Action." Philosophical Review 74 (1965): 27–46.
Plantinga, Alvin. "On Ockham's Way Out." Faith and Philosophy 3 (1986): 235–269.
Purtill, Richard L. "Fatalism and the Omnitemporality of Truth." Faith and Philosophy 5 (1988): 185–192.
Stump, Eleonore, and Norman Kretzmann. "Eternity." Journal of Philosophy 78 (1981): 429–458.
Van Inwagen, Peter. An Essay on Free Will. Oxford: Clarendon Press, 1983.
Warfield, Ted. "Divine Foreknowledge and Human Freedom Are Compatible." Nous 31 (1997): 80–86.
Widerker, David. "Troubles with Ockhamism." Journal of Philosophy 87 (1990): 462–480.
Widerker, David. "Two Forms of Fatalism." In God, Foreknowledge, and Freedom, edited by J. M. Fischer, 97–110. Stanford, CA: Stanford University Press, 1989.
Zemach, Eddy, and David Widerker. "Facts, Freedom, and Foreknowledge." Religious Studies 23 (1988): 19–28.
David P. Hunt (2005)