Divine Foreknowledge and Freedom

Tomorrow morning you will either have coffee or you will not, and the choice will be yours. Now suppose that a thousand years ago God already believed that you would have the coffee. Nothing about that belief pushed you. It exerted no force, delivered no instruction, and left no trace on the world you woke up in. It just sat there in the past being true.

The trouble is what happens when you try to describe the alternative. If you are genuinely able to skip the coffee, then you are able to act in such a way that a belief held a thousand years ago was false. But that belief is over and done with. It is not the sort of thing anyone can now do anything about. So it looks as though you cannot skip the coffee, and the reason has nothing to do with your psychology, your character, or the laws of physics. It is that the past has already had its say.

This is one of the oldest problems in philosophical theology, and it is unusual among the problems in this tier in that it is not an argument against God's existence. It is an argument that two claims most theists want to hold cannot both be held. Every response to it gives something up.

Setting out the argument

The modern discussion begins with Nelson Pike's "Divine Omniscience and Voluntary Action", in the Philosophical Review in 1965, which set out the problem in a form precise enough to argue about. Pike's contribution was to strip away everything inessential. He did not appeal to determinism, to God's causing anything, or to any claim about time beyond the ordinary one that the past is settled. He needed only that God is essentially omniscient, so that God cannot hold a false belief, and that God's holding a belief at a time is a fact about that time.

The engine of the argument is a notion from medieval logic that the tradition kept ready for exactly this purpose.

Aristotle registers the underlying thought in book six of the Nicomachean Ethics, where he quotes the poet Agathon to the effect that not even a god can make what has been done not to have been done. The medievals turned that observation into a technical notion and used it in exactly this debate, and the whole difficulty is that God's past beliefs look like they qualify.

Every premise here is doing work, and every published solution to the problem is the denial of one of them. That is the useful thing about having the argument written out: the solutions stop looking like different moods and start looking like a small, complete set of options.

What the argument is not

Two confusions have to be cleared before the responses make sense, because both make the problem look easier than it is.

The first is the idea that the argument accuses God of interference. It does not. Nothing in the argument says that God's belief made you do anything, and it is no reply to point out that a belief is not a cause. The claim is about entailment and about power, not about causal influence. A perfect weather forecaster who never causes any weather still generates the puzzle if her forecasts are infallible and recorded in the past.

The second is that the argument is just logical fatalism dressed up. Logical fatalism runs from the claim that it was already true a thousand years ago that you would have the coffee, and there are respectable ways to resist it, most of which turn on denying that a truth about the future is a fact about the past at all. The theological version cannot be handled that way, because a belief is not merely a truth. It is a state that somebody was in at a time, and states that somebody was in at a time are as past as anything gets. That is what makes the theological version the harder one, and it is why philosophers who are relaxed about logical fatalism can still find this problem serious.

It is also worth noticing which conception of freedom the argument targets. It attacks the ability to do otherwise. Someone who holds that freedom is a matter of acting on one's own desires without external constraint can accept the whole argument with a shrug, since nothing in it says your act fails to flow from your character. The problem is a problem for those who think freedom requires alternatives genuinely open at the moment of choosing, which is the conception the free will defence needs.

The Boethian route

The oldest answer denies the first premise by denying that God ever believed anything a thousand years ago. Boethius, and Aquinas after him, hold that God has no temporal location at all, so there is no past time at which God held a belief and therefore no accidentally necessary fact to be stuck with. Divine timelessness, met in rung 4, arrives here as a solution to a different problem.

The picture is that God knows what you will do the way you know what is in front of you, not by inference from what came earlier but by having it present. Aquinas puts the point in terms of vision in the first part of the Summa Theologiae, question 14, article 13: what God sees is your act as actually occurring, and seeing an act occur does not make it necessary, since anything that is actually occurring is occurring. The necessity is conditional and harmless.

Pike's reply, and it has been the standard one since, is that the move relocates the problem without removing it. If God timelessly knows that you will have coffee, then either that timeless knowledge is fixed or it is not. If it is fixed, then your power to refrain is the power to make a fixed item different, which is the same predicament in new vocabulary. If it is not fixed, then it was not settled after all and the timelessness has bought nothing.

Linda Zagzebski develops this in The Dilemma of Freedom and Foreknowledge (Oxford University Press, 1991), arguing that a notion of necessity parallel to accidental necessity applies to the timeless realm: God's timeless knowledge is no more within your power than the past is, so a timeless analogue of the argument goes through. Her book examines the Boethian, Ockhamist, and Molinist solutions in their best-known forms, finds each inadequate as usually stated, and then proposes revised versions. The interesting result is not the verdict on any one solution but the demonstration that the problem is robust: it survives translation into every framework offered to dissolve it.

Timelessness also has the costs catalogued in rung 4, and it sits badly with the personalist conception from rung 5. A solution available only to classical theists is not much use to the analytic philosophers of religion who feel the problem most keenly.

Ockhamism

The label comes from William of Ockham's treatise on predestination, divine foreknowledge, and future contingents, which distinguishes propositions that are about the present or past in the strict sense from those that are verbally about the past but really about the future. The distinction was brought into the modern debate in Marilyn McCord Adams's "Is the Existence of God a 'Hard' Fact?", in the Philosophical Review in 1967, which supplies the hard fact and soft fact vocabulary the literature has used since.

The proposal has an appealing consequence: you retain a kind of power over the past. Not the power to change it, which nobody has ever claimed, but the power such that had you exercised it, the past would always have been different. Alvin Plantinga defends this in "On Ockham's Way Out", in Faith and Philosophy in 1986, arguing that counterfactual power over the past is not the incoherent thing it sounds like, and that we exercise a mundane version of it whenever a present action determines the truth of a past-tensed description.

The difficulty is producing a criterion. Everyone agrees that some facts stated in the past tense are partly about later times, and everyone agrees that being told the concept exists is not enough. What is needed is a principled way of sorting facts into hard and soft that does not simply classify as soft whatever the Ockhamist needs to be soft. Attempts have a habit of overgenerating, so that facts nobody thinks are soft come out soft, or undergenerating, so that the divine beliefs come out hard again.

There is also a deeper worry that the whole enterprise is beside the point. Grant that God's belief is a soft fact. It still has a hard core: God, a concrete being, was in a certain state in the year 1000, and the being in that state looks as intrinsic and as past as anything. Ockhamists reply that the state is individuated partly by what it is a belief about, so there is no state-in-itself left over once the soft component is removed. Whether that reply works turns on questions in the philosophy of mind about whether mental states are individuated by their contents, which is a strange place for a debate about providence to end up, but that is where it has gone.

Molinism

The most ambitious response denies neither premise about the past but changes the account of what God knows and when.

Molina's part four is available in English as On Divine Foreknowledge, translated with an extensive introduction by Alfred Freddoso (Cornell University Press, 1988), and Thomas Flint's Divine Providence: The Molinist Account (Cornell University Press, 1998) is the fullest modern development.

The attraction is not just that it answers this rung's question. It gives a theist something no other view on offer supplies, which is a detailed and specific providence that never overrides anyone. God can arrange a life so that a person, left entirely free, does what is wanted, because God knows what they would do under each arrangement. Prophecy becomes straightforward. Answered prayer becomes straightforward. And, most importantly for this tier, the free will defence becomes intelligible, since Plantinga's claim that some possible person suffers from transworld depravity is precisely a claim about counterfactuals of freedom. Without middle knowledge, or something like it, the free will defence has no propositions to be about.

That exchange is the grounding objection, pressed in its influential form by Robert Adams in "Middle Knowledge and the Problem of Evil", in the American Philosophical Quarterly in 1977, and developed further by William Hasker in God, Time, and Knowledge (Cornell University Press, 1989). It remains the central dispute about the view, and the reply above is roughly Flint's. Whether it succeeds depends on how seriously one takes truthmaker demands generally, which is a question in metaphysics that this debate cannot settle by itself.

Hasker adds a second objection that does not depend on truthmakers. If the counterfactual stating what you would do in your present circumstance was true before you existed, and God used it in deciding to create, then it is now outside your power in the way past facts are. Molinism has moved the fixity rather than eliminating it. Molinists answer that you have counterfactual power over which counterfactual is true, which is the Ockhamist move appearing again in a new setting. The two solutions turn out to lean on each other more than their proponents usually advertise.

Open theism

The last response denies that there is anything for God to have known.

The position was given its name and its programme by The Openness of God: A Biblical Challenge to the Traditional Understanding of God, published in 1994 by InterVarsity Press with contributions from Clark Pinnock, Richard Rice, John Sanders, William Hasker, and David Basinger. Its philosophical case is Hasker's: if the arguments against the compatibility of foreknowledge and freedom work, and the Boethian, Ockhamist, and Molinist repairs all fail, then a theist who wants libertarian freedom has to give up the foreknowledge.

Open theists insist this is a claim about the future rather than about God. Compare the standard treatment of omnipotence, where the tradition long ago accepted that not being able to make a round square is no limitation, because there is no such thing to make. Open theists apply the same manoeuvre: not knowing what has no truth value yet is no limitation, because there is nothing there to be known.

The disagreement is therefore not really about God's cognitive powers. It is about the metaphysics of time and truth, and specifically about whether propositions about future contingents are true now. An open theist is committed to something like presentism about time and to a rejection of bivalence for future contingents, and those commitments have to be defended on their own terms rather than as conveniences for the theology.

The costs are substantial. Prophecy has to be reinterpreted as announcement of what God intends to bring about or as extrapolation from present conditions, which fits some texts poorly. Providence becomes risky: God takes chances, and some may not come off, which the tradition consistently denied. And the assurance that God's purposes will be accomplished has to be grounded in God's resourcefulness rather than in advance knowledge, which is a weaker guarantee even if it is not nothing.

There is also a sharper cost for this tier specifically. Several later rungs run arguments that assume God knew what would result from a decision. The problem of evil in its evidential form asks why God permitted a particular suffering, and the natural theistic answers appeal to goods that the permission secures. If God did not know that the suffering would occur, the shape of both the objection and the defence changes, and not obviously to the theist's advantage: a God who creates without knowing what the free creatures will do has taken a risk with other people's lives.

The same problem turned on God

There is a version of the argument that has nothing to do with creatures, and it is usually noticed late because the debate is framed around human freedom.

God knows what he himself will do. If that knowledge is infallible and already settled, then divine freedom faces the argument from the start of this rung in an especially sharp form, since there is no question of God's being unaware of the relevant belief or of its being about someone else. Suppose God knows that he will answer a particular prayer next Tuesday. Then either he can do otherwise, in which case he can falsify his own infallible knowledge, or he cannot, in which case his acts are as unavoidable as ours were supposed to be.

The classical tradition has a tidy answer, though it is tidy only because it has already paid for it elsewhere. A timeless God does not know in advance what he will do, since there is no advance; there is one eternal act, and the question of whether he could do otherwise than he foresees does not arise, because foreseeing is not what is happening. Rung 4 bought that answer with the doctrine of simplicity, and rung 5 recorded what the personalists think it costs.

Molinism has a distinctive difficulty here that is worth stating plainly. On the Molinist account God does not choose which counterfactuals of freedom are true; he finds them true and works with them. So there are worlds God cannot actualise, not because they are impossible, but because the creatures in them would not cooperate. This is exactly the structure of Plantinga's free will defence, where transworld depravity is the claim that some person goes wrong in every situation God could set up for them. What follows is a constrained deity: powerful over the circumstances, but handed a fixed set of facts about how free creatures respond that he had no part in fixing. Molinists accept this and argue that it is the price of creatures who are genuinely other than God. Critics call it a limitation on aseity, which links this rung back to rung 6, since the counterfactuals are contingent truths that God did not make and that constrain what he can do.

What each solution costs

The arguments do not converge, and it is worth being explicit about the shape of the disagreement rather than pretending someone has won.

Timelessness costs you the personal God of rung 5, and by Zagzebski's argument may not deliver even then.

Ockhamism costs you a clean account of the past, and owes a criterion for hardness that has proved difficult to supply.

Molinism costs you an explanation of what makes the counterfactuals true, and buys in exchange the strongest account of providence anyone has constructed.

Open theism costs you foreknowledge outright, along with a traditional reading of prophecy and a risk-free providence, and buys the simplest resolution of the problem.

Compatibilism about freedom costs you the ability to do otherwise, and dissolves the problem entirely, which is why a considerable number of theists in the Reformed tradition have never regarded it as a problem at all. That option is underrepresented in the analytic literature because the analytic literature mostly assumes libertarian freedom, but it is the majority position across the history of Christian theology, and leaving it out distorts the map.

Next

The responses in this rung all took for granted that we understand what is being said when we say God knows, believes, or decides. The classical tradition of rung 4 has already suggested that we do not. The next rung takes up that suspicion directly and asks whether talk about God says anything determinate at all, whether the predicates are used in the same sense as when applied to us, and what is left of theology if the answer is no.