Philosophical Arminianism

Philosophical Arminianism is expounded in Kvanvig's book, Deliberation and Destiny (Kvanvig Reference Kvanvig2011). This is really two short books, bound back-to-back to make one volume of moderate length. ‘Destiny’ is the theme of the first three chapters, which are concerned with the doctrine of hell, especially with regard to the finality of hell and the debate with different versions of universalism. ‘Deliberation’ occupies the last five chapters; the deliberation in view is the divine deliberation which precedes the decision to create. Chapter 4 is devoted to open theism, and argues that God as conceived by open theism would know virtually nothing with certainty about the future. A key premise of this argument is that, according to open theism, ‘God's nature changes in response to the indeterministic unfolding of the world he has created’ (68). Regrettably, Kvanvig does not inform us of his source for this remarkable claim about open theism.

Chapter 5 narrates a three-way discussion between an occasionalist (Occ), a Molinist (Moll), and an open theist (Ott). Moll, speaking for the author, argues that ‘actions by God that entail that an individual performs a certain action do not threaten the claim that the person displays libertarian freedom in performing that action’ (xvi). Chapter 6 renews a debate with Thomas Flint over ‘Maverick Molinism’, and argues that ‘Molinism cannot equate a counterfactual's being true in a way logically prior to God's will with the truth of the counterfactual being outside God's control’ (xvi). Chapter 7 consists of a detailed discussion of the ‘deliberative conditionals’ that God consults in deciding which sort of world to create. An important conclusion is that ‘the types of Molinist views found in the literature do not provide conditionals which are capable of function in such a deliberative model of creation’ (xvii). Each of these chapters is densely argued, and the arguments deserve thorough consideration. Our concern here, however, will be with the final chapter, in which Philosophical Arminianism is expounded.

This final chapter is entitled ‘An epistemic theory of creation’, signalling the fact that the view proposed ‘relies on epistemic conditionals in place of counterfactuals, and thus substitutes an epistemic approach to creation in place of the more standard metaphysical and semantic approach relied on by Molinists and their opponents’ (xvii). The view shares with Molinism (which Kvanvig formerly embraced) the aspiration to combine libertarian free will with a view of ‘complete providence’ in which God controls every detail of the world's history – a combination Thomas Flint has termed ‘libertarian traditionalism’. Kvanvig terms his view ‘Philosophical Arminianism’ because he wished to associate it with a historical figure who would be readily recognizable as having held libertarian views on free will.Footnote ¹

Philosophical Arminianism, like Molinism and some deterministic views, offers a deliberative model of creation. Kvanvig asserts (plausibly, I think) that: ‘there is no alternative for understanding creation except in terms of a deliberative model of the process God uses in determining what to do’ (140). This is so, even though we cannot suppose that God literally deliberates. The deliberative model begins by supposing a particular divine creative act or set of acts, and then uses appropriate conditionals to determine the overall state of things – the possible world or, as Kvanvig sometimes terms it, the complete fiction,Footnote ² or complete story – that will result from such a creative act. Once this has been done for a suitable ensemble of initial divine creative acts, the resulting worlds can be assessed and the best (or one of the best) selected for actualization. Since what God decides to actualize is a complete possible world, full providential control is guaranteed. What is distinctive about Philosophical Arminianism is the nature of the conditionals employed by God in this deliberative process. These are open, indicative, epistemic conditionals. Open, because the antecedents are consistent with what is known (or, in hypothetical reasoning, supposed) to be true. The conditionals are indicative, as opposed to the subjunctive conditionals featured in Molinism. Furthermore, these are epistemic conditionals; this is what makes the view an epistemic account of creation. This means that the antecedent of the conditional must provide adequate evidence for the truth of the consequent. But what does it take for evidence to be adequate? Kvanvig gives a double answer: the evidence, if undefeated, is sufficient for knowledge of the consequent, given that the consequent is in fact true. And the evidence is strong enough to ‘legitimate closure of inquiry’ (148) concerning the truth of the consequent.

Now, no one will doubt that epistemic conditionals of this strength will be available when they are backed by deterministic laws of nature. Some will question, however, whether such conditionals will be available when the consequents describe libertarian free acts of human beings. If there are no sufficiently strong epistemic conditionals in these cases, the process of hypothetical reasoning cannot reach its conclusion in the depiction of the resulting ‘complete story’, and the deliberative process will be blocked. Kvanvig takes this objection seriously, and does not claim to have a conclusive refutation for it. However, he makes some points that, he claims, ‘minimize this worry sufficiently that the proper conclusion to draw from them is that it is at least an open epistemic possibility that the group [of complete stories resulting from suppositional reasoning about worlds containing free agents] is not empty’ (156). One of these points is made by pointing out that global scepticism concerning the future is highly implausible. More importantly, however, global scepticism about free action is also implausible. Kvanvig states: ‘Knowledge doesn't require infallible grounds, and the quality of our epistemic support regarding the behavior of other humans is regularly as good as our epistemic support regarding the behavior of, say, our favorite espresso machine’ (156f.).Footnote ³ There will nevertheless be some sets of conditions for which no well-supported epistemic conditionals are available; in particular, Kvanvig mentions ‘Buridan's ass’ situations, in which ‘there are equally attractive options for some individual, and there is no basis on which to conclude that the individual will take one of the options over the others’ (158). Even where the conditionals are available, they are acknowledged to be both fallible and defeasible; obviously this will need to be remedied at some point if they are to serve in the divine deliberation concerning creation.

A bit more needs to be said, however, about the basis for these epistemic conditionals, and also for God's knowledge of them. Essentially, the same answer suffices for both questions: the epistemic conditionals are determined by the facts that obtain, or that are presumed to obtain, in conjunction with the necessarily true principles of epistemic support. It is in virtue of the way the world is that some propositions are evidence for others. Given the facts plus the epistemic principles, the epistemic conditionals are set. And whereas we humans have to learn from experience how evidence works, and what confirms what, God has no need of such learning; God is not only logically omniscient but also epistemically omniscient.Footnote ⁴

One final point needs to be made: it is stipulated that there are no inscrutable truths – truths that leave no possibility of detection whatsoever (144). This stipulation would be seriously controversial if applied to human beings; it is certainly plausible that there might be truths (for example, concerning the divine nature) which would forever escape the most expert and insightful human scrutiny. But the model is a model for divine knowledge, and it seems safe to assume that no truths are inscrutable for God. As we shall see, this stipulation becomes important at a later stage in our discussion.

With this much in place, we can give a brief outline of the process of divine suppositional reasoning which takes deliberation ‘from the original information state W [consisting of logical and epistemic omniscience] to a full fiction or story. . . . We begin by considering the range of things that God suppositionally considers as candidates for strongly actualizing [i.e., for causally necessitating] . . . whether the object of necessitation is some particular matter of fact or some law of nature’ (155). We can distinguish between total and partial suppositional units; a total suppositional unit ‘is one that contains everything God is considering strongly actualizing in a given piece of suppositional reasoning’ (156). Each total suppositional unit, or ‘plan’, consists of a set of suppositions, ordered in terms of the temporal sequencing of the parts of the plan. These parts are then considered in order, and with each new supposition there is or may be a change in the epistemic conditionals that are supported; as was noted previously, it is in virtue of the way the world is that some propositions are evidence for others. Conditionals may be lost as well as gained in this process; additional information may defeat what was formerly a well-supported epistemic conditional. This is because ‘strengthening’ is not valid for epistemic conditionals. If a material conditional ‘p ⊃ q’ is true, so is ‘(p & r) ⊃ q’, and the same is true for strict conditionals, but the corresponding inference for epistemic conditionals is invalid.

Among the complete suppositional units, or plans, we can distinguish complete plans from incomplete plans:

(Thus, a complete suppositional unit may fail to result in a complete plan, if the ‘final state of information’ generated from the supposition leaves the truth-values of some propositions open.) These complete stories, then, represent options that are available to God as he decides which world to create.

We are not yet home free. The epistemic conditionals with which we began were acknowledged to be defeasible; it follows that knowledge or beliefs derived from them must be fallible. But God can hardly be thought to base his creative decisions on a fallible foundation. Something, then, needs to be done. Kvanvig's response, perhaps surprisingly, is that whereas the epistemic conditionals with which we began were defeasible, the ones that are left at the end of the process of suppositional reasoning are indefeasible. To see this, consider what it would be like for God, at the end of this process, to hold a defeasible belief. God, for instance, believes the conditional, ‘C → A’, where ‘C’ represents a total state of circumstances in which an agent performs a free action, and ‘A’ represents the action that is taken. And knowing that C obtains or will obtain, God also believes that A is true. Now, if the conditional C → A is defeasible, the following is possible: God holds the beliefs as stated above, but the agent fails to perform action A. At this point our no-inscrutable-truth stipulation comes into play. If A is false, its falsity will leave traces, and these traces will constitute a defeater both for A and for C → A. Now, this defeater must be either internal or external to God's cognitive system. If it is internal, God will be aware of this, and will be aware that the beliefs in question are defeated; thus, he will not in fact believe either A, or C → A. So, given that God does hold those beliefs, there cannot be an internal defeater. But neither can there be an external defeater. Kvanvig explains, ‘The reason there can't be any external defeaters is that, by the end of the reasoning process, the story is complete’ (165). It is the completeness of the story (the possible world) that makes the difference: at the beginning, there was always the possibility that more information would be added, and this added information might prove to be a defeater for a previously justified belief. But at the end of the reasoning process, this possibility no longer exists. Since there cannot be either an internal or an external defeater, the beliefs in question are indefeasible, contrary to our original supposition. Kvanvig goes on to argue that this shows that God is infallible, at least with respect to a certain set of propositions: namely, all those propositions included in the complete stories which result from various divine creative plans.

On this account, then, God is infallible in the indicated respect, but he is still not essentially omniscient. For there are those possible worlds in which God's creative actions do not result in a complete plan, and in those worlds God will not know with certainty everything that will occur. Kvanvig does not find this to be obviously unacceptable; he states: ‘When the plan is a complete one, the world that results is a divine accomplishment; when the plan is an incomplete one, the world that results is a divine adventure. For God to display full providential control, the actual world must be an accomplishment rather than an adventure’ (171f.). But if we do consider essential omniscience to be an essential attribute of God, we can embrace the idea that ‘it is incompatible with the divine nature to take risks with creation’ (171).Footnote ⁵

Kvanvig sums up his model in the slogan, ‘God knows the beginning from the end’ (165). I take this to mean that at the end of the process of suppositional reasoning God has infallible knowledge of each of the complete stories – the possible worlds – that will result from whichever creative plan he selects. In this way, God is in the position to select the plan that will be actualized, and to begin the enactment of the chosen story.

Now we are in a position to compare Philosophical Arminianism with other views of divine foreknowledge and providence. Unlike deterministic models, it does have a place for libertarian free will for creatures. Unlike open theism, as well as other models with full divine foreknowledge or timeless knowledge but without middle knowledge, it allows for full providential control, because God knows, prior to his decision concerning what to create, exactly what will result from such a decision. The closest comparison, then, is between this view and Molinism, both of which affirm ‘libertarian traditionalism’. The view does have certain possible disadvantages as compared with Molinism. As was noted above, the view does not immediately guarantee divine essential omniscience; in order to secure that, an additional assumption is needed, namely that God is essentially incapable of taking risks with his creation. (Admittedly, most Molinists would accept this assumption, but for them it does not need to be added as an extra premise, as is the case for Philosophical Arminianism.) Another possible disadvantage is that the view does not guarantee, as Molinism does, that full providential control is possible in a world containing libertarian freedom. It is claimed that the view renders this claim epistemically possible, and perhaps even plausible, and Kvanvig contends that this should be sufficient. Nevertheless, the lack of a guarantee on this point must be accounted as a cost of the model.

Offsetting these costs, however, there are substantial benefits. It has been seriously questioned whether true counterfactuals of freedom even exist, but there is no real doubt about the existence of true epistemic conditionals of the sort favoured by Philosophical Arminianism. Nor is there room for doubt about the grounding of such conditionals, in contrast with the persistent ‘grounding objection’ to Molinism. It is also fairly clear how such epistemic conditionals can be known – very imperfectly known for us humans, to be sure, but there is no problem at all concerning God's knowledge of them. And they are not subject to the objection, raised against Molinism in the preceding chapter of the book, that the conditionals are not suitable to be used in God's deliberations concerning which world to actualize. But are there further objections to Philosophical Arminianism, which have not been addressed so far? That is the topic of our next section.

Some objections

I will now argue that there are at least two major difficulties remaining for Philosophical Arminianism, difficulties that need to be addressed if the view is to be viable. The first of these concerns the availability of the epistemic conditionals that are basic for the model. I will not claim that there are no true conditionals of this sort in situations involving libertarian free will. I contend, however, that there will not be enough of them to serve the purposes of the model. In these conditionals, we recall, the epistemic support provided by the antecedent is sufficient for knowledge, given the truth of the consequent, and is also sufficient to legitimate closure of inquiry. The conditional is not as such an affirmation about the likelihood of the consequent given the truth of the antecedent, but it surely has implications for this. It is out of the question, for example, that a conditional would legitimate closure of inquiry if, assuming the antecedent, the consequent were only as likely as its denial, or even if it were slightly more likely than its denial. As a very rough estimate, we might say that the conditional probability of the consequent given the antecedent must at least be in the range of 80 to 85 per cent.

Now I will grant, for the purposes of this discussion, that we sometimes do have evidence of this strength concerning the future actions of free agents. And given this, it is epistemically possible, and perhaps even probable, that there are possible worlds involving free agents which yield a ‘complete story’ as described by the model. But here is the question I would like us to ponder: is it at all plausible that our actual world, the ‘story’ in which all of us are characters, is of this sort? Keep in mind that if there is, in the entire history of the cosmos, even a single episode in which there is no conditional of the requisite strength predicting the action of a non-determined agent, that history will not yield a complete story, and full divine providential control will not be possible. Now, Kvanvig has granted that there are possible situations in which no such epistemic conditional will be available. The example he gives is of a ‘Buridan's ass’ situation, in which ‘there are equally attractive options for some individual, and there is no basis on which to conclude that the individual will take one of the options over the others’. Ask yourself this question: Is it really at all plausible that there has never, in the entire history of our world, been an actual Buridan's ass situation? A determinist, to be sure, would answer that question in the affirmative, citing the possibility (on her view, the certainty) of undetected causal factors which lead to the individual's making one choice rather than the other. But once we have conceded the reality of undetermined free choices, it seems difficult indeed to justify an affirmative answer to the question.

Buridan's ass situations are not the only ones to be considered here, however. To take another class of examples, consider the huge number of athletic contests that are engaged in every day of the year. A good many of these contests are decided by a narrow margin, often in the very last minutes of the game. The results in such cases depend on the presumably free actions of one or several agents – for instance, whether a soccer ball is struck in a very slightly different way, so that it glances off a goal-post and goes into the net rather than rebounding into the field of play. (This brings out another element of difficulty for the model. We may be able to predict the ‘gross behaviour’, so to speak: ‘the player will kick the ball towards the goal’. We are, however, almost never in a position to predict the finer details of that behaviour, such as the one just mentioned. Are we asked to assume that, once the gross behaviour has been decided on, the details are deterministically controlled? And why should we assume that?) Is it in any way plausible to suppose that, in each and every one of this multitude of cases, there are available strong epistemic conditionals that meet the requirements of the model?Footnote ⁶

Another group of difficulties arises if we suppose, as our best contemporary science seems to indicate, that the fundamental laws of nature are in part probabilistic rather than deterministic. I submit that it is out of the question to suppose that there are epistemic conditionals of the sort described by Kvanvig for each and every quantum-level indeterministic event. The decay of a radioactive atom, for instance, is in the typical case purely spontaneous; there is nothing at all in the circumstances that enables anyone to predict that the atom will decay at one time rather than another – and this in spite of the fact that for large ensembles of such atoms, statistical predictions can be extremely reliable. Surprisingly, Kvanvig does not discuss this problem at all in his chapter on Philosophical Arminianism. He does refer to it in the preceding chapter, in the course of developing a difficulty for Molinism (133–139). But I do not see that he provides any solution for the problem that would save the day either for Molinism or for his own view. There are, to be sure, a couple of possible solutions, neither of them very appealing. One could insist that the true, ultimate laws of nature must after all be deterministic. (This is not absolutely ruled out by current scientific knowledge, though it is held by only a small minority of scientists.) Or, one could grant that so far as natural causes are concerned the indeterminism is ultimate, but hold that these naturally undetermined events are all directly controlled by God. Kvanvig, however, considers both of these options and rejects them (133–134). (The latter option he terms a ‘god-of-the-gaps’ solution.) But if these possibilities are dismissed, what options remain? Kvanvig offers little hope here.

But suppose we assume (or pretend) that there is an ample supply of epistemic conditionals to fill out a complete plan for a world such as ours. At that point another major difficulty rears its head, with regard to the alchemy by which the base metal of defeasible, fallible belief is transmuted into the gold of infallible knowledge. As we have seen, Kvanvig gives an account of how this is to happen, but I believe there is an important difficulty concerning how the process can work.

The process of suppositional reasoning generates at each stage a stock of epistemic conditionals, conditionals stating what will happen under given conditions. Our main concern, of course, is with those conditionals that involve the undetermined actions of free agents. The suppositional reasoning will conclude that each such agent will do the thing he or she is most likely to do, as spelled out by the epistemic conditionals that apply to such an agent. Now we know very well that free agents don't always do the thing that, antecedently, they were most likely to do. And this means that, if God were to perform the creative actions that inaugurate a particular story, the story thus told would almost certainly fail to come out the way the reasoning would have led us to expect. (That's why the reasoning at this point is defeasible.) So these misleading conditionals have to be identified and corrected, and when this has been accomplished the resulting story is indefeasible and God has infallible knowledge of what would emerge from that particular ensemble of creative actions. The misleading conditionals are identified, says Kvanvig, because ‘falsity leaves traces’, and these traces, showing up in the story as it develops, alert an astute reasoner (namely, God) that something has gone wrong and corrections need to be made at an earlier point in the story. (In correspondence Kvanvig likened this to a situation where, in writing a novel, the novelist finds that developments later in the book make nonsense of what has gone before. Clearly, revisions to the earlier episodes are required, in order to make the overall story coherent.)

The difficulty with this becomes clear when we ask the question: what is it about the situation that triggers the occurrence of the ‘trace’ which alerts God that something has gone wrong with the story? ‘Falsity leaves traces’, but what is the basis for saying that the prediction generated by the conditional in question is false? To be sure, if we were thinking about an actual, concrete world, God could, so to speak, ‘look and see’ that the predicted action has not taken place. But that is not possible here; the progression from defeasibility to indefeasibility must occur entirely within the process of suppositional reasoning. At this stage, there is no actual universe; God's creation of the actual universe comes later, after the process of suppositional reasoning is complete and God knows the options that confront him.

Consider, then, the process of suppositional reasoning. We have an agent, in circumstances we designate as ‘C’. We have an epistemic conditional, C → A, which is derived in the light of all the facts supposed to obtain in this particular story, guided by the necessarily true epistemic principles. Given this, what basis could we possibly have for saying that the epistemic conditional is wrong, and that the action designated by ‘A’ will not occur? The basis cannot be found in the actual choice made by an agent, for at this stage there is no actual agent. The basis cannot be found (as it perhaps is in the novel) in a conflict between the agent's character, prior experiences, and so on and the predicted action, for all that will have been fully taken into account in the reasoning which derived the epistemic conditional C → A. But what other options are there?

There is, however, one possible way (and it may be the only way) in which we could say that the prediction is false. It seems that in order for the conditional to be false, there must be a true conditional stating what the agent in question would do if placed in such a situation – a conditional, that is, such as C □→ ~A. If there were such a conditional, then the conflict between that conditional and the one generated by the process of suppositional reasoning might perhaps trigger the ‘traces’ which would lead to the needed revision. (The ‘no-inscrutable-truth’ stipulation would guarantee that the presence of such conditionals could not indefinitely escape the scrutiny of a sufficiently astute reasoner.) But the acceptance of conditionals such as C □→ ~A is a momentous step, one which a Philosophical Arminian might well prefer to avoid. These conditionals are indistinguishable from the counterfactuals of creaturely freedom which are a foundational assumption of Molinism. To agree with Molinism in embracing them would leave Philosophical Arminianism vulnerable to many of the well-known objections to Molinism,Footnote ⁷ and would cancel out much of the advantage that Philosophical Arminianism claims in comparison with Molinism. But how this is to be avoided is not at all clear.

Summary and prospects

Without doubt, Kvanvig has succeeded in introducing a fundamentally new position into the debate about divine foreknowledge and providence. Recently it has become fairly widely accepted that the only view on these topics that can combine libertarian free will with a doctrine of complete providence is Molinism.Footnote ⁸ Philosophical Arminianism competes with Molinism for this sector of the battlefield, and in doing so it alters the overall strategic situation. It might very well prove to be a mistake to assume that the difficulties for the view pointed out in this article will mean that it is unable to maintain itself as an option. Opponents of Molinism have been producing counter-arguments to that position for a considerable time now,Footnote ⁹ and the Molinists have proved themselves to be impressively resourceful and tenacious in holding their ground. There is no reason why Philosophical Arminians should not turn out to be equally capable in defending their chosen option.

On the other hand, predictions that Philosophical Arminianism will advance triumphantly over the battlefield, over-running previously well-defended positions, are likely to prove unwarranted. The view will have little attraction for open theists, who find it distinctly preferable to view the world's history as a divine adventure rather than as (in all respects) a divine accomplishment. Nor will the view hold much appeal for theological determinists, who have no use for libertarian free will in the first place. I would venture, furthermore, that Molinists who previously have found no major difficulties with their own position are unlikely to find Philosophical Arminianism attractive. If they are unmoved by the standard objections to their view, they are unlikely to find Kvanvig's new argument (set out in chapter 7) more impressive. Furthermore, they will be reluctant to trade the certainty of full divine providential control afforded by Molinism, for the mere epistemic possibility offered by Philosophical Arminianism. I propose, then, that this view is most likely to appeal to two groups of philosophers: First, to those who hold to divine foreknowledge (or to divine timeless knowledge) and are unable to subscribe to Molinism, but who harbour a desire for a stronger doctrine of providence than their views otherwise seem able to support. And second, to erstwhile Molinists who, like Kvanvig himself, have become uncomfortable with their own position, yet who desire to retain as many of its benefits as possible. These comments of mine, however, are very much in the category of plausible conjectures; they fall far short of the strength of support required for epistemic conditionals as described by Kvanvig.

Among the possible worlds it is interesting to contemplate is one in which Jacob Arminius, in the final months prior to his death in 1609, was vouchsafed a preview of the theory that would later be promulgated in his name. He couldn't have read Kvanvig's book, of course, not only because it had not yet been written but because it is full of references to concepts, theories, and philosophers unknown to him, which would have made it extremely difficult for him to understand the view. No doubt, however, a suitable paraphrase into the then-available terminology could be prepared. We would picture him as being intensely interested, poring at length over the manuscript as he lay on his sickbed. Eventually, however, we see him setting the new work on his bedside table with a sigh. He then picks up his well-thumbed copy of Molina's Concordia, and begins to read.Footnote ¹⁰