In Defence of Open Theism

Richard Swinburne

University of Oxford

Editor | Gregg D. Caruso, Corning Community College, SUNY (USA).

Correspondence | Richard Swinburne, 50, Butler Close, Oxford, OX2 6JG, UK; Email: richard.swinburne@oriel.ox.ac.uk

Citation | Swinburne, R. (2017). In defence of open theism. Science, Religion and Culture, 4(2): 56-62.

DOI | https://dx.doi.org/10.17582/journal.src/2017/4.2.56.62

In 1965 Nelson Pike presented his very clear version of an ancient argument purporting to show that a temporal God (that is, one who exists at all moments of time), could not have infallible knowledge of any future causally undetermined contingent event, and so – if there are any such events - could not be essentially omniscient, in the sense of having infallible knowledge of all true propositions. Given the assumption that human free actions are causally underdetermined, this entails that such a God cannot at any time have infallible knowledge of which future actions humans will do freely. Almost all discussions of this issue by analytic philosophers have been a series of responses to Pike; and this collection of John Martin Fischer’s previously published essays is a collection of such responses. This is no criticism of the collection, because Pike’s argument is a very powerful argument with a very important conclusion about the nature of God and his interactions with humans. Also, the various ways of attempting to avoid Pike’s conclusion have considerable intrinsic philosophical interest. All the previously published essays in the collection conclude that Pike’s argument remains undefeated. I endorse this conclusion. A temporal God cannot have total foreknowledge of future free human actions.

Fischer’s New Argument

However in a new 50-page introduction to the collection. Fischer himself presents – somewhat hesitantly – a new way of avoiding Pike’s conclusion. On almost all theories of knowledge, S knows that p iff S believes that p, p is true, and certain other conditions are satisfied. When S believes that p and those other conditions are satisfied, S is said to be (on Fischer’s definition) ‘in a knowledge conferring situation’ (KCS) with respect to p. Humans can often be in a KCS with respect to logically contingent propositions about the future, for example about ‘the Sun will rise tomorrow’; and they could be in this situation even in an indeterministic universe, for example when they have evidence which makes it 99.99% probable that the relevant proposition is true. In that case, Fischer plausibly suggests, God could often be in that situation. But then, Fischer (2016,38) suggests, God could have an additional piece of evidence known to him with certainty, that he is essentially omniscient and so that his knowledge, unlike human knowledge, is not fallible. ‘God can thus “bootstrap” his way to certainty in this distinctive way’! Fischer emphasizes that God can (or, at any rate, will) only ‘bootstrap’ to infallible knowledge those propositions with respect to which he is already in a KCS in virtue of satisfying conditions of the same kind as put humans in a KCS. So he would not have total omniscience about the future in an indeterministic universe, but he could know infallibly much more about human free actions than all the other papers in this collection suggest to be possible. This view however does seem open to an apparently conclusive objection. God cannot know something which is logically impossible. Given Pike’s argument and the failure of other objections to it, it is logically impossible that any being at all could know infallibly the truth-value of all logically contingent propositions in an indeterministic universe. But of course, as Fischer (this time on collaboration with Neal Tognazzini) points out (2016, 228-9) elsewhere in the collection, if it really does seem to someone more obvious that both there is a temporal God who is essentially omniscient and that humans have indeterministic free will, than that any of the arguments against this are cogent, then that person is fully rational in maintaining the former conjunction. I am confident that hardly anyone is in that position unless they consider that that conjunction is an item of revealed knowledge, which - for a Christian - means revealed in the Bible and/or the teaching of the Church. And I feel fairly confident that no biblical text or church teaching would support Fischer’s latest view that God infallibly knows a lot more than any human who knows all that God knows about the past and his decrees for the future (and is able to discover the logical consequences of these) could ever know fallibly, but that there are still some propositions about the future which God cannot know at all.

Three Senses of ‘Hard Fact’

While I have no objection to any of the lines of argument of any of the previously published essays in the volume, they do seem to me to need filling out in three important ways – in respect of how a ‘hard fact’ can be defined, in respect of whether the past is ‘fixed’, and in respect of whether a Pike-style argument has equal force against the view that a timeless God is essentially omniscient. I shall make my points about these issues mainly by repeating claims of mine which were published only last year (in Swinburne 2016) and so ones which few of the readers of this volume are likely to have read.

All the lines of arguments in Fischer’s book depend on the unargued principle of the ‘fixity of the past’, that the past is ‘fixed’ in respect of ‘hard’ facts about the past. The expression ‘hard fact’ is a philosopher’s term, and we can define it how we like, but there seem to be three different ways of understanding it in different places in the book, only two of which are explicitly distinguished. There are two purely temporal senses of ‘hard fact’, which Fischer does not distinguish from each other. He writes that ‘hard facts … about a time T are genuinely about a time T and not even implicitly about times after T. Soft facts about a time T may be genuinely about T but are also (in some genuine sense) about times after T’ (Fischer 2016, 12). It is natural to spell this out more precisely as ‘hard facts do not entail but soft facts do entail facts about the future’; and I will call that the first definition of ‘hard fact’. But, as I shall illustrate, Fischer seems sometimes to have in mind a slightly different temporal sense of ‘hard fact’ an alternative sense, that a ‘hard fact F at T is one which is genuinely about a time T, even if it is also implicitly about times after T. That is to say, F is a hard fact at T iff F has sufficient and necessary conditions for its occurrence at T, whether or not those conditions and so F itself have necessary and/or sufficient conditions at times after T. I shall call this sense of ‘hard fact’ the second sense. Then a soft fact about T in the second sense is a fact about T which is not a hard fact about T in this sense. So while all hard facts in the first sense are also hard facts in the second sense, the converse does not hold. And then there is the other sense of ‘hard fact’ which Fischer explicitly discusses, the ‘independence’ sense, that a hard fact about a time T is one which does not ‘depend on’ or is not ‘determined by’ any facts after T, whereas a soft fact is one so dependent or determined. Todd in particular is cited (Fischer, 2016,25-6) as supporting the latter sense as the relevant one in this context. To avoid circularity we need to understand by ‘facts after T’ ‘hard facts after T’ in some other sense of ‘hard fact’, and I suggest that we understand by that ‘hard facts after T’ in the second sense. I shall call this independence sense of ‘hard fact’ the third sense. A fact about T which is not a hard fact about T in the third sense is a soft fact about T in that sense. In the first sense of ‘hard fact’, since God is supposed to be essentially omnipotent, as well as essentially omniscient, both his decrees at T for the future and his beliefs at T about the future count as soft facts about T. Whereas in the third sense of ‘hard fact’ God’s decrees about the future are clearly hard facts, but – if God has infallible foreknowledge- God’s beliefs about the future , or at least those about future free human actions, are soft facts in the third sense. God believes that Jones will do X at T2 because Jones will do X at T2; which belief God holds is determined by what will happen.

The first sense needs careful spelling out to clarify what makes a fact ‘genuinely about a time T’, since all facts seem to entail facts about both future and past. Does not ‘Jones does X at T1’ entail ‘it was true at T0 that Jones would do X at T1’, and ‘it will be true at T2 that Jones did X at T1’, where T0 is earlier than T1,’ and T1is earlier than T2? The obvious way of avoiding this problem is to define a fact as a hard fact about T1 iff it would be a fact even if there were no times earlier than or later than T1. Hasker (1986,81-90) proposed that we call a fact ‘hard’ iff it does not entail the existence or non-existence of later times; but that introduces an unnecessary past/future asymmetry into ‘genuinely about a time T1’, which can make arguments for the fixity of the past question-begging. But even my proposed definition would not settle the issue of whether some fact at T1 is a hard fact at T1, because whether some fact would be a fact if there were no times later or earlier than T1 may depend on what else happens at T1.Thus whether the fact that John met his future wife Mary in 1988 is a hard fact or a soft fact about 1988 depends on when John married Mary. If he married her later in 1988, it is a hard fact about 1988; but if he married her only in 1990, it is a soft fact about 1988. So I now present the definition which I offered in Swinburne (2016,161-3). It is phrased in terms of ‘events’ rather than ‘facts’; but , given that facts are facts about the occurrence of events, nothing turns on that. It also assumes that events take place over periods of time and not at instants of time, and that instants are just the boundaries of periods; and so it assumes that talk about instants can always be analysed in terms of talk about periods (for example, that to say that the match finished at exactly 4pm, is to say that the match took place for a period ending at 4pm and did not take place during any period beginning at 4pm.) So the ‘T’’s in the definition are names of periods of time. I believe however that the subsequent argument which depends on this definition can be rephrased so as to allow that events may occur at instants. Here is the definition and illustrations of how it works:

“I define an event E as a hard event at T iff the occurrence or non-occurrence of times before or after T is neither logically necessary nor logically sufficient for the occurrence of E—given the occurrence of the other events that occurred at T, for the occurrence of which the occurrence or non-occurrence of times before or after T is neither logically necessary nor logically sufficient. (To avoid too clumsy a definition, I am counting the non-occurrence of an event of some kind at T as an event at T.) An event at T that is not a hard event at T is a soft event at T.

I now illustrate how this definition works. Me-having-a pain-at-midday is a hard event at midday, because whether it occurred is logically independent of the occurrence of any time before or after midday, given everything else that happened at midday for the occurrence of which the occurrence or non-occurrence of times before or after T is neither logically necessary nor logically sufficient. Whether or not there were such times, I could still have had or not have had a pain at midday. Likewise a-world-war-beginning-in-1914 is a hard event at 1914. Its occurrence is independent of there being any time before or after 19142, given everything else that happened in 1914, for the occurrence or non-occurrence of which the occurrence or non-occurrence of an earlier or later year is neither logically necessary nor sufficient. But John-being-30-years-old-in-1988 is a soft event in 1988, since it is logically necessary for its occurrence that there were thirty years before 1988. Necessarily, if there were no times before 1988, that event could not have occurred. Intuitively John-meeting-his-future-wife-in-1988 is a soft event in 1988. But neither the occurrence nor the non-occurrence of times before or after 1988 is logically necessary or sufficient for its occurrence, for John could have married Mary later in 1988. To get the result that this is a soft event in 1988 we need the clause ‘given the occurrence of the other events that occurred at T, for the occurrence of which the occurrence or non-occurrence of times before or after T is neither logically necessary nor logically sufficient’. For, given what else happened in 1988, including the event of John-not-marrying-Mary-in-1988, the occurrence of times later than 1988 is necessary for the occurrence of John-meeting-his-future-wife-in-1988; and the occurrence of other times apart from 1988 is not logically necessary or sufficient for the event of John-not-marrying-Mary-in-1988. The-beginning-of-the-First-World-War in 1914 is a soft event in 1914, because the non-occurrence of any years before 1914 is logically sufficient for its occurrence, given the occurrence of a-world-war-beginning-in-1914 (for the occurrence or non-occurrence of which, as noted above, neither the occurrence nor the non-occurrence of times before or after 1914 is logically necessary or sufficient.)

An event that is not a hard event at a time T may be a hard event at a different time, and in particular at a time that includes T. Thus, while John-meeting-his-future-wife-in-1988 is a soft event in 1988, it is a hard event at the period 1988–90 inclusive; the existence or non-existence of years outside that period makes no difference to whether or not it occurred.” (Swinburne, 2016, 161-3)

It then follows - given that God has total foreknowledge of human free actions– that God’s beliefs at T1 , about some future event at T2 are ‘soft facts’ at T1 in both the first and the third senses. They are soft facts in the first sense, since necessarily God’s beliefs are true, and so the occurrence of the later time T2 is necessary for their truth . In the third sense of ‘hard fact’ - given again that God has total foreknowledge of human free actions - God’s beliefs about human future free actions are soft facts because which beliefs he has depends on those actions; they will vary in response to which actions humans freely do.

However in the normal sense of ‘belief’ a belief at T1 is what it is solely in virtue of how things are at T1. And if God is to have ‘beliefs’ in a normal sense, those beliefs must be hard facts in the second sense which Fischer does not explicitly distinguish from the first sense. Fischer claims that the soft facts of the beliefs of a necessarily omniscient God about human future free actions in the first sense, soft in the first sense that they entail the existence of future times, seem to contain a hard element; they are what Fischer calls ‘either hard-core or hard-type soft facts’ (Fischer, 2016,146-7). On the assumption that God is necessarily omniscient (as surely is the normal theological assumption) and so ‘God’ is a name and not a role-term, it would be a hard-type soft fact. ‘There is a state which intuitively obtains at T1 (a state of God’s mind) which in fact counts as believing that S does X at T2, and which would so count, no matter what happens in the future’. This way of understanding the ‘hard’ aspect or element in God’s beliefs about future human free actions involves those beliefs being ‘hard’ in my second sense. Is this possible? Could the soft facts (‘soft’ in both the first and third senses) of God’s beliefs at T1 about T2 contain this hard element, as Fischer suggests that they must do if God is to have essential foreknowledge? If a ‘soft’ belief at Tr in the first sense is defined in the way that I defined it, as entailing the existence of times later or earlier than T1, and given that there is no problem with God essentially having a lot of knowledge, there is no problem with God having beliefs which are ‘soft’ in the first and third senses but hard in the second sense. For God’s or any one else’s essential knowledge of the past would exhibit just this feature. If God has an infallible belief today about what Jones did freely yesterday, this is a soft belief in the first sense since it entails the existence of yesterday, a soft belief in the third sense since it depends on what Jones did yesterday, and a hard belief in the second sense since it has necessary and sufficient conditions for its occurrence in the state of God’s mind today. The problem lies with the fact that God’s beliefs about future free human actions can only be infallible if they are soft in the third sense, that they depend for their truth on those future free actions. Given our assumption that free actions are not totally caused, God’s beliefs about them can only be infallible ones if the actions cause God’s beliefs, and that involves backward causation – an effect preceding its cause – being logically possible; and so God can only have such infallible beliefs if the past is not ‘fixed’. Fischer (2016,147) rightly sees the obstacle to God having infallible knowledge of human free actions to be constituted by ‘the fixity of the past’. If backward causation were logically possible, then there would be no problem with God’s beliefs at T1 about free human actions at T2 being ‘soft’ facts at T1 in the first and third senses, and a ‘hard’ fact in the second sense.

The Past is Fixed

So is the past fixed? What is it for a ‘fact’ or ‘event’ to be ‘fixed’ at a time T? There is an odd passage in Fischer’s book (2016, 190) in which he claims that some soft facts are fixed: ‘it is a soft fact about early this morning that the sun rose twenty four hours prior to another sunrise, but presumably noone has a choice about this fact since no one can prevent the sun’s rising tomorrow.’ But any powerful enough demigod could easily prevent the sun from rising tomorrow. Surely in this context, what we should understand by a fact (or event) at T being ‘fixed’ at T* is that no agent however strong could at T* – it is logically possible – cause it to be the case that that fact (or event) occurred or did not occur; and so the past is fixed iff no agent, however strong, could cause a past event to occur or not occur. So is the past fixed in this sense? Fischer assumes without question that it is. While it may seem obvious to almost all philosophers (but not to all physicists) that the past is always fixed, I believe that it can be shown to follow from an even more obvious principle – that the future is never fixed, always fixable. Here is my argument for the fixity of the past (based on Swinburne, 2016, ch. 9). The argument is phrased in terms of an assumption (which is, I believe, dispensable) that events are caused by substances (e.g. physical objects or persons) rather than by other events, as well as in terms of the assumption that events take place over periods of time and not at instants of time.

Given these assumptions simultaneous causation is to be understood as a substance exercising its causal influence over exactly the same period of time as the effect that it causes, and backward causation is to be understood as a substance beginning to exercise its causal influence after the beginning of the effect that it causes. The events referred to are hard events (in the second sense). One may cause an event E by causing another event that in turn causes E. A cause that is the most immediate (or direct) cause of an event, and actually brings it (the whole event) into existence, may be said to ‘fix’ it. By an event being ‘fixable’ at a time T I shall mean that the event is such that it is logically possible that a strong enough agent could at T have fixed either it or its non-occurrence, whichever he chose, whatever else logically possible might be the case at any other time.

One could bring an event E into existence at a time only by beginning to act at a time when E does not exist. One can of course cause an event to continue in existence, but to cause it to continue to exist is to cause a later part of the event to exist, and one can only do that by acting at a time when that later part does not exist. So E cannot be fixed by a direct cause acting at exactly the same time as E. For this simple reason, I suggest that simultaneous direct causation is impossible. And not merely must the direct cause C of some effect E exert causal influence at some time other than when E occurs, but it must continue to do so over the whole time while E occurs, if C is to fix E. For, if C exerted causal influence so as to cause E for only a part of the time when E occurred, the substance involved in E itself or some other substance could fix the other part of E. So, I claim, a substance that fixes an event must exert its influence for a period of time that includes both a time when the event does not occur and the whole time when it does occur. We can see this principle at work in mundane examples. When a moving billiard ball A hits a stationary billiard ball B and causes B to move, A must be moving as it touches B and so for some period ending with the instant at which it touches B, in order to transmit its motion; and it will be the direct cause of B’s subsequent motion only for the period while it is still in contact with B. After contact is broken, the direct cause of B’s subsequent motion is B itself. Among other paradigm examples of a substance being the direct cause of an event are a person having an intention to cause some bodily movement being followed immediately by a brain event that causes the movement (‘followed immediately’ in the sense that the effect begins after the beginning of the cause and ends at the same instant as it).

I now suggest that our understanding of the future involves the understanding that we could always make a difference to it if we were strong enough. To deny that is to embrace fatalism. So I offer my principle of the fixability of the future: that a powerful enough agent could directly cause and so fix any hard event at all (hard in the second sense) at a time T by a causal act beginning before T and continuing throughout T, whatever else might be the case at any other time. So every event at T is fixable before the beginning of T. Yet, to repeat the reason given above, no (whole) event E at a time T2 is fixable at any time T0 ending earlier than the end of T2. This is because, however strong an agent was at T0 it remains possible that he changes his mind at a later time T1, beginning at the end of T0 and ending at the end of T2, and so stops (some part of) E occurring. Yet, since an agent at any earlier time T0 could (unless he changed his mind, or was prevented by the act of a stronger agent before the end of T1) cause what happens at T1, and thereby what happens at T2 what happens at T2 cannot be fixed at T0. Hence every event is unfixed at every time ending before it ends.

I now show that it follows that no event E at T2 can be caused directly and so fixed by any agent D acting at any later time. As we have seen, a direct cause of an event E at T2 must act for the whole time while E occurs, as well as for some further time when E does not occur. In the case of a supposed later cause D, that further time T3 would be immediately subsequent to E. But only what is fixed can fix some other event; until D’s action is fixed that action cannot fix E. For, while an event is not fixed, an agent F of sufficient strength acting at an immediately earlier time could always prevent that event from happening. So D’s action at T3 could be caused not to occur by an agent of sufficient strength acting at T2, and then its effect (E) would not happen.

So no future event can fix, that is directly cause, an earlier event, because until it occurs it is still fixable. Yet, if an event cannot have a later direct cause, it cannot have any later cause. For such a later cause would be connected to the direct cause by a chain of causes, each of which was the direct cause of the next one; and some of these direct causings would be of an earlier event by a later one, and so ruled out by the previous pattern of argument. So backward causation is not logically possible; every event is fixed at all times after it has occurred. In summary and very loosely, since forward causation is always logically possible, backward causation is never logically possible.

It follows that not merely direct simultaneous causation, but also indirect simultaneous causation, are impossible. For, if C caused an event E simultaneous with the act of causing it by directly causing F, which directly caused E, F would have to occur either (1) simultaneously with C’s act, or (2) after C’s act, or (3) before C’s act. (1) is impossible because of the impossibility of direct simultaneous causation; (2) involves backward causation in the respect that F causes E; and (3) involves backward causation in the respect that C causes F. And the same problems arise for any postulated longer chain of causes by which C causes E. Hence the logical impossibility, not merely of all backward causation, but of all simultaneous causation.

I conclude that it is not possible that a human free action can cause God’s earlier belief that it occurred. Hence a temporal God cannot have essential foreknowledge of human free actions. Further, we cannot avoid the need for some restriction on the omniscience of a temporal God by supposing that humans do not have freewill. For God himself is normally supposed to have uncaused free will limited by an inability ‘to do evil’, and, I suggest, also limited by an inability not to do the best possible action when there is such an action. However God will surely often have a choice between incompatible equal best actions, and between an infinite number of incompatible good actions, each less good than another such action. In these circumstances a truly omnipotent God is surely free to choose what to do. Yet he would not have a free choice if he infallibly foreknew what he would do.

A Timeless God could not Foreknow Human Free Actions, or Cause Temporal Events

But of course the view of most Christian theologians from the third century until the present day is that God is outside time, and so his knowledge of any human free action occurs timelessly, and is neither before, simultaneous with, nor after that human action. However I do not think that a timeless God could know or cause actions in time. I argued earlier for the claim that for an agent to cause an effect at a time T, the agent must act directly at that time (as well as at a contiguous time). If that is correct, then a free human agent acting at T could only directly cause God’s belief that it occurred if God exists at T, that is in time. And if the human agent causes God’s belief indirectly by causing some other event which in turn causes God’s knowledge, then the latter causing requires God to be in time. Not everyone may find it plausible to hold that a temporal action cannot cause timeless effects, but they are more likely to find it plausible to hold that a timeless agent cannot directly cause, and so fix, effects in time. For given that a strong enough agent can always fix an event E occurring at T, by an act beginning before T and continuing through T, a timeless agent could not prevent that and so could not fix E. The intuition about causation to which I am appealing is that to cause an effect directly at a time T, an agent has to act on the world at the time T; for that is when and where the causal agency effects its result. So a timeless God could not create or interact with the world, and (probably) could not even learn about it; such a god would not be the God of Christianity, Judaism, and Islam. Hence – to my mind – the claim that divine omniscience can be shown compatible with human freedom by supposing God to be timeless, fails.

For all these reasons I conclude that God’s omniscience is limited by his omnipotence which prevents him from knowing what he will himself choose to do in future, and allows him to create free creatures whose future free choices he will not know infallibly. Various writers, including Fischer (2016, 45) claim that ‘Scripture’ implies that God foreknows future human actions. It certainly implies that God foreknew some human actions, but it is surely not a Christian (Judaic, or Islamic) doctrine that every intentional human action is free; nor is it psychologically plausible to suppose this. And – in my view - Scripture implies that God did not foreknow some human actions. But for justification of that view, I refer readers to Swinburne (2016, 197-9).

Bibliography

• • Fischer, John Martin, 2016 Our Fate, Oxford University Press.
• • Hasker, William, 1989, God, Time and Knowledge, Cornell University Press.
• • Pike, Nelson, 1965 ‘Divine omniscience and voluntary action’ Philosophical Review, 74, 27-46.
• • Swinburne, Richard, 2016 the Coherence of Theism, second edition, Oxford University Press.

End Notes

[1] The central arguments of my paper are based on material contained in Swinburne (2016). I am much indebted to Alexander Pruss for his criticisms of earlier definitions of mine of a ‘hard event’, which led to my present definition of a hard event in the first sense. The argument in Swinburne (2016) against the possibility of backward causation assumed the first sense of ‘hard event’, but I give the same argument here, since clearly if backward causation of events which are ‘hard’ at some time in the first sense is impossible, that is because of the features that they possess at that time and so because they are ‘hard’ in the second sense.

[2] I am assuming that an expression denoting a period (or instant) of time, such as a particular year, picks out the time it does on our current usage, independently of what in fact happens before or after that time. This aspect of our usage is shown by the fact that the names of years (e.g. as ‘1988’ or ‘1990’), originally given to them on the basis of their supposed distance in years from the year of the birth of Jesus (1 ce) have been retained, despite the current general belief of scholars that Jesus was born a few years earlier than previously believed.