Open Theism and Promises
November 1, 2007 — 20:16

Author: Alexander Pruss  Category: Open Theism  Tags:   Comments: 21

Thesis: If Open Theism is true, then either possibly there are some things some humans know that God doesn't know or possibly God has some false beliefs.  The conclusion is absurd, hence we should reject Open Theism.

Argument:

For simplicity I will abbreviate "x does not fulfill the promise P and nothing prevents x from keeping P, nor does x conclude a defeater arises, nor does x forget" as "x violates P". 

  1. Let x and y be humans.  Often when x knows y's character to be solid and knows y to have promised to do something for x, x is not only justified in believing y will not violate the promise, but x knows it.
  2. If Open Theism holds, there is no possible divine doxastic policy such that, necessarily: (In every case like the one described in (1), God believes y will not violate the promise, and God has no false beliefs).
  3. Therefore, whatever possible divine doxastic policy is adopted, either in some worlds there will be a situation like that described in (1) where x knows y will not violate the promise, but God doesn't believe it and hence doesn't know it, or in some worlds God will have a false belief.
  4. This is absurd, and so Open Theism should be rejected.

Moreover, it is very likely that in the actual world either there are going to be cases of (1) where God doesn't know y will not violate the promise, or cases where God has some false belief.  For there are so many cases of (1) in the actual world, that it is highly likely that any divine doxastic policy that doesn't involve foreknowledge of free actions will either miss some cases of (1) or will lead to belief in some cases where y in fact violates the promise.

To fill out the argument, I need to argue for (1) and (2).  I think (1) is true.  We certainly say things like: "I knew you wouldn't let me down."  I think we should take them at face value.  If we don't because we're worried about the fact that y might still freely choose to do the wrong thing, then likewise we should deny that the testimony of a reliable witness of solid character gives knowledge, because we should likewise be worries that such a witness still might have freely chosen to lie.

What about (2)?  Well, the intuition here is this.  If God doesn't have foreknowledge of future free events, his policy will be based on his evaluating evidence.  When the evidence that y will not violate the promise is strong enough, God will believe it;  when not, he won't.  Now the bar for strength of evidence is either higher than that which we use when coming to believe that someone's character is solid enough that we can count on her not violating the promise or not higher.  If it is not higher, then just we sometimes judge falsely based on the evidence, so, too, God will sometimes judge falsely based on the evidence.  But if the evidential bar is higher, there will surely be at least some logically possible, and probably some actual, cases like (1) where we will be more epistemically bold than God, and where we'll nonetheless get it right and have knowledge.  So then sometimes we'll know something God doesn't know.

The argument works best against Open Theisms that allow for non-trivial truth-values of future tensed contingent propositions, but I think a variant of it may work in general.

Comments:
  • John Alexander

    Alex
    I think that you are using a stronger concept of ‘know’ in 3 regarding x then the one used regarding God. If the evidence for x and God re believing y is the same then the strength of their claims regarding the truth that y will keep her promises are the same.
    Do you think God can be surprised? I know that I can be. What if God and I believe, based on past our past experiences with y, that y will keep her promise, but it turns out that she does not. God and I were not inductively incorrect to believe what we did regarding y. It simply turns out that y’s actual action fits the negative probabilities that she will not keep her promises; small though they may be.

    November 2, 2007 — 9:14
  • John:
    Whether the argument in your second paragraph succeeds will depend on whether the content of the belief is that y will not violate her promise, or that y probably will not violate her promise. It is true that if the latter is the content of the belief, then it is not falsified by a violation of the promise. But I think we do go further–we not only believe that it is likely that y won’t violate the promise, but we believe that y won’t violate the promise. That is how trust works–if it’s about probabilities, it’s not trust but qualified suspicion.

    November 2, 2007 — 11:38
  • Acknowledgment: I’ve been told by Jon Kvanvig and David Alexander that they have given arguments like this before.

    November 2, 2007 — 12:17
  • Justin Capes

    Alex,
    I’m wondering how the argument is supposed to work for Open Theists like (our own) Alan Rhoda who deny that there are any contingent future-tensed propositions of the form “S will do A.” If they are correct, then premise (1) seems false, for assuming a JTB+ analysis of knowledge, we don’t really have knowledge when we say things like “I know Alex will have a new post next week,” since, if you are free, there is no determinate truth about whether you will post next week. When we say things like this, it would seem, we mispeak, or are making the weaker probability claim.

    November 2, 2007 — 12:55
  • Justin:
    A good question. I left a promissory note in the last paragraph of my paper about this.
    I now think it’s a bit of a harder issue than I thought when I posted. But here’s roughly what I was thinking–maybe someone can improve on it.
    Response 1. Open Theists who deny that there are true propositions about future contingent actions contradict point (1). But point (1) is true.
    Response 2: Consider cases where we appropriately say after someone has kept a promise: “I knew she would come through.” Taken literally, this can’t be right on those versions of Open Theism.
    However, we should be able to say that in retrospect such-and-such a past prediction was right. This cannot be parsed as implying that the past prediction at the time it was made had some property, viz., of being right, but that it now has that property. Rather, this is a kind of retrospective attribution. Here’s a similar example. Consider the claim: “An hour ago, Smith was drinking the fatal cup.” There, “fatal” does not refer to some property which the cup had had then. For it might well depend on the then-future free actions of others whether the cup would be fatal (e.g., if they induce vomiting in Smith, the cup won’t be fatal). But now, from the standpoint of the future in which Smith is dead, “An hour ago, Smith was drinking the fatal cup” is quite true.
    So there may be a retrospective rightness that we can attribute to predictions. Then the claim would be that either sometimes God made predictions that in retrospect have turned out to be wrong, or that sometimes people have made predictions that were right, predictions that God did not make.
    This argument is less than compelling. For if the predictions weren’t true when they were made, they are no threat to the claim that God knows more than anybody else. I prefer Response 1.

    November 2, 2007 — 14:44
  • Alex,
    In addition to the issue suggested by Justin, your argument equivocates on “know”. The kind of knowledge that is relevant to questions of divine omniscience is infallible knowledge. The kind of knowledge that is relevant in human contexts is something like JTB, in which J does not entail infallibility.
    So your (1) is true iff “knows” means “fallibly-knows” (F-knows, for short), whereas your (2) clearly has infallibility in view as expressed in the consequent:
    necessarily: (In every case like the one described in (1), God believes y will not violate the promise, and God has no false beliefs)
    Hence, in your (3), God’s knowing has to mean “infallibly knows” (I-knows, for short).
    When the ambiguities are clarified, the apparent absurdity vanishes.

    November 2, 2007 — 15:08
  • Alan:
    I use “knows” to mean “F-knows” throughout. So, my claim is that if Open Theism holds, then either possibly there are some things which humans F-know and God doesn’t even F-know, or possibly there some things God believes that are false.
    I don’t see why (2) is supposed to imply that I am talking of I-knowledge. After all, in (2) the only persons who knowledge is talked about is the humans referenced in (1).
    The relevant part of (3) reads “where x knows y will not violate the promise, but God doesn’t believe it and hence doesn’t know it”. This works just as well whether “know” is “F-know” or “I-know”, since both F-knowledge and I-knowledge imply belief. We should read this as “F-know” for consistency.

    November 2, 2007 — 15:21
  • One more point.
    Your “Response 1” also depends on an equivocation. The sense in which (1) is true is the sense in which “knows” means “F-knows”. But the sense in which I deny that there are true future contingents is the sense in which “will” and “will not” connote inevitability, it’s being now-unpreventably the case, which I argue to be requisite for God to I-know the future.
    If you prefer to use “will” and “will not” in a weaker sense that does not entail now-unpreventability, fine. But then you shouldn’t be pressing analogies between human F-knowing and divine I-knowing. They do not have the same necessary and sufficient conditions.

    November 2, 2007 — 15:21
  • Let x and y be humans. Often when x knows y’s character to be solid and knows y to have promised to do something for x, x is not only justified in believing y will not violate the promise, but x knows it.
    Alan, I think Alex is claiming that, under the assumption of open theism, there are cases in which both of the following are true: (1) x is human and knows that y will do A and (2) God is omniscient and does not so much as believe that y will do A. It does not matter whether you think God has I-knowledge or F-knowledge. The fact is that he has neither, if Alex is right. The problem is how we might F-know that y will do A when God does not (F or I)-know that y will do A. So I don’t think the argument equivocates. Did I miss something?

    November 2, 2007 — 16:03
  • Just read your reply, Alex.
    So you don’t have I-knowledge in mind in (2). Okay. I had thought you did because apart from that reading (2) strikes me as false.
    First, it is clear that you want y’s not violating his promise to be a future contingent. In that case, it can only be true that “y will not violate the promise” if “will” does not connote now-unpreventability. Thus, you are committed to holding that “will” is compatible with “might not”, which in turn is compatible with “does not”. Hence, by transitivity of compatibility, “y will not violate the promise” is compatible with “y violates the promise”. Hence, the latter does not falsify the former. So, in this sense of the word “will”, God can truthfully believe that y will not violate the promise even if y does violate the promise.
    Now, that seems to me an effective reductio ad absurdum of the supposition that “will” (understood in a strict and unqualified sense) is compatible with “might not”. Either future contingency has to go, or the truth of “y will not violate the promise” has to go. I opt for the latter.

    November 2, 2007 — 16:15
  • Hi Mike,
    Thanks for your simplified reconstruction. It helps.
    As you suggest, let’s set the F-knowledge/I-knowledge issue aside. I continue to maintain that the human and divine cases are parallel.
    The shape of my argument depends upon whether one grants that “will”, when used in a strict and unqualified predictive sense, implies now-unpreventability or not.
    If it does, and if y’s keeping his promise is not now-unpreventable, then it is simply false that y will keep his promise, and so neither humans nor God can properly be said to “know” that proposition.
    If it doesn’t, then “will” is compatible with “might not” and with “does not”, in which case there is no reason why both God and humans can’t both “know” that y will keep his promise.

    November 2, 2007 — 16:26
  • Thus, you are committed to holding that “will” is compatible with “might not”, which in turn is compatible with “does not”.
    I understand the first compatiblity claim. But I can’t see the second compatibility. If it is true that I do not keep the promise (in the future, I take it) then how could that I will keep the promise?

    November 2, 2007 — 16:53
  • Mike

    correction–
    “how could [it be true] that I will keep the promise”

    November 2, 2007 — 16:55
  • Alan:
    In your dichotomy, you’re assuming that if p is true now and p is incompatible with the non-occurrence of some future event, then that future event is now unpreventable. That’s an assumption I don’t grant, but it is, of course, the founding assumption of open theism, so we’re not likely to make progress there.
    I do not understand any sense of “will” such that “will p” is compatible with “will not-p”. I don’t know what the truth conditions for such a “will” would be. I suspect that if one spelled it out, this “will” would turn into a “possibly” or something like that.

    November 2, 2007 — 17:24
  • I just got back from lunch, and before I refresh the browser and see if there have been any more responses, I’d like to make an announcement:
    I hereby retract the “transitivity of compatibility” principle that I appealed to above.
    The principle is false, and obviously so, as I realized about 5 minutes after my previous reply. What was I thinking?

    November 2, 2007 — 17:33
  • Now that I’ve established my own fallibility, I’d like to revise my reply to Mike above.
    The first half of my reply still stands. If “will” connotes now-unpreventability, then neither humans nor God can know that y will not violate the promise because the proposition in question is not true, and so is not a possible object of knowledge.
    The second half of my reply needs to be fixed, however. In the looser sense of “will” that is compatible with “might not”, a human can F-know that y will not violate the promise provided that y does not violate the promise. But in that same sense, God could also F-know it if God could believe it. Could God believe it? Sure, provided that we understand the content of the belief as something like “y will probably not violate the promise.”
    But even if I’m wrong about this and there is a doxastic asymmetry being humans and God on this point, I’m inclined to just shrug it off. So what if humans can believe and F-know things that God does not know because God cannot believe anything that is possibly false? God will then not believe “y will not violate the promise” but he will believe and I-know something in the neighborhood: “y will probably not violate the promise”. The difference between humans and God at this point, insofar as there is a difference, is not one that arises from any imperfection in God’s doxastic performance, but rather one that arises from human limitations.

    November 2, 2007 — 18:08
  • John Alexander

    Alex
    You write’ “if it’s about probabilities, it’s not trust but qualified suspicion.” I do not agree with this. You seem to make trusting someone reducible to beliefs about statements regarding that person that are necessarily true in the sense that conclusions in deductive arguments are true if the premises are true and the conclusion follows from them. I do not see how the practice of promising and the beliefs we develop about promising can ever fulfill this degree of certainty. We derive our beliefs re promising based on the sample size of those whom we are considering regarding their keeping promises. If y has kept her all her promises to me, and others that I know she has made promise too, then I know to a high degree of probably, but not certainly, that she will keep her promise in a particular situation and this is sufficient for me to trust that she will. I will be surprised if she does not. I do not think that God’s knowledge is any different. Imagine that in the past y has failed to keep her promise but that she had an excuse that justified her doing so. If this has occurred this does not diminish her credibility in making new promises or by ability to trust her to keep it. I know that if she does break her promise that she probably has a good reason that explains and justifies her doing so, but that the probability of this occurring is very small. We cannot rule out that conditions will not arise between the making of the promise and fulfilling the promise that will warrant y in not keeping her promise. This is one reason why we cannot have certainly even if our sample is 100% regarding past instances of her keeping her promises. I think that you are using ‘know’ in the sense of being certain in #1 above and I do not see that this is a proper definition of ‘know’ in these types of scenarios. As is said earlier I think this problem of usage extends into other premises in your argument.

    November 2, 2007 — 20:05
  • Alexan

    John:
    Knowledge doesn’t require probability 1, but does require belief. There is a difference between believing that something is the case and believing that it is probably the case, and I think in cases of trust we have the former. That had better be the case, because most of what we know we know on trust without probability 1. How do I know that the earth is round? Because people say so–I have no first person evidence of it (I have never circumnavigated the globe, nor have I seen the earth from space, nor have I observed in the desert or at sea the tops of buildings or ships appearing first at the horizon). Could some of them be lying and the rest be deceived? Well, that possibility can’t be excluded completely. But nonetheless, I know that the earth is round, not just that the earth is probably round.

    November 2, 2007 — 23:19
  • John Alexander

    Alexan: You write: “Knowledge doesn’t require probability 1, but does require belief.” I agree. This is going to be pretty standard stuff, but let us agree that I believe x to be true. Can I believe x to be true when in fact x is false. If I can, which I have, then my believing that x is true is a necessary but not sufficient condition for me knowing x. So I need some additional conditions to be meet that will be necessary and jointly sufficient for me to claim that I know x is true. This focuses us on the criterion of ‘being justified in believing x to be true.’ One way to be justified in believing x to be true is to maintain that it is certain that x is true, e.g., all bachelors are unmarried men. This is certainly true and I also believe it to be true. But promising does not fulfill the condition required for certainty; it is not analytically true. There is nothing in the act of promising that entails that if y promises something to x that y necessarily will keep her promise. If it did then no promises would ever be broken, or those that appear to be broken would not be promises. I can establish a degree of probability that y will keep her promises based on the sample of y’s promises and her keeping them that I know. It may even reach the point where I become ‘certain’ that she will keep her promises because she has never, to my knowledge, broken any promises she has made, but this is not epistemic certainly, it is psychological certainty.
    Please note that I have brought forth nothing in this discussion that has not been already brought forth by JL Austin in his discussion of promising and excusing, GE Moore in his discussion of different types of certainly, and Hume and Ayer on the distinction between knowledge based on ‘relationships of ideas’ as compared to knowledge based on ‘matters of fact’ (to use Hume’s terminology). I happen to think they are right.

    November 3, 2007 — 9:36
  • John Alexander

    I should have made it clear as my discussion evolved that the statement “I agree” made above is in relationship to knowledge that is certain. I am sure everyone realizes this, but, just in case, I want to make it clear.

    November 3, 2007 — 9:49
  • John Alexander

    I should additionally point out that under this analysis God can believe x to be true and have x turn out to be false. The practices of promise making and promise keeping do not provide a framwework for rejecting open theism. I will be quiet now!

    November 3, 2007 — 9:53