[cross-posted at Parableman]
I’m working on a chapter for the forthcoming Blackwell Philosophy and Harry Potter on the topic of destiny, and one of the things I’m trying to do in the chapter is distinguish between different metaphysical analyses of prophecy. I’ve come up with three, and I’m inclined to think that it might be exhaustive enough for the purposes of a popular-level work like this, but I’m curious if anyone here can think of any others.
Here’s what I’ve got (and how I’m presenting it in the draft I’m writing):
1. They involve mere likelihoods. No one has access to the actual future, but someone might have magical access to information that’s derived from what’s likely. Given what’s true about the various people involved, it’s very likely that a certain outcome will happen. That means prophecies, even the ones Dumbledore is inclined to call genuine, are not infallible. They can turn out get it wrong.
2. They do not derive their content from the actual future. Rather, they make the future happen. When a genuine prophecy occurs, it influences those who hear it in such a way that they end up doing things that will fulfill the prophecy. This kind of prophecy is self-fulfilling in a very literal sense.
3. The seer has some intuitive connection with the way things will really happen, such that the words of the prophecy are true about a future that really will be that way. If it’s a genuine prophecy, it can’t be wrong, because its origin lies in the very future events that it tells about. In the same way that a report about the past can bring knowledge about the past only if there’s some reliable connection with the actual events in the past, a genuine prophecy in this sense must derive its truth from a reliable method of getting facts about the future.
My understanding of J.K. Rowling’s view of prophecy, judging by this interview and my sense that the Albus Dumbledore character represents her views when he discusses this issue with Harry Potter, is that she wants to treat Professor Trelawney’s two genuine prophecies as the first kind, a kind of prophecy an open theist could accept.
There are hints in at least two of Dumbledore’s conversations with Harry that he thinks something like the second kind is going on, but it’s clearly not a reduction of prophecy to what happens in #2, because the characters in question (mostly Lord Voldemort) still make free choices and aren’t simply caused by the prophecy to do anything the way some ancients thought Laius was caused by Apollo’s prophecy to do what he did that led to Oedipus eventually killing him.
My argument at this point is that there isn’t really a way for Dumbledore to distinguish between Trelawney’s two genuine prophecies and all her vague predictions that can often be interpreted as coming true unless the genuine ones are of the third kind (because the pseudo-prophecies are of the first kind, and the genuine ones can’t be completely explained by the second kind). Rowling doesn’t seem to want to accept that, and Dumbledore is clearly with her, so there’s a consistency issue here both for the character and the author. But my argument depends on the options I’ve listed being exhaustive. Is that true?
Let me add a third beat-up-on-Molinism post. Here is the idea. Ordinary language makes use of counterfactuals of free will (CFWs). But ordinary CFWs are not Molinist in kind. For one, Molinist CFWs have no truthmakers, or at best have Platonic kinds of truthmakers (a pair of abstract propositions standing a property of subjunctive conditionality). Ordinary CFWs have concrete truthmakers.
Here are some examples of ordinary CFWs:
- George grabs some money off the table not knowing or caring whether it is his or not. As a matter of fact, it is his. A standard criticism to make of George is: Were the money not his, he would still have taken it.
- Martha kisses a man she just spent two hours talking with at a bar. Had this man not been her husband, she might not have kissed him.
- Patrick is an optimist, and as gets up in the morning, he chooses to believe that he won the lottery for which the draw was last night. He looks at his newspaper, and finds he was right. Nonetheless, we say that Patrick did not know he won until he looked at his newspaper, and one reason he did not know was because he would have believed that he won even had he not won.
- Curley takes a smaller bribe. Had he been offered a larger, he would have taken it. (This example is Plantinga’s.)
The four above CFWs are made true by the causes, motives and reasons behind the agent’s actual choice. Thus, that George took the money and did not know or care whether the money was his is what makes it true that he would have taken it even if it were not his. What makes it true that Martha might not have kissed the man had he not been her husband is that among the considerations guiding Martha’s decision to kiss the man, there was the fact that he is her husband. What makes it true that Patrick would have come to the same conclusion had he not won is that he chose to believe he won on grounds that would have been the same had he not been the winner. And what makes it be the case that Curley would have taken the larger bribe is that he took the smaller, and all the reasons for taking the smaller bribe are available for taking the larger, and there are no additional reasons against taking the larger bribe (if there are, e.g., if the penalties are higher, then we ordinarily cannot affirm the counterfactual in (4)).
There is a way in which these kinds of ordinary CFWs do not state some further counterfactual fact–rather, they are a convenient restatement of the facts about the causes, motives and reasons behind the agent’s actual choice.
Because these CFWs are made true by concrete things, they are not Molinist CFWs. In fact, I suspect that most Molinists (I am thinking of Tom Flint in particular, who made a remark that commits him to this) would say that the corresponding Molinist CFWs can have different truth value from the ordinary ones. For instance, take (2). On standard Molinist views, it is highly plausible that there is a possible world w where Martha kisses her husband, and where the fact that he is her husband is one of the reasons for kissing him, which is sufficient to make true the might conditional in (2), but where in w there also holds the Molinist CFW that had he not been her husband, she still would have kissed him. Thus, the Molinist CFW would conflict with the ordinary one. Likewise, Molinists are apt to agree that Molinist CFWs could be such that were Patrick not to have been the winner, he would not have chosen to believe that he was the winner. The Molinist CFWs hang loose from the ordinary ones, then, and are irrelevant for the kind of nomative evaluation that we make. If Patrick is lucky enough that for him the Molinist CFW holds that had he not been the winner, he would not have chosen to believe he was the winner, that doesn’t make his belief knowledge.
But now we have a problem for the Molinist. Nomic counterfactuals and ordinary CFWs are different kinds of counterfactuals from Molinist CFWs. But then how do we ever acquire a concept of a Molinist CFW? And if we don’t ever acquire such a concept, then it seems all the Molinist stuff is, literally, nonsense.
Suppose that a Molinist God creates a world where there is a sequence of 1000 indeterministic throws of a fair coin, and suppose that middle knowledge extends to stochastic non-agential events. (My argument will also apply in the case of Thomist God who determines indeterministic events.) Suppose 514 of the coin throws, let us suppose, are heads and 486 are tails. Consider the fact p that approximately half of the throws landed heads. A standard scientific explanation of p would involve the following facts:
- The coin was fair: heads and tails each had probability 1/2.
- The individual throws of the coin were independent of one another.
- If (1) and (2) hold, then by an appropriate version of the Law of Large Numbers, it is likely that a sequence of 1000 throws of the coin would have approximately half of them be heads.
Fact (3) is a mathematical fact. Facts (1) and (2) are concrete facts about the situation at hand, and both are essential. If (1) is false, we might well expect a different heads-to-tails ratio. If (2) is false, then the Law of Large Numbers need not apply.
But this scientific explanation is unlikely to be correct if Molinism holds. For if Molinism holds, then God in effect controls what sequences of throws come up, by choosing the antecedents of counterfactuals. God makes the choice of sequence based on global providential considerations. Since the sequence is chosen on the basis of considerations of the sequence as a whole, it seems unlikely that the items in the sequence will be independent.
Suppose we say, as I suggested in the previous thread in response to Mike’s related concern, that God deliberately chooses a sequence of events that is statistically apparently random. Then p will still be true–about half of the throws will land heads. However, (2) will not be true, at least not if we condition on God’s choosing a sequence of events that is statistically apparently random. For, if (1) and (2), hold we have a non-zero probability that all the throws will be heads. But conditionally on of (1) and the claim that God chose a sequence of events that was statistically apparently random, we get a zero probability that all the throws will be heads, since if all the throws were heads, the sequence could not be statistically apparently random.
Perhaps we shouldn’t condition on God’s choosing a sequence of events that is statistically apparently random. But if we don’t condition on that, then to check whether (1) and (2) we need to compute the probabilities of all the possible choices God could have made. And we have little reason to think (1) and (2) will hold then.
In Molinist worlds nothing is left to chance, not even undetermined events. God does not choose any object or person or circumstance randomly in anyone’s life-history. Tom Flint describes the traditional view of providence captured in the Molinist account.
Isn’t it natural to think that He has arranged it so that, not just some things, but everything fits together in such a way that his love is made manifest? Isn’t it natural to think that nothing is left to chance, that nothing haphazard or unexpected from the divine perspective occurs . . .? (p. 13, DPMA).
If God is directing each and every undetermined event toward his chosen goals, then we should not observe a chance pattern in the occurence of those events. We should rather observe evidence of God’s direction. We should find the frequency of undetermined events diverging from their chances. The problem for Molinism is that there is no evidence that God is using counterfactuals of creaturely freedom to direct undetermined events in the world. We simply do not observe any divergence between frequency and chance.
Here’s a moderate Molinism. It has two parts.
1. Accept that subjunctive conditionals of free will (“F-conditionals”) have non-trivial truth values if their antecedents are sufficiently determinate, and God knows these truth values.
In this respect, the moderate Molinism is exactly like standard Molinism. Whatever qualifiers one wants to put on when F-conditionals have non-trivial truth values can be put in.
2. Deny that God can make use of his knowledge of an F-conditional with antecedent p as part of a reason for bringing it about that p.
Why go for (2)? In order to avoid the priority in the order of explanation argument: if God makes use of his knowledge of an F-conditional with antecedent p as part of a reason for for his bringing it about that p, then the truth of that F-conditional, as well as the truth of p, are explanatorily prior to the free action in the consequent, and hence the free action is entailed by facts explanatorily prior to it, which contradicts a plausible principle of alternate possibility.
If it’s logically impossible for anyone to make use of knowledge of F-conditionals with antecedent p as part of a reason for bringing it about that p, then this is not a problematic restriction on omnipotence.
Moreover, the idea that God has to bracket some of his knowledge of F-conditionals when making decisions seems quite plausible. For instance, plausibly, he has to bracket his knowledge of F-conditionals that, together with his knowledge of their antecedents, entail that he is going to do A, when he is deciding to do A.
Careful readers will have noted that I posted an argument against Molinism earlier this morning, and it committed a modal fallacy. I took down the argument as soon as I realized the fallacy. Here’s an argument that doesn’t seem to commit the same modal fallacy, but the cost of it is that it has some much more controversial premises. Let C a complete description of the circumstances at the time of Jones’ choice. The main point of Molinism is to make possible situations like this:
- Were Jones in C, he would freely choose to mow the lawn.
- Because of (1), God brings it about that Jones is in C.
Now add some statements that are, plausibly, conceptual truths, for a reductio:
- If p is explanatorily prior to Jones’ choosing what to do in C, and p entails that Jones will choose to mow the lawn, then Jones does not freely choose to mow the lawn. (This is a version of the Principle of Alternate Possibilities.)
- Explanatory priority is transitive.
- If, because of q, God brings it about that p, then q is explanatorily prior to p.
- If C is a complete description of the circumstances at the time of Jones’ choice, then that Jones is in C is explanatorily prior to Jones’ choosing what he chooses.
- If p and q are explanatorily prior to r, then p&q is explanatorily prior to r.
The argument now is easy. By (1) and (2), Jones is in C and freely mows the lawn. By (2) and (5), conditional (1) is explanatorily prior to Jones’ being in C. By (6), Jones’ being in C is prior to Jones’ choosing to mow the lawn. By (4), it follows that conditional (1) is explanatorily prior to Jones’ being in C. Let p be the conjunction of (1) with the claim that Jones is in C. By (7) and what we have already shown, p is explanatorily prior to Jones’ choosing to mow the lawn. But p entails that Jones chooses to mow the lawn. By (3), Jones does not freely choose to mow the lawn. But by (1) and (2) he does. Hence, a contradiction ensues.
Suppose Molinism is true. It seems that there will be an interesting and non-trivial probabilistic structure to the space of F-conditionals (subjunctive conditionals of free will). Some examples:
1. Almost implication where the reasons are better: P(Curley would accept a $10,000 bribe in circumstances C | Curley would accept a $5000 bribe in circumstances C) is high (but maybe not automatically 1).
2. Probabilification where the reasons are similar: P(Curley would accept a $5001 bribe in C | Curley would accept a $5002 bribe in C) is moderately high, but lower than the probability in 1.
3. Independence between non-interacting possible persons: P(Curley on earth would accept a $5000 bribe in C and Jones would accept a $5000 bribe in D) = P(Curley would accept a $5000 bribe in C) P(Jones would accept a $5000 bribe in D), if C and D are such that there is no causal contact between Curley and Jones.
4. P(Curley would accept a $5000 bribe in C | Curley would accept a $5000 bribe in C*) = 1 if C and C* differ only in respect of something not in the causal history of Curley's bribe decision.
Now some of these can be questioned, but I think they all have something going for them. But never mind the details. All I need is the claim that there are some probabilistic laws governing the relationship between different F-conditionals.
I don't think the probabilities here are just epistemic.
So now let me ask: Where does the objective probabilistic structure of F-conditionals come from? What grounds it? I am inclined to think that objective probabilities tend to flow from laws of nature (the other alternative is some sophisticated version of the principle of indifference, but nothing like that will yield laws like (1)-(4)). But that doesn't seem right here, especially if one is inclined as I am to think that the laws of nature are grounded in the powers and propensities of existing (and not counterfactual) entities.
(One can make of the above an argument against Molinism: If Molinism is true, there should be such a probabilistic structure, but there can't be. But one can also make of it an argument for Molinism: (1), (2) and (4), at least, are highly plausible in themselves; but they do not make sense apart from Molinism.)
I've never had trouble with the idea of God creating a world with an infinite past. (Perhaps being a B-theorist and seeing God as outside of time helps?) But for some years now, I've been wondering whether it would be possible for God to create a world containing a backwards infinite sequence of interconnected libertarian-free (l-free) choices. This would be a sequence where for each l-free choice there was an earlier l-free choice, and the earlier choice contributed substantially to the conditions under which the later one was made.
Why would this be any more of a problem that God just creating a run-of-the-mill world with an infinite past? Well, I still don't have a very clear argument, but maybe if I share my considerations one of you will be able to help me formulate a good argument–or show me how I am confused. As background, I am assuming that both theological compatibilism and Molinism are false. The former assumption is essential to setting up the problem. I do not know if the latter is.
Now in the case of God creating a world containing free agents but no backwards infinite sequence of interconnected l-free choices, it is easy to find a model for creation. God directly creates (i.e., strongly actualizes) an initial state, and then lets things evolve, intervening at appropriate points, letting l-free choices happen, and of course cooperating in the choices of creatures and in other causal events in whatever way the correct doctrine of continual creation requires, thereby weakly actualizing the whole world. No problem.
There is perhaps also no problem with God's creating a deterministic world with an infinite past. God simply strongly actualizes an infinite past all at once, with all the events in it in their appropriate causal relations (assuming this makes sense–I am not completely sure at this point). However, if the infinite past contains l-free choices and if theological compatibilism is false, this model fails. I, however, the infinite past only contains finitely many l-free choices the model can be salvaged–God just creates all at once the past prior to the first l-free choice.
Another model for God's creating a deterministic world with an infinite past is that God sets the boundary conditions at minus infinity, and lets things evolve according to the laws. But that, too, will not work here. For the boundary conditions at minus infinity are going to be a limit of the conditions at large negative finite times. But these conditions include the choices of the agents at these past times. Thus among the limiting conditions there will be limiting properties of the sequence of choices as one goes back in the sequence. For instance it might be that the choices as one goes further and further back increase in virtue, tend asymptotically to a single level L of virtue. Then one of the boundary conditions will be that the limiting level of virtue in choices as one goes back is L. But if to create something with an infinite past one sets boundary values, then this fact about the limiting level of virtue in choices is one that God has to set. But how can God do that if the choices are l-free?
The basic difficulty here is in spelling out how the explanatory interaction between human choices and God's creation works in this case. The conditions, internal and external, under which a choice is made are partly explanatory of the choice. (The problem reminds me of the following question: Suppose two people are given qualitatively the same choice at the same time. Can God bring it about that they make the same choice? Can God bring it about that they make different choices?)
This is vague. I know. I can't do better right now.
Since there’s still little going on here, I thought I’d direct readers to another post in my series based on my introductory philosophy course lecture notes. This time it’s on foreknowledge and freedom. Again, I don’t expect it to include anything newsworthy for many readers of this blog, since we’ve discussed all these issues here in much more depth in the past, but I’ve tried to summarize the main moves in the discussion at a level someone in an introductory course could understand, and some may want to take a look at that or offer feedback. Newer readers less familiar with our discussions on this topic or with the literature on the issue may find it informative as well. I did try to include the most current work on the subject.