The analogy of God as the author and us as his characters has a venerable history. Here I want to object to one use of the analogy as a way of resolving the tension between providence and creaturely causation, deterministic and especially indeterministic. The puzzles the analogy is addressing are like this:
- How can it be that horses evolved fully under the influence of random stochastic processes, and yet we can also explain the existence of horses in terms of the way they glorify God?
- How is it that Francine freely chose to accept baptism in the name of the most holy Trinity, and yet the choice was entirely caused by God's grace?
- Colonel Mustard was murdered because the author believed that books about murdered colonial colonels sell well.
- Colonel Mustard was murdered because he knew that Captain Catsup was not as great a tiger hunter in India as he claimed to be.
This use of the author analogy is mistaken for a simple reason. The "because" in (4) is in the scope of a fictionalizing operator. What (4) really says is:
- According to the story (Colonel Mustard was murdered because he knew that Captain Catsup was not as great a tiger hunter in India as he claimed to be).
Objection 1: This is unfairly pressing the analogy. The case of God is like that of the author, except that the story is true.
Response: I am inclined to agree with the second sentence. But the author analogy was given to solve a puzzle about apparently competing explanations in (1) and (2). No solution is given by an analogy to a case where there is only one genuine explanation and one truth-canceled explanation, since what made the non-competitiveness claim about (3) and (4) plausible was, I submit, precisely the fact that (4) is in the scope of a truth-canceling operator.
Objection 2: There is, of course, an "According to the story" operator in (4). But it does not appear as in (5). Instead, (4) should be read as:
- (According to the story (Colonel Mustard was murdered)) because (according to the story (he knew that Captain Catsup was not as great a tiger hunter in India as he claimed to be)).
Response: This reading is mistaken. Facts of the form of "According to the story p" are explained in terms of the author's authorial activity--they are not facts in the story (generally speaking), in the way p is, but facts about the story. Moreover, consider other cases of this kind of reading. If (6) is to be a paraphrase of (4), the "because" in (6) would have to be the same sort of "because" as in (4). Now, the "because" in (4) is causal--Colonel Mustard's knowledge caused him to be murdered. But the "because" in (6) need not be causal, because from (3) we learn that the author had a colonel murdered because that would sell books, and that could well be a complete causal explanation, with no overdetermination.
Moreover, whether backwards causation is possible or not, it's not easy. But if the "because" in (6) were a standard causal "because", backwards causation would be too easy. For the fact that according to the story Colonel Mustard was murdered could have been made true before the fact that according to the story Colonel Mustard knew about Captain Catsup's poor show in India. Suppose that the author had serialized the novel, and in the first installment printed the fact that Colonel Mustard was murdered, and only later decided that Colonel Mustard knew about Captain Catsup's Indian exploits or lack thereof. Then at the time at which it was already true that according to the story Colonel Mustard was murdered, it was not yet true that according to the story Captain Catsup was found out by the colonel.
But suppose you wanted to explain why Francine was murdered.You would not say that it was fully explained by the free action of Smith, though it was Smith's free action. If God hadn't knowingly placed Smith in circumstances C, he would never have killed Francine. So, what the Author does seems important to explaining what happened in "the story". It is also important to responsibility attributions. Of course, this is a crucial disanalogy to the storybook case, where the author never is given responsibiliyt for the demise of a character (except in Misery, of course).
Some might argue that, regarding authors and their fictional characters, causes are occasionalistic. Character-actions based on intentions revealed in the story are explained not by the characters’ own causal powers but by the author’s. Thus the character Smith’s evil intention to burn down Francine’s house (with Francine inside it) is adequately explained by reference to the author (who creates the heat, as it were), not Smith. Although to the reader it looks as if Smith’s action may be explained by reference only to his free choice, this cannot be true because (as far as the story goes) the claim, ‘Smith could have acted otherwise’ is false. He could have acted otherwise iff the author had decided otherwise, but she didn’t. Nevertheless, we would also want to say (as readers) that while the author is responsible for Francine’s death, Smith is both responsible and blameworthy. But could we properly assert the latter if causes were indeed occasionalistic?
If the analogy holds, it looks as if the system of causes in the actual world is also occasionalistic, given that events (including the outcomes of human intentions) never come about in the absence of divine providence. This troubled St Thomas, who claimed that occasionalism threatened omnipotence. In ‘Disputations’, he wrote:
‘It must be granted without qualification that God operates in all voluntary and natural activity. Through not appreciating the situation accurately, some have made the mistake of attributing all action exclusively to God and denying that natural things perform by their natural powers, as though fire did not heat that God creates heat . . . This, however, is impossible, for it would abolish the order of cause and effect among created things, and this would reflect ill on the power of the creator, for to impart power of action to an effect comes from the very strength of causality.’
And in ‘Contra Gentes’ he adds:
‘The same effect is not ascribed to God and to the natural cause as though each were responsible for a part, for the whole effect proceeds from each, though in different ways.’
Accordingly, creation incorporates both remote and proximate causes. Both divine and creature causes are simultaneously necessary and sufficient for whatever effects are produced. Although, on this account, God brings it about that Judas sins, the sins remain Judas’s and not God’s in virtue of the causal powers God has invested in Judas which he employs in the act of sinning.
Can St Thomas’s account work equally well with authors and their fictional characters? If not, then the analogy proposed cannot hold.
Hugh McCann thinks the story analogy is a good way of explaining Aquinas's account. I am not so sure.
I think that on the story account, we can say that it's true in the story that Smith could have acted otherwise. In fact, the narrator might simply tell us "Smith could have acted otherwise", and then it's true in the story (assuming it's a story with a reliable narrator). I think this is unhelpful, because to help with the free will case, the question shouldn't be whether:
(*) True in the story (Possibly (Smith acts otherwise)).
For then the possibility is just a fictional possibility. Rather, the question is whether:
(**) Possibly (True in the story (Smith acts otherwise)).
Good point. Perhaps, however, the narrator’s stating explicitly (in or of the story) that Smith could have acted otherwise amounts to just this: that there is (a second) possible story in which Smith does in fact act otherwise. But if, in the original story, Smith did act otherwise, then he would no longer feature in that story but in the second. Of course, it would still be true in the original story that the narrator made that statement; but the problem (it seems to me) would then be one of trying to attach any meaning to it.
Alexander, I think that the attraction of the author/novel analogy goes beyond the use of an "in the story* operator.
Suppose I'm writing a story. I have in mind some things I want to happen: Captain Catsup will kill Colonel Mustard. So there must be a motive: let's say Colonel Mustard knows some humiliating truth about Captain Catsup. I've decided that this will involve colonial colonels, so I make the truth something related to their time in India.
But a good novelist cannot simply make the characters into vehicles for the plot. For example, in order for Captain Catsup to feel threatened, he must believe that Colonel Mustard will reveal this truth. So I make Colonel Mustard an honest up-standing character. But I also want people to think that Reverend Green has a motive to murder Colonel Mustard, so I invent an affair between Colonel Mustard and Miss Scarlett (who is really Reverend Green's illegitimate daughter). But how does that fit with Colonel Mustard's honesty?
The author has to control the plot, and so the characters must perform certain tasks. But it must be credible that these characters would act in this way: we must believe that all of these actions spring from the character's being a particular kind of person placed in this particular situation.
So the author is engaged in a balancing act: the plot must be well-crafted and the characters credible. The two requirements are in tension, and we can recognize when an author fails to perform it well. (E.g. I recall reading that E.M.Forster agreed with many readers who felt the affair between Bast and Helen in Howard's End was unrealistic, but it was necessary because he wanted to move the plot in a particular direction).
Then, we naturally think of God as performing perfectly what human authors attempt. The universe has a plot that runs smoothly - providence. But the characters all behave with integrity: the actions they perform come naturally from who they are - freedom.
Incidentally, Rowan Williams discusses this at length in his Dostoyevsky: Language, Faith and Fiction. He argues that there is a close connection between language and freedom. This is because, as a familiar argument runs, if someone claims that my actions are completely predictable, I can ask them to predict my next action. Then I can falsify the prediction. It is my ability to understand language that enables me to exercise freedom. So, argues Williams, the fact that we are language-users proves that our behaviour cannot be completely determined by determinate physical causes. Fictional characters are exist within the realm of language and, as such are free. As I said, Williams explores this theme at length: what he fills out in the rest of the book provides an answer as to what this freedom involves, and why it matters. I had been planning to refer to some specific passages of the book, but really, the whole book is about freedom and the limits of authorship.
Ben,
Thanks for these very helpful suggestions.
But notice that integrity-based explanations can backtrack. If x's character is chosen so as to make action A happen, then x's character does not explain the action, except within the scope of the fictionalizing operator. On the contrary, in that case, the action explains the character, outside the scope of the fictionalizing operator. But character-based compatibilistic freedom requires that the action not only harmonize with the character, but that the character actually enter into the explanation of the action.
I also want to note one discrepancy between fiction and life. There are things that happen in life which we are uncomfortable with in fiction--we would say that an action is out of character, not sufficiently motivated, etc. if it happened in fiction. Now we may posit deep springs of change in the character, and post hoc we may even try to guess what they were. But if we do that, we'll simply be giving a just-so story that we have little real reason to think is true. (In fact, we don't have much reason to think the just-so story is true even if we do assume compatibilism. For even if compatibilism is true, a person might act in a way that is out of character, not sufficiently motivated, etc.--perhaps the action will be unfree then, but it will in any case be possible.) A strained and improbable affair--well, improbable things do happen quite often.
By the way, the prediction argument seems to me to be a poor one. Here's why. Let T(A,B) be the following conditional: "Were I to tell you that you would do A, then you would do B." Complete predictability at most shows that for any pair of actions A and B, I can know whether T(A,B) is true. So suppose that for every pair of actions A and B, I know whether or not T(A,B) is true. (If the choices are indeterministic, this is Molinist knowledge. If the choices are deterministic, this is knowledge of causal or nomic counterfactuals.) Then there is a further question whether there exists any action A such that T(A,A) is true. There may not actually be any such action--predictability guarantees that for every A, I know whether or not T(A,A) is true, but it does not guarantee that there is any A for which T(A,A) is true.
So complete predictability is insufficient to guarantee that I will be able to correctly tell you what you will do, since it is only the case that I can correctly tell you what you will do if there in fact is an A such that T(A,A), and there might not be such an A.
Now, I suppose if complete predictability holds, there might be an A such that T(A,A), and if there is such an A, then I can tell you what you will do, and you will indeed do it. But predictability is a red herring even in that case. For suppose there is an A such that T(A,A). Then whether or not I can predict your actions, it is possible that I tell you that you will do A (I might just guess), and if I tell you that you will do A, then in fact you will do A (because of T(A,A)).
I may be missing something from your quick sketch of the argument.
Alexander, I think that's a pretty decisive argument against the prediction argument. Williams only mentions this argument in an off-hand way, as though he thinks it is something pretty well established. I've always had my doubts about it - I remember debating this with Swinburne one time: he probably has a better defence of it than Williams does.
Just to explain the integrity-based idea a little more. A character,C, is created to perform some action, A. So the A explains C. But then, it turns out that a character who would perform A would also, in the ensuing circumstances, perform B. So then C explains B. Then either, the author must accept a situation in which A performs B, or else change the situation somehow so that the circumstances that lead to C performing B, perhaps by introducing another character.
Authors then often talk about the process as one in which characters start to take on a life of their own, asserting their independence. I suspect that one reason for the attraction of the analogy between author and God lies in the phenomenology of writing a novel.
Ben:
Do you have a reference for Swinburne's version of the prediction argument? I didn't realize the prediction argument was taken seriously in the literature. If it is, I might publish my refutation.
You make authorship sound like an engineering problem. We have different parts, each with its own fairly fixed properties, and various conditionals about how the parts would function... This isn't a criticism of what you say--I think you're right (but it's not freedom).
Alexander, Swinburne discusses the issue of prediction at length in The Evolution of the Soul. However, his argument is not the simple argument from predictability that I presented above. His conclusion is
"I conclude that the phenomenon of human counter-suggestibility is strong evidence of indeterminism in the production of purposes and so intentional actions, because it would only be compatible with determinism on the assumption of an a priori very unlikely mechanism for the production of beliefs."
(p. 257)
So my previous post about him was somewhat misleading. I remember discussing this issue with him, but it must have come up as a digression when talking about some other topic, so even at the time, I had only a vague impression. When I read Williams on the topic, I was reminded of the conversation with Swinburne, and wondered whether Williams might have had Swinburne's argument in mind: but that is unlikely. Williams seems to think that counter-suggestibility (our ability to falsify predictions about our future behaviour) rules out determinism. Swinburne's argument is that it is not probable that a deterministic system would result in so many cases where I cannot reveal to you what your next action will be.