I recently attended a talk by philosopher Stephen Law on fine-tuning and the Evil God hypothesis. What follows is an e-mail I sent afterwards to Dr Law, making some responses to his criticisms of the fine-tuning argument. I don’t have the time to turn this into a proper article, but I hope it will still be useful. Do feel free to skip to the section on Evil God if that is your main interest.
Dear Dr Law,
Thanks again for your talk at Oxford Brookes last night – very much enjoyed it! This might turn into quite a long e-mail, but I hope some of it is useful for your reflection, and I’ll try to make it as concise as possible. I had some possible responses for everything except the 3rd concern: I haven’t read a huge amount on metaphysics or the internal coherence of theism, so don’t have much to say on that 3rd concern that you won’t have already heard.
1st concern: Stenger et al, and multiverse
This might be an area for caution, but I think the fine-tuning advocate still has the support of the majority of the scientific community (including atheists), so I think it probably ought to be granted. I know you mentioned some of Stenger’s own calculations, but I think those are pretty problematic. Collins and Barnes, in particular, have offered a fair bit in response to Stenger’s work in this area, so (for example) Collins notes that whether stars exist seems to be the only factor Stenger considers to be necessary for life in his model. He doesn’t seem to deal much with other possible life-inhibiting factors (such as the lack of atoms with greater atomic number than 1), and doesn’t seem to take into account many of the fine-tuned constants (e.g the strong nuclear force), or even the ostensibly fine-tuned initial conditions (pertaining to entropy, and so on). Collins also has some particular queries about the simplicity of Stenger’s equation:
“the equation he uses is based on a simple star model of stellar evolution. The equation does not take into account the complexities of a stellar evolution, such as whether the energy transport from the center of the star to the surface is by convection or radiative diffusion. More importantly, it assumes that the star is made mostly of hydrogen, which would not be the case if the strong force were increased beyond a small amount; further, it does not take into account the effects on star stability of quantum degeneracy, which require much more sophisticated codes to take into account.”
The relevant physics is pretty beyond me, but I think the pertinent points are that objectors to fine tuning are in the minority, and that the fact of fine-tuning has the support of not only most, but also the most prominent, physicists. Stenger’s models have been critiqued from within the physics community, and I don’t think his concerns (in this regard, at least), ought to be regarded as too damaging to the fine-tuning case.
I’ve actually written a paper on the multiverse which I’m hoping will be published, and I expect it’d be too long to recount now! In informal terms, though, I think it ought to be noted that, so long as the fine-tuning argument is only construed as an inductive (or perhaps Bayesian/C-inductive) argument to the conclusion that some fact related to fine-tuning constitutes evidence for theism, the mere possibility of a multiverse will not do much to damage the case. Other possible explanations can be advanced for any observed datum in experimental science, but obviously that doesn’t prevent us from concluding that the datum constitutes evidence for some particular hypothesis! I think this point can be demonstrated formally, too (you can skip this section if you’re particularly uninterested in probability theory!): the relevant probabilities for deciding whether some fine-tuning related fact constitutes evidence for theism are P(EPU|T & k) and P(EPU|¬T & k), where EPU =df there exists an embodied moral agent-permitting universe, T =df theism is true and k =df other background information (but, importantly, not everything we currently happen to know). Then,
P(EPU|¬T & k) = [P(EPU|M & ¬T & k)·P(M|¬T & k)] + [P(EPU|¬M & ¬T & k)·P(¬M|¬T & k)]
Where M =df there exists a multiverse. Now, supposing we grant that P(EPU|¬M & ¬T & k) is negligible (for the sake of argument – supposing that the multiverse is the only objection we have), P(EPU|¬T & k) will approximate to P(EPU|M & ¬T & k)·P(M|¬T & k). But then it is clear that the multiverse is only relevant to the argument insofar as P(M|¬T & k) and P(EPU|M & ¬T & k) are both reasonably high. But clearly most people will be reluctant to say that a multiverse probably exists, conditioned only on atheism and some other background information: after all, there seems to be no independent evidence for it with which we can raise P(M|¬T & k). So I don’t think the theist ought to think that the multiverse does damage to a C-inductive argument from fine-tuning to theism. Much more can be written about this (and I can send my paper if you like), but I’ll spare you for the time being!
2nd concern: Kinds of probability
As I intimated in the debate, I found this a bit misleading. Clearly theists (mostly) mean to use epistemic probability in their formulations, since physical probability is far more inscrutable when discussing kinds of universes, and since many fine-tuning proponents have explicitly outlined accounts of epistemic probability in advancing the argument (for a section on epistemic probability in this context, see e.g. Collins, “The Teleological Argument” in Craig & Moreland eds. The Blackwell Companion to Natural Theology. 226-239). So the most that could be said, I think, is that they ought not use epistemic probability.
Now, obviously you did not have time in your talk to go into the different accounts of epistemic probability, but you did make one objection to the use of epistemic probability which I don’t think the fine-tuning proponent (or target) ought to find convincing, viz. that the epistemic probability of my (or any other embodied moral agent’s) existence is 1. I don’t think this is true. I ( and any sophisticated fine-tuning proponent) would agree that P(EPU|my experiences) = 1, but the fine-tuning proponent is not arguing that P(EPU|my experiences) << 1, so that particular conditional is quite irrelevant to the argument. Rather, she is arguing that P(EPU|¬T & k) is low, where k is some background information which, importantly, does not include the fact of our existence. There seems to me to be nothing in the probability calculus itself which precludes us from coming up with conditional probabilities where some known data are not included in the conditional, so I think any objection (e.g. Sober, 2003) which insists that all known information is included in the conditional is palpably unjustified – especially as it would lead to some quite counter-intuitive and counter-scientific consequences! Indeed, the eminent (atheist) Colin Howson thinks such “old evidence” objections to Bayesian arguments are so bad that the section of his book responding to them begins: “We have tried to deal with the objections which we feel merit serious discussion. We shall end, however, with one that doesn’t.” He then goes on to explain how one can (and must, if one is to pursue science rationally) provide an axiomatisation of the conditional which removes certain known data from the conditional, to come up with these other epistemic probabilities. But the summarised point is that the epistemic probability that I exist is only 1 conditioned on certain facts, but these facts are (justifiably) excluded from the conditionals in Bayesian fine-tuning arguments, so I do not think this will be a successful response to the argument.
4th concern: Doesn’t prove one god in particular
Again, I think this only ought to be a concern if it is construed deductively. If the fine-tuning argument is framed in Bayesian terms, then the objection seems trivial: of course it doesn’t prove one god in particular, but it does constitute evidence for theism, and that is what it was intended to do. Of course, a related objection is that the argument does not prove the God of a particular religion, even if it demonstrates theism to be true. But this seems to me to be is a complete red herring. Take Christianity as an example. For Christianity to be true, its essential doctrines ought to be true, theism being one of these. The question of “the Christian God’s” existence presumably depends on some essential properties of the “Christian God” which are dissimilar to other conceptions of God. But Christianity has traditionally emphasised what God has done in history, and used this ostensible action as the basis for other theological doctrines. So let us concede for the sake of argument that the question of whether the Christian God exists is a meaningful one, separate from the question of whether God exists. Such a being’s essential properties will then presumably include, for example, the property, “became incarnate in Jesus, and raised Jesus from the dead”. In this case, it will be clear how the kalam cosmological argument, if sound, contributes to the Christian case. This is because the probability of God becoming incarnate in Jesus, and raising Jesus from the dead, is far greater (indeed, infinitely greater) given theism than given atheism. Thus, any arguments providing strong evidence for theism generally are likely to make Christianity significantly more probable than previously. Whatever one’s assessment of Christianity and of whether belief in these doctrines is warranted, then, they should nevertheless agree that this objection – that the fine-tuning argument, even if sound, cannot strengthen a particular religion’s case – is clearly mistaken.
5th concern: Evil God
I find this quite an intriguing suggestion, and indeed agree that almost all theists (including perhaps the majority of theistic philosophers) would be unequipped to meet this challenge. I don’t find moral or ontological arguments convincing (though I think they can be made more rigorous than is often assumed, such as Maydole’s modal perfection argument from 2003, with which I only really take issue with the use of the Barcan Formula and perhaps one of the premises). I think there are plausibly asymmetries in arguments from miracles and arguments from (first person) religious experience (if one accepts, for example, the principle of credulity). For example, one might reasonably suppose that, if God existed, he would make certain facts pertaining to the resurrection obtaining, whereas this is not so clear (and, at any rate, it seems far less simple) on Evil God. But perhaps this can be brought out further by considering what I think is a far more powerful objection to the Evil God challenge: simplicity.
Of course, the Evil God challenge is an instance of the classic problem of under-determination and empirical equivalence – so we have to be clear that this is no problem unique to theism: indeed, it impinges on every scientific theory. And it seems to me that the primary criterion involved in deciding between empirically equivalent theories is something like simplicity, or prima facie believability, or something like that. And clearly one could come up with all sorts of sceptical hypotheses which are empirically equivalent to the ordinary scientific account of the world: that the universe was created 5 minutes ago with the appearance of age, for example, or Putnam’s brain in a vat scenario, or some Cartesian demonic hypothesis. And it seems to me that the overriding criterion for theory appraisal (or, for Bayesians, intrinsic probability) in empirically equivalent theories is simplicity or believability. Indeed, Graham Oppy, with whom I am sure you are familiar, insists that “theists will not (and indeed ought not to) concede that the case for an orthodoxly conceived monotheistic god can be paralleled by the case for [evil gods]” and that “theists reasonably judge that the cases are not parallel” (Arguing About Gods, 2006),on grounds of simplicity/believability alone. (NB: Oppy also thinks that an atheist can reasonably accept the symmetry thesis, so he makes clear an important distinction in uses of these challenges: “there are two quite different contexts in which arguments involving [evil god] appear – and different considerations must be appealed to in the assessment of these arguments in these different contexts … one suggestion is that there is an argument against belief in an orthodoxly conceived monotheistic god … another suggestion is that there is a reply to theistic arguments for belief in an orthodoxly conceived monotheistic god that can be based upon [evil god].” Oppy thinks the challenge succeeds in the latter task, but fails in the former task: I think it fails in both.) I can think of no good reason why the standard scientific, naïve realist account of the world should be thought of as simpler than the aforementioned sceptical hypotheses, without maintaining that theism is also simpler than the evil god hypothesis.
But I think some more analytic considerations can be taken into account, to explicate theism’s simplicity more fully. The reasonable theist, I think, might do well to appeal to some particular concept of God to appraise through Bayesian confirmation, and I think the most promising line of reasoning is to appeal to some Swinburnian considerations re: what is involved in personal explanation, and what, precisely, the theist is positing as an explanation. Here, while I think Daniels’ account is mistaken, I also think that there is something to it, namely a kind of complexity in the idea of a being knowing all moral truths (and therefore all moral “goods”) and knowingly acting contrarily to the good, which isn’t symmetrically present in theism. I don’t think it is incoherent, but I think it has an added complexity. Now, when further Swinburnian considerations are added, such as the idea of God as essentially perfectly free, I think the case is strengthened manifold. The point here is that, according to the theistic explanation, personal explanation consists of explaining some datum/data by positing an agent with intentions and so on, who generally act for some good. According to the theistic account, the only reason persons do not act for the good is in case of ignorance of the good or because of some irrational, external influence (such as some malignant desire). But clearly an omniscient being (such as Evil God) would not act against the good in the first case, since he would know all moral truths. And since God is defined as perfectly free, that is, not being subject to irrational, external influences, it seems to follow directly, according to the account of personal explanation, that God would be all-good. Now, in the case of Evil God, one would have an omniscient being who is not perfectly free (since they could otherwise not do evil), and who was thus subject to irrational, external influences. But the idea of an omnipotent, omniscient, eternal being who was nevertheles, not only partially affected by, but completely engulfed by and determined by irrational, external influences, seems to have an enormous amount of prima facieunbelievability and complexity, even if not incoherence. So it seems plain to me that the traditional theistic account is simpler in that respect.
Additionally, if one finds some of the arguments for a free cause of the universe to be remotely persuasive, this would make a perfectly free being not only simpler, but also more explanatorily powerful than Evil God. For example, the kalam cosmological argument and some arguments from contingency might argue that freedom (or some other kind of indeterminacy) is the only thing which can account for the existence of the world – in the first case because it is alleged that only a free cause could exist “eternally” without giving rise to an “eternal” effect, and in the second case because freedom might seem to be the only way that contingency can arise from a necessary being (with some argument for a necessary being independently adduced, though I tend to find arguments for a necessary being persuasive, even if we cannot get much of a grasp on what the necessary being is – Oppy, for example, is inclined to think that the Big Bang singularity was necessary).
I’m also inclined to think that there might be some kind of argument relating to meta-ethics and modality, particularly if one was inclined to think that moral truths were necessary. Since I don’t know your position on whether moral truths are necessary, and since I haven’t thought about such an argument in much detail, I’ll leave that for now, but I’d be very interested to hear whether you think moral truths are necessary or what they are grounded in, if you wouldn’t mind?
Of course, you do concede in your paper that theism seems simpler, and you even give an argument independent from any of these to that conclusion. But, as I noted after the talk, your only response to this is to say that simplicity is not a big factor when both hypotheses are strongly disconfirmed by the evidence. But this assumes, of course, that both hypotheses are strongly disconfirmed. It seems to me perfectly plausible that one might think that theodicies and reverse theodicies (or defences à la Plantinga, or assurances à la M. Adams) might all be successful in rejecting arguments from evil or from good, and that theism is simpler than Evil God, and thus that there is very good reason to prefer theism. Without giving a fuller account of the problem of evil (and of good), then, I don’t think anyone who already rejects the problem of evil ought to be troubled by Evil God. I attach a spreadsheet allowing for a formalisation of this: T is theism, EG is the Evil God hypothesis, and E is our observational evidence about the world. The problems of evil and good seem to suggest that P(E|T) and P(E|EG) are both very low with respect to P(E|¬T & ¬EG). If this is indeed the case then, indeed, the posterior probabilities of T and EG are likely to be low (this can be seen by filling in values for those three probabilities, on the left hand side). But supposing one considers the theodicies/defences/assurances persuasive against the problem of evil, or even that one considers evil to have evidential power against T and EG, but nevertheless thinks that other observational evidence overcomes this improbability and provides even better evidence for T and EG. In that case, simplicity becomes important (as can be noted by making P(E|EG) ≈ P(E|T) ≈ P(E|¬T & ¬EG), and making P(EG) and P(T) discrepant on grounds of simplicity). Note that this does not even require the success of theodicies, depending on how strong other evidence for T and EG is (and what the priors of T and EG are). But clearly any reasonable theist is going to a) reject the problem of evil and consider it to have no or little evidential force against theism, or b) accept the problem of evil, but think that there are other evidences for theism, such that the totality of the available evidence acts somewhat in theism’s favour. In that case, simplicity will be perfectly adequate to reject the symmetry thesis. And so, I think, in order to persuade the theist that their position is problematic, one has to show that either a) theism is not simpler (but this is implausible), or b) the problem of evil works and is significantly stronger than all the evidence for theism. This is a much bigger task. That is not to say that it cannot be done, but I think the theist ought to feel comfortable in their position until these tasks are satisfactorily completed. (NB: There are also relevant considerations pertaining to priors here: it may be the case that the prior for theism is so high that even if the problem of evil works and is much stronger than the positive evidence for theism, theism is still overall probable. I don’t think many people would be comfortable with such a high prior for theism (including most theists), but it is worth mentioning purely for completeness).
Much more that could be said, of course, but I’ve taken enough of your time for now! Look forward to hearing your thoughts.