Introduction
In Alvin Plantinga's (Reference Plantinga and Beilby2002a, 11) evolutionary argument against naturalism (EAAN), he contends that one who holds both naturalism (N) and evolution (E) acquires an undefeated defeater for her belief that her cognitive faculties are reliable (let R be the proposition ‘human cognitive faculties are reliable’) and as a result an undefeated defeater for everything else she believes, including N and E. This is because we have good reason to think that the probability of R is low or inscrutable given N and E, i.e. P(R/N&E) is low or inscrutable. This argument, if successful, shows that naturalism is self-defeating or irrational. The purpose of this paper is to show that EAAN does not work if P(R/N&E) is inscrutable, because someone who concludes that P(R/N&E) is inscrutable may or may not have an undefeated defeater for her belief that R. Besides, one does not acquire a defeater for all one's other beliefs even if one has an undefeated defeater for R.
An undefeated defeater for R?
I shall first argue that, when a naturalist comes to believe that P(R/N&E) is inscrutable, she does not acquire any defeater for R at all, unless the cited probability is her only source of believing that R.
We have to first understand why Plantinga thinks that someone who comes to realize that P(R/N&E) is inscrutable will get a defeater for R. He assumes the following relationship between the rationality of belief and the probability that a certain proposition is true. If after some reflection or investigation, I conclude that I cannot make an assessment of the probability of a proposition X, the probability of X will be inscrutable to me. In that case, I will also not be justified to believe X or ~X. Therefore, my judgement that the probability of X is inscrutable constitutes a defeater for my belief that X, because that judgement requires me to withhold my belief in X. Assuming this principle, Plantinga claims that someone who comes to realize that P(R/N&E) is inscrutable will get a defeater for R if she also accepts naturalism and evolution.
However, it is not the case that someone who comes to realize that the probability of any proposition X on a certain evidence E is inscrutable will necessarily get a defeater for X. Whether you get a defeater for X depends on whether you have other grounds for believing X other than the judgement of the probability of X on evidence E. Let say your other grounds for believing X is evidence O. So the probability of X on evidence O is more than 0.5. When you come to realize that the probability of X on evidence E is inscrutable, i.e. when you think that you do not know what the probability of X on evidence E is, that will not affect your assessment that the probability of X on evidence O is more than 0.5. Thus, your ignorance about the probability of X on evidence E does not constitute a defeater for your belief that X.
Perhaps I can illustrate the idea with a concrete example. Suppose an engineer is given the task of testing a new device to see whether it is reliable. She subjects the device to a comprehensive set of tests that examine the performance and reliability of the device from many different angles. Let's call the set of tests, Test A. The result of Test A shows that the probability that the device is reliable is 0.9. Suppose a few months later, some engineers in Japan introduce a new test for this device. Let's call it Test B. But Test B is different from Test A in that it is not a set of tests. It is a single test that examines only one small aspect of the device's performance. But after subjecting the device to Test B, the engineer finds that the probability of the reliability of the device according to Test B is inscrutable. In this situation, her agnosticism about the probability of the reliability of the device on Test B does not entail or require agnosticism about the reliability of the device, i.e. it does not require her to withhold her belief that the device is reliable, because she has already had a good reason to believe that the device is reliable, that is, the probability that the device is reliable on Test A is 0.9. Those are her original grounds for believing that the device is reliable, and the result of Test B has no adverse effect on those grounds.
The situation will be different, if, let us say, a few months later, the company that produced the original Test A introduces an updated version of Test A – and they name it Test A Vista. The Company claims that Test A Vista examines all the areas examined by the older Test A with greater accuracy and is thus supposed to replace Test A. If the engineer tests the device with Test A Vista and finds that the probability that the device is reliable is inscrutable, she will have to withhold her belief that the device is reliable, because in this case, her original grounds for believing in the reliability of the device have already been cast in doubt. Why is that? Her original grounds are the finding that the probability that the device is reliable on Test A is 0.9. But that finding has already been cast in doubt when a more accurate later version of the test shows that the probability of the reliability of the device is inscrutable. However, Test B is different from Test A Vista. There is no claim that Test B is a more accurate and updated version of Test A. In fact, Test B examines only one small aspect of the device's performance, whilst Test A is much more comprehensive in scope.
I hope this better illustrates my earlier claim that when one comes to realize that the probability of X given evidence E is inscrutable, it does not imply that she will necessarily get a defeater for her belief that X. When a naturalist comes to realize that P(R/N&E) is inscrutable, she may or may not get a defeater for her belief that R, because she may have other grounds for believing R. Richard Swinburne makes a similar observation about the inscrutable probability thesis:Footnote 1
If he affirms that the probability is inscrutable, he can perhaps say that, given only N and E, we have no reason to believe that R; but not that R is probably false. But the former allows the possibility that we have, beside N and E, plenty of other reason for believing R. Plantinga's argument has significant force only against naturalism … if we take him to be affirming that the probability of R, given N and E, is low … . (2004, 350, n. 3).
When someone comes to believe that P(R/N&E) is inscrutable, she does not acquire any defeater at all, unless P(R/N&E) is her only grounds for believing that R. What are other possible grounds for her belief that R? I can think of two examples. She may have R as a basic belief, or she may believe R on the basis of inductive reasoning. In either case she does not ground her belief that R on the assessment of P(R/N&E). So when she comes to believe that P(R/N&E) is inscrutable, this belief is neither a rebutting nor an undercutting defeater for her belief R. Suppose I believe in the existence of God on the basis of sensus divinitatis, i.e. I take the belief that God exists as properly basic and not on the grounds of evidence. I later examine a set of theistic arguments and come to the conclusion that, on the basis of these arguments, the probability of God's existence is inscrutable. This does not constitute a defeater for my belief that God exists unless my only grounds for that belief are the assessment of that probability on the basis of examining the set of theistic arguments. But that is clearly not the case since my belief in the existence of God is based on sensus divinitatis, and an agnostic attitude towards the probability of that set of theistic arguments does not constitute a defeater for that belief.
Let's consider the case of someone who believes R on the basis of some form of inductive reasoning. Suppose John is a naturalist and has thought about the problem of induction. Suppose he has read James van Cleve (Reference Van Cleve1984) and is convinced that according to the reliabilist solution to the problem of induction, we can have inductive knowledge as well as reason to think that induction is reliable. He then comes to believe and defend R by way of an inductive reasoning. Let's suppose further that he later comes to read Plantinga's EAAN and accepts that P(R/N&E) is inscrutable (but not low). Does it follow that at that point he can no longer believe R? It seems not, because he had already believed R and had a good reason for that belief before he came to accept that P(R/N&E) was inscrutable.
Perhaps some may think that the reliabilist solution to the problem of induction is hopeless, as I'm inclined to. Well, we can modify our example a little bit. Suppose John is convinced that the rationalist solution to the problem of induction works. So he believes that we can show that induction is a truth-conducive method of inference by arguing that the truth-conduciveness of the criteria used in induction is an a priori truth. That is how a rationalist would argue for the truth-conduciveness of induction (see Bonjour (Reference Bonjour1998) ; Swinburne (Reference Swinburne2001), ch. 4). After seeing the truth-conduciveness of induction, he then comes to believe and argue for R by way of an inductive reasoning. Let's suppose further that he later comes across Plantinga's evolutionary argument against naturalism and accepts that P(R/N&E) is inscrutable (but not low). Does it follow that at that point he can no longer believe R? The answer is negative, because, as in the reliabilist case above, he had already believed R and had a good reason for that belief before he came to accept that P(R/N&E) was inscrutable. Unlike the case of the low probability thesis, the acceptance of the inscrutable probability thesis does not affect his original belief and reason for that belief. Thus, there is a good reason to think that agnosticism about P(R/N&E) does not entail agnosticism about R, because one may have R as a basic belief, or have other reasons for holding R. Put differently, one who comes to think that P(R/N&E) is inscrutable does not acquire a defeater for her belief R if her initial grounds of believing that R are something other than the assessment of P(R/N&E).
A defeater for every other belief?
Even if we grant that the agnostic about P(R/N&E) does acquire an undefeated (undercutting) defeater for R and thus has to withhold her belief R, it still does not follow that she has a defeater for all her other beliefs (formed by those faculties).
This can be illustrated with the example of our belief in the reliability of induction and how that affects our inductively formed beliefs (that is, beliefs formed by induction). It is true that if one comes to think that the probability of the reliability of induction is low, that is, less than 0.5, he has a defeater for all his beliefs formed by induction and thus has to withhold them. That's one way of interpreting someone's ‘coming to doubt the reliability of induction’. But there is another way of interpreting it: that is, coming to the realization that there is neither reason to think that induction is reliable nor reason to think that it is unreliable. I think the latter interpretation is more in line with Hume's inductive scepticism. What he claims and argues is not that induction is unreliable, but that we have no reason to think that it is reliable. And one who has neither reason to think that induction is reliable nor reason to think that it is unreliable should be agnostic about its reliability. So someone who comes to think about the problem of induction by reading Hume's Enquiry Concerning Human Understanding will not have a defeater for his beliefs formed by induction if he understands Hume correctly, and can still have rational or even warranted inductive beliefs.
In short, if we acquire reason to think that we should be agnostic about the reliability of the source of our belief, and do indeed take that agnostic view about the reliability of the source, it does not follow that we have a defeater for that or any belief produced by that source. And I think both externalist and some versions of internalist epistemologies support this position. According to externalist epistemologies such as reliabilism, our belief can still be warranted even if we do not have any belief about the reliability of the source of that belief, because a belief is warranted as long as it is produced by a reliable belief-forming mechanism and as long as we do not have reason to think that its source is unreliable. The fact that one is agnostic about the reliability of the source of that belief does not prevent one from being rational or warranted in holding that belief, because knowledge of the reliability of the source is not necessary for warrant and agnosticism about the reliability of the source is just the same as having no belief and therefore no knowledge about the reliability of the source.
Even though internalists typically require that one's justification for holding a belief must be internally accessible to the subject who holds that belief, internalism by itself is not committed to the claim that one must always have a justification for holding a belief in terms of another belief. In fact, this claim will get you into an infinite regress. For example, foundationalist internalism does not and cannot require that someone can only be rational or justified in holding a belief if he is justified or rational in believing that his cognitive faculties are reliable. Because if it requires this, then it will also require that he is rational in believing that his cognitive faculties are reliable only if he is rational in believing the reliability of the source that produced her belief that her cognitive faculties were reliable. And this will get into an infinite regress.
This point – the contention that even if we grant that the naturalist has an undefeated (undercutting) defeater for R, it does not follow that she also has a defeater for everything else she believes – has been made by Ernest Sosa (Reference Sosa2002, 98–101) and James van Cleve (Reference Van CleveFootnote 2002, 121–123). Plantinga's (Reference Plantinga2002b, 238–247) replies to them do address some of the objections contained in this contention. But I shall argue in the following section that Plantinga cannot successfully deflect the contention that it does not follow that the naturalist who has an undefeated (undercutting) defeater for R also has a defeater for everything else she believes, if we focus only on the inscrutable probability thesis.
In a personal communication, Plantinga agrees that agnosticism about R does not always prevent one from rationally holding beliefs produced by her cognitive faculties. He attempts to resist my second thesis by making a distinction between agnosticism about R before one raises the question of the reliability of her cognitive faculties and agnosticism about R after raising that question. His point is that, if you are agnostic about R but you have never thought about the question of the reliability of your cognitive faculties, then you do not have a defeater for all your other beliefs produced by those faculties. But if you have raised the question of the reliability of your cognitive faculties, and after considering that question you come to take an agnostic attitude towards the reliability of your cognitive faculties, then you acquire a defeater for all your other beliefs produced by those faculties. This is a very interesting suggestion because it has an element of epistemic contextualism, which is roughly the view that whether we can say that someone knows a proposition or is justified in believing the proposition depends on some contextual factors (see Lewis (Reference Lewis1996) ).
Agnosticism about R before and after considering the question of R
Plantinga's key contention is that there is a principled distinction between agnosticism about R before one raises the question of the reliability of her cognitive faculties and agnosticism about R after raising that question, and that the former does not prevent one from rationally holding or having warrant for the beliefs produced by her cognitive faculties, whereas the latter furnishes her with an undercutting defeater for every other belief produced by her cognitive faculties.Footnote 2 In other words, if someone fails to believe R after considering the question of R, she will acquire an undercutting defeater for every other belief produced by her cognitive faculties.Footnote 3
If this claim stands, it seems that my second thesis will be threatened. But it is far from clear that the distinction is epistemically significant vis-à-vis the question of whether we can rationally hold every other belief produced by our cognitive faculties. Suppose at some time t 0 Ron holds some inductively formed beliefs but has no belief about the reliability of induction, because he has not thought about that question. According to Plantinga's distinction, Ron can rationally hold these inductively formed beliefs at t 0. His lack of belief at the time does not prevent him from having rational inductively formed beliefs. Suppose further that later at t 1 Ron reads about the problem of induction and begins to question whether induction is reliable. After some deliberation he concludes that there is neither reason for thinking that induction is reliable nor reason for thinking that it is unreliable, and thus holds an agnostic attitude towards the reliability of induction (i.e. neither believes it nor its denial).
According to Plantinga's distinction, Ron's lack of belief that induction is reliable at t 1 constitutes a defeater for his inductively formed beliefs. But what exactly has changed between t 0 and t 1 that makes us think that Ron's lack of belief about the reliability of induction at t 0 does not prevent him from rationally holding his inductively formed beliefs, while his lack of belief that induction is reliable at t 1 makes him irrational in holding to those inductively formed beliefs? Put differently, if Ron can rationally hold his inductively formed beliefs at t 0 when he has no belief about the reliability of induction, why assume that he cannot rationally hold these beliefs at t 1 when he has no belief about the reliability of induction, just because he has considered the question of the reliability of induction?
Similarly, if at both t 0 and t 1 Ron has no belief about R, and if he can rationally hold the beliefs produced by his cognitive faculties at t 0, there seems to be no reason to think that the mere fact of having considered the question of R at t 1 requires that he must from that point onwards believe R if he wants to remain rational in holding those beliefs. There is no reason to think that the mere fact of having considered the question of R is epistemically significant to whether one can rationally hold the beliefs produced by his cognitive faculties, unless one assumes that some sort of implicit belief about R must exist at t 0 even when Ron has not thought about the question of R and thus doesn't have any explicit belief about R. This requirement of implicit belief is implied in Plantinga's answer to Sosa:
No doubt a child or for that matter a full-grown adult can rationally believe much without believing R … perhaps she doesn't so much have the concept of reliability. (Even so, though, if she does have that concept, doesn't she do something like assume R? Or perhaps implicitly believe it, just as I implicitly believed, a moment ago, that 348+1=349, even though I had never entertained that thought?) (Plantinga (Reference Plantinga2002b), 243).
This indicates that Plantinga seems to think that the fact that Ron has not thought about the question of R at t 0 does not prevent him from having an implicit belief that R at t 0, which is necessary for him rationally to hold the beliefs produced by his cognitive faculties. Perhaps this is what makes the crucial difference between t 0 and t 1, when the question of R has been considered: having considered the question of R and having decided that he can only be agnostic about it, Ron no longer has and can't have any belief about R, whether explicitly or implicitly, whereas at t 0 he implicitly believes that R.
This assumption provides a prima facie case for the claim that the distinction is epistemically significant and is the only way I can think of that could possibly make the distinction epistemically significant.Footnote 4 Unfortunately, there are two problems with the assumption that one has to (at least implicitly) believe R if she is rationally to believe anything produced by her cognitive faculties. First of all, as pointed out by his critic, Van Cleve, this assumption is quite out of keeping with Plantinga's externalist epistemology. This is because typical externalist theory of epistemic warrant does not require that, if you are to be warranted in holding a belief, you must be aware of the grounds for that belief. As long as the belief is produced by a reliable process, one can be warranted in holding it, whether or not one is aware of the reliability of the process which produced that belief. Unless Plantinga can try to fit this assumption consistently into his epistemology, I think that he is not entitled to make this assumption, not to mention using it to resist my second thesis.
The second problem I see is that, even if we grant that people do implicitly believe the reliability of their cognitive faculties whenever they form any beliefs, that still does not show that this implicit belief has any bearing on one's rationality in holding one's beliefs formed by one's cognitive faculties. This is because two problems will arise if we insist that this implicit belief has bearing on one's rationality in holding his beliefs formed by his cognitive faculties. Firstly, if an implicit belief in the reliability of someone's cognitive faculties is necessary for him to be rational in holding any belief, then even his implicit belief in the reliability of his cognitive faculties can only be rationally held if he has a further implicit belief in the reliability of his cognitive faculties, because the first implicit belief is also a product of his cognitive faculties. But this generates an infinite regress and is but an illustration of the old problem of the structure of epistemic justification – how far should the chain of epistemic support go?
It might be argued in response that if the requirement is that only explicit beliefs require an implicit belief in the reliability of cognitive faculties for those beliefs to be rationally held, the infinite regress is avoided. But unfortunately, this takes us into the second problem. If what you mean by implicit belief is a belief that you are not aware of – an unconscious belief – my question is: how could that implicit belief that you're not aware of have any bearing whatsoever on your rationality of holding your explicit beliefs? The matter becomes a bit complicated here because there are two kinds of rationality – objective rationality and subjective rationality. If we are looking at subjective rationality from the first-person perspective of the subject whose rationality we are assessing, the subject's unconscious belief can have no bearing on his rationality in holding his explicit beliefs. Of course, if we are looking at the objective rationality from the third-person perspective of someone who attributes rationality to a subject, the subject's unconscious belief can have bearing on his rationality in holding his explicit beliefs. However, if objective rationality is in view here, then we will get back to the first problem of getting into the infinite regress, because a person's implicit or unconscious belief does have bearing on the rationality of his holding other beliefs. Therefore, you can no longer say that only explicit beliefs require an implicit belief in the reliability of cognitive faculties for those beliefs to be rationally held. So you either get trapped in problem 1 (i.e. the problem of infinite regress) or problem 2 (i.e. one's unconscious belief has no bearing on the (subjective) rationality of one's other beliefs).
Conclusion
Let me note, in closing, that it seems to me that to salvage EAAN, it would have to be held that P(R/N&E) is low. In fact, Plantinga does offer good arguments for the claim that P(R/N&E) is low, i.e. less than 0.5. But he does not think that we can insist on the low probability claim:
Then P(R/N&E) will be at most .45, less than ½. Of course we could be easily be wrong; the argument for a low estimate of P(E/N&E) is by no means irresistible; our estimates of the various probabilities involved in estimating P(R/N&E) with respect to that hypothetical population were (naturally enough) both imprecise and poorly grounded. You might reasonably hold, therefore, that the right course here is simple agnosticism: one just doesn't know what P(R/N&E) is. You doubt that it is very high; but you aren't prepared to say that it is low: you have no definite opinion at all as to what that probability might be. Then this probability is inscrutable for you. (Plantinga (Reference Plantinga and Beilby2002a), 10)
However, if at least one of my theses in this paper is correct, it seems that there is no choice but to insist on the claim that P(R/N&E) is low, making it irresistible (or at least harder to resist) by providing a more precise and well grounded estimation of how low that probability is. And it seems to me that there is no reason to think that this is unachievable. It is not that I deny the central claim of EAAN that naturalism is self-defeating. In fact, I am inclined to think that P(R/N&E) is low, but I just can't see how this (the central claim) can be defended when we allow that P(R/N&E) is either low or inscrutable rather than just low, and when we think that we can be easily wrong about the low estimates. In short, Plantinga's argument against naturalism is unsuccessful in so far as it is formulated as a disjunction of the low probability thesis and the inscrutable probability thesis. Given that the problem lies in the inscrutable probability thesis, EAAN can be salvaged by removing the thesis as a disjunct and thereby formulating the argument in terms of only the low probability thesis.
The inscrutability probability thesis in EAAN is the peculiar focus of this paper, and is where its original contribution lies. Of course, the focus on the inscrutability probability thesis in locating the weakness of EAAN in general cannot be claimed to be my original contribution, given that Swinburne has already independently pointed out that it is not the case that the awareness of the inscrutability of P(R/N&E) implies the acquisition of an undefeated defeater for R (see ‘An undefeated defeater for R?’ above). Rather, the main original contribution of this paper lies in my focus on the inscrutable probability thesis in EAAN in my defence of the second thesis.
By focusing on the inscrutability probability thesis, I have strengthened Van Cleve's and Sosa's objections to Plantinga's contention that one's realization of the low or inscrutable value of P(R/N&E) constitutes a defeater of one's every belief.Footnote 5 Plantinga's (Reference Plantinga2002b) replies to their objections seem to have force because he uses examples in which someone comes to think that P(R/N&E) is low in his replies. However, Plantinga's replies are unsuccessful when we single out the inscrutability probability thesis, and this is a point that neither he himself nor Van Cleve and Sosa notice. Had Van Cleve and Sosa singled out the inscrutability probability thesis in formulating their objections, the sort of replies provided by Plantinga (2002b) would have been pre-empted or could have included some of the moves considered in my discussion of the second thesis (e.g. the distinction between agnosticism before raising the question of R and agnosticism afterwards).Footnote 6
