(Prepublication draft of paper published in Int. J. Phil Religion (78) 2015)
THE PANDORA'S BOX OBJECTION TO SKEPTICAL THEISM
ABSTRACT: Skeptical theism is a leading response to the evidential argument from evil against the existence of God. Skeptical theists attempt to block the inference from the existence of inscrutable evils (evil for which we can think of no God-justifying reason) to gratuitous evils (evils for which there is no God justifying reason) by insisting that given our cognitive limitations, it wouldn't be surprising if there were God-justifying reasons we can't think of. A well-known objection to skeptical theism is that it opens up a skeptical Pandora’s box, generating implausibly wide-ranging forms of skepticism, including skepticism about the external world and past. This paper looks at several responses to this Pandora's box objection, including a popular response devised by Beaudoin and Bergmann. I find that all of the examined responses fail. It appears the Pandora's box objection to skeptical theism still stands.
1. The skeptical theist response to the evidential argument from evil
Skeptical theism is currently one of the most popular theistic responses to the evidential argument from evil, a typical version of which is outlined below.
Let an inscrutable evil be an evil that (even after careful reflection) we can think of no God-justifying reason for God, if he exists, to permit. And let a gratuitous evil be an evil there's no God-justifying reason for God, if he exists, to permit. Then an evidential argument from evil runs:
(1) There are inscrutable evils.
(2) Therefore, probably there are gratuitous evils.
(3) God, if he existed, would not permit gratuitous evils.
(4) Therefore, probably God does not exist.
Skeptical theists challenge the inference from (1) to (2). They maintain our inability to think of a God-justifying reason for an evil does not allow us reasonably to conclude there probably is no such reason. Inferences of this form are often termed ‘noseeum’. Noseeum inferences can be sound: the fact that I can’t see any elephants in my garage allows me reasonably to conclude there are probably no elephants there. However, I can’t reasonably conclude there are probably no insects in my garage given only the fact that I can’t spot any (looking in from the street). Given my perceptual limitations, there might still easily be insects present. The skeptical theist maintains that, given our cognitive limitations, the inference from (1) to (2) is similarly flawed. Michael Bergmann, a leading defender of skeptical theism, puts the objection thus:
The fact that humans can’t think of any God-justifying reason for permitting and evil, doesn’t make it likely that there are no such reasons; this is because if God existed, God’s mind would be far greater than our minds so it wouldn’t be surprising if God has reasons we weren’t able to think of. (2012: 11)
According to Bergmann, the skeptical theist’s skepticism (detached from their theism) includes as a main ingredient endorsement of such skeptical theses as:
ST1: We have no good reason for thinking that the possible goods we know of are representative of the possible goods there are.
ST2: We have no good reason for thinking the possible evils we know of are representative of the possible evils there are.
ST3: We have no good reason for thinking that the entailment relations we know of between possible goods and the permission of possible evils are representative of the entailment relations there are between possible goods and the permission of possible evils.
ST4: We have no good reason for thinking that the total moral value or disvalue we perceive in certain complex states of affairs accurately reflects the total moral value or disvalue they really have. (2012: 11-12)
Bergmann maintains that, given the truth of ST1-ST4, we are simply in the dark about whether there exist God-justifying reasons to permit the evils we observe. But then the evidential argument from evil fails.
2. The Pandora’s box objection to skeptical theism
Skeptical theism has been criticised on the grounds that it opens up a skeptical Pandora’s box, generating forms of skepticism that are implausibly wide-ranging and strong. In particular, it is argued that skeptical theism requires we also embrace skepticism about the external world and the past. Why so? Well, how do we know God doesn’t have good reason to create a false impression of an external world, or good reason to create the false impression that the universe and myself are more than five minutes old? Skeptical theism blocks any attempt to justify the belief that there are unlikely to be such God-justifying reasons by means of a noseeum inference: ‘I can’t think of a good reason why God would deceive me in that way, therefore there probably is no such reason.’ But then skeptical theism would seem to have the consequence that, for all I know, God does indeed have a good reason to deceive me in this way and is deceiving me for that reason.
Wilks points out one of the more outlandish skeptical consequences he supposes skeptical theism generates. He imagines an ‘eccentric theist’ who claims God has created a sub-10,000 year old Earth orbited by the sun, with pink elephants. When compelling scientific evidence against these claims is pointed out to our eccentric theist, he replies: ‘We cannot fathom God’s reasons. For all we know, God has good reason to present us with misleading evidence against these claims, despite their truth. But then I have been supplied with no good reason to suppose my claims about a sun-orbited young earth with pink elephants are false.’ Wilks maintains that if skeptical theists are to be consistent, they should accept the reasonableness of this reply, and that if they do so, then
theism comes off looking less rational than it did before the defense… [O]ne might as well spare the effort of dispute and simply pronounce belief in God to be irrational. (2009: 76)
Call the suggestion that skeptical theism leads to such absurd skeptical consequences concerning the external world and past the Pandora’s box objection. My first aim in this paper is to spell out why one of the leading responses to this objection – a response made by, among others, Beaudoin and Bergmann – fails.
3. Bergmann and Beaudoin’s response to the Pandora’s box objection
In response to the Pandora’s box objection, Bergmann appeals to what he calls commonsensism:
Commonsensism: the view that (a) it is clear that we know many of the most obvious things we take ourselves to know (this includes the truth of simple perceptual, memory, introspective, mathematical, logical, and moral beliefs) and that (b) we also know (if we consider the question) that we are not in some skeptical scenario in which we are radically deceived in these beliefs. (2012: 10)
Having defined commonsensism, Bergmann asks us to consider Sally, a hypothetical agnostic who endorses skeptical theses ST1-ST4 but who, given her commonsensism, can still know many things via perception and memory:
Take for example her knowledge that she has two hands. Given Sally’s commonsensism – in particular, clause (b) – she knows, in addition to the fact that she has hands, that’s she’s not a brain in a vat being deceived into thinking she has hands. And similarly, she knows that if God exists, then God doesn't have an all-things-considered good reason for making it seems that she has hands when in fact she doesn’t. She knows this despite her endorsement of ST1-ST4… By endorsing ST1-ST4, Sally is committing herself to the view that we don't know, just by reflecting on possible goods, possible evils, the entailment relations between them, and their seeming value or disvalue, what God’s reasons might be. But it doesn't follow that we have no way at all of knowing anything about what reasons God might have for doing things… In general, for all the things we commonsensically know to be true, we know that God, (if God exists) didn't have an all-things-considered good reason to make them false (2012: 15)
Beaudoin suggests a similar move in response to the objection that skeptical theism entails skepticism about s, where s is the state of affairs in which God created an old-looking universe just five minutes ago. This objection, counters Beaudoin,
presupposes that the basis on which any skeptical theist believes God does not actualize s is an… inference from ‘I can’t see what would justify God’s actualizing s’ to ‘probably there is no reason - probably God does not actualize s.’ This basis for believing that s does not obtain is unavailable to the skeptical theist… But the point is other… reasons… might still be available to the skeptical theist… Consider an analogy. Suppose I know nothing about Smith’s honesty, or lack thereof. For all I know, Smith is an inveterate liar. Now I claim to believe something (P) Smith told me, but not on the basis of Smith’s telling me; instead I’ve confirmed with my own eyes that (P). Clearly in this case it wouldn’t do for someone to challenge the rationality of my belief by pointing out that for all I know Smith is a liar; my belief that (P) isn’t based on Smith’s testimony… Perhaps there is some theologically neutral, telling philosophical argument for rejecting skepticism about the past. If there is, then on this basis the skeptical theist can conclude that God has no [morally sufficient reason] for actualizing s, since he has not actualized it. (2005: 44-45)
According to Bergmann and Beaudoin, then, given there are other ways of knowing about the external world and the past (ways that don’t rely on any noseeum inference regarding God’s reasons), skeptical theism constitutes no threat to such knowledge. But then, granted the fact that the skeptical theist does indeed possess knowledge of the external world and past, they can conclude that God has not, for some unknown reason, radically deceived them about such things.
Call this the Bergmann/Beaudoin response to the Pandora’s box objection. As I explain below, the Bergmann/Beaudoin response fails.
4. Why the Bergmann/Beaudoin response fails
In the terminology of epistemic defeat, the reason why skeptical theism might appear to require we embrace skepticism concerning the external world and past is that it appears to generate an undercutting defeater for all our beliefs grounded in perceptual experience and memory. A stock illustration of an undercutting defeater involves widgets on an assembly line. Given the widgets appear perceptually to me be red, I am prima facie justified in believing that they are red. However, if I'm subsequently informed by a reliable person that the widgets are illuminated by a red light (to reveal imperfections) that makes them appear red even if they are not, then, it’s suggested, I come to possess an undercutting defeater for my original belief. Why so? Well I now possess good grounds for thinking that the method by which I acquired by original belief, is, in the circumstances in which I formed it, unreliable and not to be trusted.
But what, exactly, is defeated in such cases? Typically, it's supposed that justification, and thus knowledge, are defeated. On acquiring that new evidence about the red light, I can no longer be said either to justifiably believe or to know that the widgets are red.
Now, it is controversial whether, in such a case, justification and knowledge really are lost. Lasonen Aarnio (2010) suggests that the intuition that knowledge is lost in such cases is often misleading. The implications of Lasonen Aarnio’s view for the Pandora’s box objection will be discussed towards the end of this paper. For argument's sake, I shall accept for the time being that the widespread intuition that justification and knowledge are lost in such cases is indeed correct.
Why suppose skeptical theism generates a defeater for beliefs about the external world and past? Well, given that it appears to me both that I ate toast for breakfast this morning and that there is an orange on the table in front of me, perhaps I am prima facie justified in believing I ate toast for breakfast and that there is an orange before me. But if I now learn that, (i) God exists, and (ii) for all I know, God has an all-things-considered good reason to deceive me about these things, then, runs the objection, I can no longer justifiably believe I had toast for breakfast or that there is an orange there. At the heart of the Pandora’s box objection lies the thought that, just as learning about that red light generates an undercutting defeater for the belief that the widgets are red, so learning that (i) and (ii) generates an undercutting defeater for beliefs about the external world and past.
Consider what appears to be an analogous case.
Olly’s orange. Suppose I see what appears to be an orange on the table in front of me. Let’s assume I'm thereby prima facie justified, and indeed can be considered commonsensically to know, that there’s an orange there. But suppose I then discover the following. Someone – call him Olly – possesses a holographic projector capable of producing entirely convincing-looking visual appearances onto the table in front of me. Now suppose the probability that Olly is using his projector is inscrutable to me. Suppose, for example, that I learn Olly has an urn of balls. Prior to my observing the table, Olly selected a ball at random from this urn. If the ball was black, Olly projected an entirely convincing-looking holographic image of an orange onto the table. If Olly selected a non-black ball, he placed a real orange on the table. I have no clue concerning what proportion of balls in Olly’s urn are black. For all I know, all the balls are black, none are black, 50% are black, etc. I can’t reasonably assign any probability to any of these hypotheses. Thus I remain in the dark about whether Olly placed a real orange, rather than a holographic image of an orange, on the table.
On being informed by a generally reliable source of this backstory to my experience, do I remain justified in believing there is an orange on the table before me? Can I be said to know there’s an orange there? Intuitively not. Even if there’s a real orange before me, it appears I’m no longer justified in believing this. For all I know, I'm observing a holographic image. The backstory appears to provide me with an undercutting defeater for my belief that there is a real orange on the table, notwithstanding the fact that I might otherwise have been justified in believing, and indeed might otherwise have been considered commonsensically to know, that there’s an orange present.
But suppose I now attempt to defend in the following manner (Beaudoin-and-Bergmann-style) my belief that there’s an orange before me. Of course I don’t believe there’s an orange there because I suppose it’s unlikely Olly picked a black ball from his urn. Rather, I have some other way of knowing there’s an orange there: in this case direct perceptual experience. Given there clearly appears to be an orange present, I can commonsensically consider myself to know there is an orange present. And, granted I do know there is an orange present, but can know this only if Olly didn’t pick a black ball, I can conclude Olly didn’t pick a black ball.
Clearly, the above argument fails. It overlooks the fact that the backstory about Olly and his urn appears to provide me with a defeater for my belief that there is an orange before me despite the fact that my belief is grounded in direct perceptual experience. Beliefs that are prima facie justified and that may be commonsensically considered known given such an experience can in principle be defeated, and such a defeater is what the backstory about Olly and the urn appears to generate.
At the heart of the Pandora’s box objection lies the thought that skeptical theism provides us with an analogous backstory to our everyday perceptual experiences. Ordinarily, perhaps I'm prima facie justified in believing, and indeed can be commonsensically considered to know, that there is an orange before me given that is how things visually appear. But if I learn there is a God who has complete control over my perceptual experiences, and that, for all I know, this God has good reason both to generate a false impression of an orange and indeed deceive me about the external world more generally, then this discovery appears analogously to supply me with an undercutting defeater for my belief that there is an orange on the table. If I can no longer be said to know there’s an orange on the table given my discovery of the backstory about Olly and the urn, how can I be said to know there’s an orange on the table given my discovery of the truth of skeptical theism?
Bergmann and Beaudoin suppose that to argue that skeptical theism provides grounds for withholding judgement about the external world and the past is akin to arguing that the fact that I am in the dark about whether Smith is an inveterate liar gives me grounds for suspending judgement about the truth of Smith’s assertion that (P). Beaudoin reminds us, correctly, that I might have independent grounds for believing (P), and thus grounds for supposing Smith isn’t lying about (P). Bergmann and Beaudoin suggest that, in the same way, I may have some independent way of knowing about the external world and the past (i.e. some way independent of inferring that God has no reason to deceive me given only that I cannot think of such a reason). They then insist that, granted the fact that I do have knowledge about the external world and past by this other route, I can conclude that God has not, for some unknown reason, radically deceived me about such things.
As should now be clear, the analogy Beaudoin tries to draw with the Smith case fails. What skeptical theism appears to generate is not just a defeater for beliefs about the external world and past based on a noseeum inference about God’s reasons, but a defeater for beliefs about the external world and past grounded in other potential methods of knowing too, including perceptual experience and memory. But then pointing out that skeptical theists don’t attempt to justify their beliefs in the external world and the past by means of such a noseeum fails to engage with the objection raised.
Notice that for atheists who embrace the skeptical part of skeptical theism, no such defeater need be generated. The atheist who accepts ST1-4 is in a position analogous to someone who justifiably believes that while there is indeed an urn containing some unknown percentage of black balls, there’s no such person as Olly who generates a deceptive perceptual appearance of an orange if the ball he draws at random from that urn is black. Such an individual does not, on learning about the urn and its mysterious contents, come to possess an undercutting defeater for their belief that there is an orange before them given only that is how things visually appear.
So, while the Pandora’s Box objection to skeptical theism might yet be successfully dealt with, the Bergmann/Beaudoin response fails.
5. Relevant disanalogies?
The skeptical theist may insist there's some relevant difference between my situation in Olly’s orange and that in which skeptical theists find themselves: a difference that explains why my coming to believe the backstory in Olly’s orange generates a defeater for my belief that there’s an orange before me, whereas coming to believe the truth of skeptical theism does not. Perhaps there is such a difference: I won’t attempt to deal here with every suggestion here that might be made, but I will look at two more obvious suggestions and explain why both fail.
First, consider the suggestion that it is the role of a certain sort of probabilistic mechanism - pulling balls from an urn at random in to determine whether or not to project a deceptive image - that leads us to suppose a defeater is generated in Olly’s orange. But then, as no such probabilistic mechanism is employed by God in determining whether or not to give us deceptive experiences, the skeptical theist is not in a relevantly similar situation.
However, in Olly’s orange, the urn/ball component of the backstory would seem to be inessential so far as the intuition of defeat is concerned. What generates the intuition of defeat is the fact that I’m in the dark about the probability of it being a real orange rather than a deceptive image that Olly placed on the table. The urn/ball component is included in the backstory to explain why I'm in the dark about that probability, but that component is optional. No explanation of why I'm in the dark about probability need be included. Alternatively, my being in the dark about that probability might be explained by my being in the dark about the probability that Olly has an all-things-considered good reason to place a deceptive image rather than a real orange on the table (this would obviously make Olly’s orange still more closely analogous to the skeptical theist’s position). Either way, the story generates the same intuition of defeat.
A second suggestion regarding a relevant disanalogy between Olly’s orange and the skeptical theist’s situation is that the skeptical theist may have good reason to suppose that God, if he exists, is morally perfect, and that a morally perfect God will not deceive us even if he has an all-things-considered good reason to do so. Thus the probability that we are being deceived by God, if he exists, is not, as it is in Olly's case, inscrutable, but low.
But why suppose a morally perfect God won’t deceive us? Descartes offers an argument for that claim in his Third Meditation, where he says God ‘cannot be a deceiver, since it is a dictate of the natural light that all fraud and deception spring from some defect’, and God is without defect. But as Maitzen (2009) points out, while all fraud and deception flow from some defective situation (a terrorist about to explode a bomb who can only be thwarted by deception, for example) it does not follow that ‘fraud and deception are defective responses to that situation’ (2009, 97). Maitzen here follows Hobbes who, in response to Descartes, points out that it
… is the common belief that no fault is committed by medical men who deceive sick people for health’s sake, nor by parents who mislead their children for their good … M. Descartes must therefore look to the this proposition, God can in no case deceive us, taken universally, and see whether it is true… (Haldane and Ross 1967: 78)
Where an all-things-considered good reason to deceive exists, our engaging in such deception does not require there be any defect in us. So why would God’s engaging in such deception require there be some defect in him?
Furthermore, those who consider the New Testament a reliable source of information about God should note that it contains passages suggesting God does indeed engage in deliberate deception. For example, St. Paul describes God as sending some people ‘a powerful delusion, leading them to believe what is false.’ (2nd Thessalonians 2:11). So the thought that God is no deceiver appears Biblically challenged, too.
To conclude this section: there may be some relevant disanalogy between the skeptical theist’s position and mine in Olly’s orange which explains why, though my belief is defeated in Olly’s orange, the skeptical theist’s beliefs about the external world are not. However, neither of above suggestions appear to succeed in identifying such a disanalogy.
6. Externalism and defeat
Finally, I want briefly to anticipate some other responses to the Pandora’s box objection – responses grounded in externalist thinking about knowledge and defeat.
Skeptical theism is usually associated with externalist epistemologies on which whether or not a subject is justified and/or warranted in believing that p is determined by factors that may lie beyond the awareness of that subject - factors such as whether the belief was formed in a reliable way and/or via properly functioning faculties. Externalists typically allow that a subject’s beliefs may be justified/warranted even if they lack information about whether such conditions are satisfied. Externalists may be right about that.
However, from the supposed fact that you do not need information about the reliability of your faculties in order to have knowledge or justified belief about the world, it does not follow that the acquisition of such information cannot affect what you know or are justified in believing about the world. Indeed, many externalists, Bergmann included, allow that if a subject comes to possess information that their belief was formed in an unreliable way, then their belief may be defeated (Bergmann 1997: 405-6).
Bergmann distinguishes three doxastic attitudes towards a proposition p: believing p; disbelieving p (believing p is false); and withholding p (refraining from either believing or disbelieving p). (He also allows one can also take no doxastic attitude at all towards a proposition (2005: 422).) Bergmann proposes that, where p*S is the proposition that S’s belief that p is formed in a reliable way, then disbelieving or even just withholding on p*S supplies S with a defeater for the belief that p (2005: 426).
Bergmann uses the following modified widget example to illustrate how withholding on p*S generates a defeater for p. Suppose Sally comes to form the belief that the widgets are red based on how the widgets look to her as they pass by on the conveyer belt. And suppose Sally has no idea whether there is a red light shining on the widgets or even how likely it is that there would be such a light shining on them. Bergmann continues:
Sally now considers the higher-level proposition that her belief The widgets are red is formed in a reliable way. Being completely uncertain about whether that higher-level proposition is true, she resists believing both it and its denial. In other words, if p is the proposition The widgets are red, she withholds p*Sally. Does this give her, in these circumstances, a defeater for her belief that the widgets are red? I think it does. (2005: 426)
So, on Bergmann’s view, a belief is defeated if one either disbelieves, or even just withholds judgement on whether, the belief was formed in a reliable way.
The above principle would explain why, in Olly’s Orange, my belief that there is an orange on the table before me is defeated. On realizing I’m in the dark about whether Olly picked a black ball from his urn (and so generated a deceptive impression of an orange) I disbelieve, or at least withhold on whether, my belief was formed in a reliable way. Thus my belief is defeated.
So now consider Sarah, a skeptical theist, who, as a result of her perceptual experience, believes there’s an orange on the table before her. On Bergmann’s view, Sarah’s belief about the orange is defeated if, as a result of her skeptical theism, she comes to disbelieve, or even just withhold judgement on whether, her belief was formed in a reliable way. Now I take it that at the heart of the Pandora’s box objection lies something like the following thought. Given her skeptical theism, Sarah really should suppose she is in the dark about whether God has an all-things-considered good reason to deceive her about the orange. But then she should disbelieve, or at least withhold, on whether her belief about the orange was formed in a reliable way. So she should consider her belief defeated.
Now, in response, an externalist like Bergmann may point out, correctly, that he is committed only to S’s belief that p being defeated if S does in fact disbelieve or withhold on p*S. Bergmann may insist that, so long as Sarah doesn’t actually disbelieve or withhold judgment on whether her belief about the orange before was formed in a reliable way, no defeater is generated. So let’s suppose Sarah fails either to consider the matter of whether her belief about the orange was reliably formed, or that, if she does consider it, she finds herself unable to do anything other than believe it was reliably formed, notwithstanding her skeptical theism. Then Sarah’s skeptical theism fails to generate a defeater for her belief. And so, assuming the relevant externalist conditions for knowledge are met, Sarah can still know there’s an orange present.
Does the above suggestion allow a skeptical theist successfully to deal with the Pandora’s box objection? I don’t see that it does. Let’s return to Olly’s orange for a moment. Suppose that, having accepted the backstory about Olly and his urn, I nevertheless continue to believe that my belief that there’s an orange on the table before me is reliably formed. On Bergmann's characterisation of defeat, given that I too fail to disbelieve or withhold on whether my belief was reliably formed, my belief remains undefeated. So, given my belief is undefeated, can I reasonably take myself to know there’s an orange present?
Intuitively not. True my belief about the orange remains undefeated (given Bergmann’s characterisation). But, given my acceptance of the backstory about Olly and his urn (that Olly has the means to deceive me, did deceive me if he picked a black ball from his urn, and I'm in the dark about whether he picked a black ball), surely I should consider my belief defeated. And if I should consider it defeated, then I shouldn’t suppose I commonsensically know it to be true. I should be skeptical about that orange.
But then similarly, if skeptical theism has the consequence that Sarah should, on reflection, consider her belief about the orange defeated, then she shouldn’t suppose she commonsensically knows there’s an orange before her either. Sarah should be skeptical about her orange. And, given his skeptical theism, Bergmann should be skeptical about his.
Here’s a second suggestion as to how their externalism might allow skeptical theists to deal with the Pandora’s box objection. When introducing the notion of defeat above, I mentioned that we might question the reliability of our intuitions with respect to widget and other cases in which it’s usually supposed that an undercutting defeater has been generated. Maria Lasonen Aarnio argues that externalists should take seriously the suggestion that knowledge can be retained even in the face of seemingly strong defeating evidence.
Suppose, for example, that I judge the widgets are red based on visual appearance. I then come to possess strong evidence that there’s red lighting in play that makes non-red things look red. Suppose that, despite my acquiring this new evidence, I nevertheless stick with my belief that the widgets are red. And suppose that, as a matter of fact, the new evidence is misleading - in fact there is no red lighting in play and the widgets really are as they appear to be. Then, according to Lasonen Aarnio, I may still know the widgets are red. For it may be that the relevant externalist conditions on knowledge are satisfied (so, for example, the method by which I arrive at my belief may still be safe).
So why do we intuit that knowledge is lost in such cases? Because, suggests Lasonen Aarnio, the policy of continuing to believe, given the new evidence, is unreasonable. But, suggests Lasonen Aarnio, it doesn't follow from the fact that my continued belief is unreasonable that I don't know. This is an example of what Lasonen Aarnio calls unreasonable knowledge.
In what sense is my continued belief unreasonable? Lasonen Aarnio suggests reasonableness
is at least largely a matter of managing one’s beliefs through the adoption of policies that are generally knowledge conducive, thereby manifesting dispositions to know and avoid false belief across a wide range of normal cases. Subjects who stubbornly stick to their beliefs in the face of new evidence manifest dispositions that are bad given the goal of knowledge or even of true belief.’ ((2010) 2)
Consider, for example, the rule or method of belief formation that tells you to believe that p when you see that p even in the presence of good evidence for thinking that your senses are not to be trusted. This method is, in a sense, good, in that if you follow it, beliefs obtained as a result will be safe (for, given you can see that p only if p is true, the policy can't produce a false belief).
However, the above method is epistemically a bad method to adopt, suggests Lasonen Aarnio, because adopting it results in a bad disposition. Lasonen Aarnio notes that a 'subject who adopts this method is also disposed to believe p when she merely seems to see that p in the presence of evidence for thinking that her senses are not to be trusted' (2010, 14 my italics). But then, if a subject were to adopt the method, they would end up believing p in a significant proportion of cases in which the evidence that their senses are not to be trusted is not misleading. So while the method is indeed safe, its adoption results in dispositions that are not knowledge conducive:
This is why the rule believe p when you see that p in the presence of evidence for thinking that your senses are not to be trusted is not part of a policy that is knowledge conducive in the intended sense. A reasonable subject would not adopt or follow such a rule, even though it is success entailing. (2010, 15)
On Lasonen Aarnio's view, someone presented with evidence that the method by which they acquired their original belief is untrustworthy should withhold belief. If they fail to withhold, they are being (in Lasonen Aarnio's sense) unreasonable. They can be properly criticised for sticking with their original belief. But that's not to say they don't know.
So, if Lasonon Aarnio is right, perhaps I might continue to know that there’s an orange on the table even after I'm presented with the evidence about Olly and his holographic projector. If I continue to believe there’s an orange there, and it so happens that Olly's holographic projector is not deceptively employed (i.e. my belief is actually a product of a safe method), I can still know there's on orange present. But then can't the skeptical theist suggest that, for much the same reason, Sarah’s skeptical theism fails to generate a defeater for her belief that there’s an orange before her. Just so long as Sarah continues to believe there’s an orange there, she might similarly continue to know (assuming the relevant externalist conditions - e.g. safety conditions - on knowledge are met).
Of course the Pandora’s box objection is not so easily dealt with. Even on Lasonen Aarnio’s view, it remains unreasonable for me to believe that there’s an orange on the table given the new evidence concerning Olly and his holographic projector. Whether or not my belief is defeated (it may not be), and whether or not I know there's an orange before me (perhaps I do), I should revise my belief about the orange given the new evidence. It's unreasonable for me not to withhold belief, not to become skeptical. But then, if the analogy drawn between Olly’s orange and skeptical theist’s position is correct, it's similarly unreasonable for Sarah to believe there’s an orange before her given her skeptical theism. Whether or not Sarah knows there’s an orange present (and she might), her skeptical theism should lead her to be skeptical about that orange. For, just as in Olly's orange, she has reason to distrust the method by which she acquired her belief.
Here’s a third and final suggestion how externalism might allow skeptical theists to deal with the Pandora’s box objection. As we have just seen, the proponent of the Pandora’s Box objection may insist that, whether or not Sarah knows there's an orange before her, her skeptical theism should lead her to be skeptical about that orange and indeed about the external world more generally. An externalist may resist that conclusion by maintaining that what one should or shouldn’t believe depends on ones cognitive design plan (which specifies how ones cognitive faculties are supposed to work), and it may be that God has designed our cognitive faculties in such a way that, while local skepticism about the orange is appropriate in Olly’s orange, we should never embrace global skepticism about the external world, not even if we have been presented with logically impeccable arguments for being globally skeptical (notice that, given we do indeed inhabit a world of the sort we seem to see around us, this particular cognitive design plan may even be aimed at truth). In Sarah’s case, unlike in Olly’s orange, it’s not just belief in the presence of an orange that’s threatened by her skeptical theism, but all her beliefs about the external world. But if Sarah’s cognitive design plan is such that no argument, no matter how good, should ever lead her to embrace that sort of skepticism, then her skeptical theism shouldn’t lead her to embrace it. The proponent of the Pandora’s box objection is mistaken in supposing otherwise.
The above response muddles two varieties of ‘should’. The proponent of the Pandora’s box objection insists that, given her skeptical theism, Sarah should embrace skepticism about the external world, in the sense that this is what logic dictates. Now Sarah’s cognitive design plan may be such that she should never accept such a skeptical conclusion, irrespective of the strength of any argument for it. But if the force of an argument is such that, logically speaking, Sarah should be skeptical about the external world, then, surely, under those circumstances, Sarah’s design plan requires that she believe illogically. Sarah should, logically speaking, be skeptical, irrespective of what her design plan dictates. But then, if the Pandora's box objection is that Sarah should embrace skepticism about the external world in the sense that this is what her skeptical theism logically requires of her, then the above response clearly fails to engage with that objection. It's that last italicised 'should' that proponents of the Pandora's box objection are presumably insisting upon, irrespective of what Sarah's cognitive design plan says she should do.
In short, I do not yet see how the resources provided by epistemic externalism allow a skeptical theist like Bergmann to deal effectively with the Pandora’s box objection.
Bergmann attempts to deal with the Pandora’s box objection to skeptical theism by appealing to commonsensism and the thought that beliefs grounded in simple perceptual experience and memory provide us with a secure basis from which we may then establish that God lacks an all-things-considered good reason to deceive us about such things. I have explained why, as it stands, that particular solution fails. I then examined a number of other suggestions as to how the skeptical theist might deal with the Pandora’s box objection - in particular, by appealing to (i) God’s moral perfection, and/or (ii) externalist thinking about defeat. None of the examined suggestions prove successful. It seems to me that, currently, there is no satisfactory skeptical theist response to the Pandora's box objection.
Alston, W. (1991). The inductive argument from evil and the human cognitive condition. In J. Tomberlin (Ed.), Philosophical Perspectives 5: Philosophy of Religion (pp. 29-67). Atascadero CA: Ridgeview.
(1996). Some (temporarily) final thoughts on evidential arguments from evil. In Howard-Snyder (ed.) (1996) (pp.311-32).
Beaudoin, J. (2005). Skepticism and the skeptical theist. Faith and Philosophy, 22, 42-56.
Bergmann, M. (1997). Internalism, externalism, and the no-defeater condition. Synthese, 110, 399-417.
(2001). Skeptical theism and Rowe's new evidential argument from evil. Noûs, 35, 278-96
(2005). Defeaters and higher-level requirements. Philosophical Quarterly, 55, 419-436.
(2006). Justification Without Awareness. Oxford: Oxford University Press.
(2009). Skeptical theism and the problem of evil. In T. Flint and M. Rea (eds.) Oxford Handbook to Philosophical Theology (pp. 374-99). Oxford: Oxford University Press.
(2012). Commonsense skeptical theism. In K. Clark and M. Rea (eds.) Science, Religion, and Metaphysics: New Essays on the Philosophy of Alvin Plantinga (pp. 9-30). Oxford: Oxford University Press.
Feldman, R. (2005). Respecting the evidence. Philosophical Perspectives, 19, 95-119,
Fitzpatrick, F. J. (1981). The onus of proof in arguments about the problem of evil. Religious Studies, 17, 19-38.
Gale, R. (1996). Some difficulties in theistic treatments of evil. In Howard-Snyder (1996), 206-18.
Haldane, E, and Ross, G.R.T. (trans.). (1967). The Philosophical Works of Descartes, Volume II (Cambridge: Cambridge University Press).
Howard-Snyder, D. (ed.) (1996). The Evidential Argument from Evil. Indiana: Indiana University Free Press.
1996a. Introduction to Howard-Snyder (ed.) (1996).
Lasonen Aarnio, M. (2010). Unreasonable Knowledge. Philosophical Perspectives, 24, 1-21.
Maitzen, S. (2009). Skeptical theism and moral obligation. International Journal of the Philosophy of Religion, 65, 93-103.
McBrayer, J. (2012). Are skeptical theists really skeptics? Sometimes yes and sometimes no. International Journal of the Philosophy of Religion, 72, 3-16.
McBrayer, J. and Swenson, P. (2012). Skepticism and the argument from divine hiddenness. Religious Studies, 48, 129-150.
Moore, G.E. 1(959). A defence of common sense. In his Philosophical Papers. London: George Allen and Unwin.
Plantinga, A. (1996). Epistemic probability and evil, in Howard-Snyder (ed.) (1996), 69-96.
Russell, B. (1996). Defenseless. In Howard-Snyder (1996), 193-205.
Segal, A. (2011). Skeptical theism and divine truths. Religious Studies, 47, 85-95.
Van Inwagen, P. (1996). The problem of evil, air, and silence. In Howard Snyder (ed.) (1996), 151-174.
Wilkes, I. (2009). Skeptical theism and empirical unfalsifiability. Faith and Philosophy, 26, 64-76.
Williamson, T. (2000). Knowledge and its Limits (Oxford: Oxford University Press).
Wykstra, S. (1984). The Humean obstacle to evidential arguments from suffering: on avoiding the evils of ‘appearance’. International Journal for Philosophy of Religion, 16, 73-93.
(1996). Rowe's noseeum arguments from evil. In Howard-Snyder (1996), 126-50.
 Proponents of a skeptical theist response to the evidential argument from evil include Alston (1991, 1996), Bergmann (2001, 2009), Fitzpatrick (1981), Howard-Snyder (1996a), McBrayer and Swenson (2012), Plantinga (1996), Segal (2011), van Inwagen (1996), and Wykstra (1984, 1996).
 After Wykstra (1996): ‘We don’t see ‘um so they probably ain’t there.’
 See for example Russell (1996), Gale (1996).
 As already noted, the accuracy of such intuitions has been question. I address this worry towards the end of this paper.
 In fact, there’s a prima facie case for saying, not just that Bergmann shouldn’t consider himself commonsensically to know there an orange present, but also that he doesn’t know there’s orange present. In Justification Without Awareness (2006) Bergmann considers a case where he supposes a subject, Jill, clearly should consider her belief defeated given her background knowledge. Jill bets her brother that both their parents are out of town that day given what she’s been told by a reputable source. Jill knows that if she wins she gets $300 that will enable her to buy a bike. Jill and her brother now see both parents walk in, yet Jill continues to believe she’ll be able to buy that bike. Bergmann observes that Jill fails ‘to put two and two together’ in the way she should. He concludes that while Jill’s belief is not defeated, neither is it known. This is because, on Bergmann’s view, Jill’s ‘defeater system is not functioning properly’ (2006: 171), this being another Bergmannian condition on knowledge. Someone like Jill should, in a case like this, ‘put two and two together’.
The proponent of the Pandora’s box objection will presumably point out that Bergmann’s own defeater system would appear not to be functioning properly if Bergmann similarly fails to ‘put two and two together’ and conclude that his perceptually grounded belief that there’s an orange before him is defeated given his skeptical theism has the consequence that he’s in the dark about whether God has an all-things-considered good reason to deceive Bergmann about that orange. Our critic will insist Bergmann should suppose his belief is defeated given his acceptance of skeptical theism in just the same way that I should consider my belief there’s an orange before me is defeated given I accept the backstory about Olly and his urn. Bergmann may insist there is some relevant disanalogy between his situation and mine in Olly’s Orange, but the onus is presumably now on Bergmann to explain what the disanalogy is. There is at least a prima facie case here for saying Bergmann does not know there’s an orange before him. However, see my final comments re Lasonen Aarnio on defeat.
 Safety conditions on knowledge are associated particularly with Williamson, Sosa, and Pritchard. A simple example of a safety condition says S knows P only if S is safe from error; that is, there must be no risk that S believes falsely in a similar case. So, for example, if Ted looks at a stopped clock when it happens to read the right time, his belief is not safe, because his belief could easily have been false. For an example of the safety view see Williamson (2000).
 Bergmann offers a 'proper function' theory of justification in which cognitive design plans play a key role. See Bergmann (2006 chpt. 5). Bergmann does not actually offer the response to the Pandora's box objection that I sketch here. It's merely a response of a sort that I anticipate Bergmann or other skeptical theists might yet make.