Sosa on scepticism
- First Online:
- Cite this article as:
- Brown, J. Philos Stud (2009) 143: 397. doi:10.1007/s11098-009-9342-5
- 82 Views
In my remarks, I discuss Sosa's attempt to deal with the sceptical threat posed by dreaming. Sosa explores two replies to the problem of dreaming scepticism. First, he argues that, on the imagination model of dreaming, dreaming does not threaten the safety of our beliefs. Second, he argues that knowledge does not require safety, but a weaker condition which is not threatened by dreaming skepticism. I raise questions about both elements of his reply.
In my remarks, I focus on one theme from Sosa’s rich discussion, the problem of scepticism. In his earlier work, Sosa developed a distinctive account of knowledge which, unlike other externalist accounts such as Nozick’s, maintains closure for knowledge. Nozick was led to deny closure by his sensitivity condition: if p were not the case, one would not believe that p. Sosa rejects the sensitivity condition instead requiring that one’s belief that p be safe, i.e. not easily would one falsely believe that p. On the assumption that the brain in the vat scenario is a far away possibility, not easily would one be in such a scenario falsely believing that one is not a brain in a vat. As a result, on Sosa’s view, one knows that one has hands and one knows that one is not a brain in a vat. A further important element of Sosa’s position is his distinction between “animal” and “reflective” knowledge. The latter, but not the former, requires “broad coherence, including one’s ability to place one’s first level knowledge in epistemic perspective” (Sosa, J Philos, 8:280, 1997). One acquires reflective knowledge by reasoning from one’s animal knowledge to yield a safe and coherent set of beliefs about one’s own epistemic position in the world.
In his book, Sosa develops this overall view to deal with a further sceptical worry, that arising from the possibility of dreaming. Unlike the brain in a vat scenario, the dreaming possibility is a nearby one. So, if in dreams I have false beliefs about the world, then this does undermine the safety of my ordinary beliefs. Sosa explores two replies to the problem of dreaming scepticism. First, he argues that, on the imagination model of dreaming, dreaming does not threaten the safety of our beliefs (Chap. 1). Second, he argues that knowledge does not require safety, but a weaker condition which is not threatened by dreaming scepticism (Chap. 2).
2 The imagination model of dreaming
Later on, Sosa asks, “how then can I not arbitrarily take myself to be awake, when I cannot distinguish my state internally from that of a realistic dream?” (16). He continues, “[a]t an unreflective level, epistemic justification can hence derive from the holding of a condition whose absence is no more subjectively distinguishable from its presence than is a realistic dream from waking life. Still, without reflective, non-arbitrary assurance that you satisfy that condition, you cannot know reflectively something you might still know at the animal level.” (16). Sosa appeals to the imagination model to argue that it is rational and nonarbitrary to take oneself to be awake.
“The dream possibility still threatens, even if it is no threat to the safety of our beliefs.
How then are dreams a threat? What they threaten is not the safety of our beliefs but perhaps their rationality. Can it be rationally coherent to grant that one could be dreaming? How can one rationally allow that possibility, as one must do if unable to rule it out? That would seem incoherent, but exactly how?” (Sosa 2007:14).
Before looking at the details of Sosa’s view, it is worthwhile asking whether this further elaboration of his view is necessary. It is unclear to me why Sosa thinks that answering the safety challenge posed by dreaming scepticism suffices to defend animal knowledge but not reflective knowledge. As he introduced it, reflective knowledge is attained by reasoning from one’s animal knowledge to attain a coherent picture of oneself and one’s epistemic position in the world. By showing that dreaming does not undermine safety, Sosa has shown how certain of one’s beliefs may constitute animal knowledge. The mere possibility that one might be dreaming does not undermine one’s ability when awake to reason reflectively from one’s animal knowledge to gain a coherent perspective on one’s epistemic position in the world. Now, assuming that knowledge requires safety, beliefs so gained constitute reflective knowledge only if they are safe. For instance, the belief that perception is a safe method of belief formation constitutes reflective knowledge only if it is safe. But, if dreaming does not involve false belief then it provides no challenge to the safety of belief at either first or second order level. Why, then, does Sosa suppose that answering the safety challenge posed by dreaming scepticism is not enough to defend reflective knowledge?
An alternative way to ask what sceptical problem dreaming poses for reflective knowledge is by examining the debate between externalist and internalist approaches in epistemology. Externalists take it that one can defuse scepticism by pointing out that, on an externalist notion of knowledge, the sceptic has not established that one lacks knowledge, or second order knowledge that one knows. On an externalist account, a belief constitutes knowledge if it meets certain external conditions. The externalist’s point is that the sceptic has failed to show that the relevant external condition does not obtain. For instance, the sceptic has not shown that the safety condition fails for either one’s first order beliefs, or one’s second order belief that one has knowledge. Perhaps, then, dreaming scepticism poses an extra threat to reflective knowledge only if one adopts an internalist approach to knowledge. If that is right, then it raises a dilemma for Sosa. On the one hand, the further elaboration of his view is not needed on an externalist approach. On the other, an internalist approach could motivate the demand for further work, but makes it harder to meet that demand. Let us keep this point in mind in evaluating Sosa’s view.
Sosa appeals to the imagination model to argue that we can distinguish waking life from a corresponding dream even though they are content identical. “What enables us to distinguish the two content-identical states is just the fact that in the dream state we do not affirm anything—not that we are veridically perceiving an external world, nor that we are not—whereas in waking life we do knowingly perceive our surroundings (18)”. Let us grant Sosa that, in one sense, one can distinguish being awake and dreaming. As Sosa himself notes, that is compatible with there being a further internalist sense in which one cannot distinguish being awake from dreaming, (18, note 16). At several points, he describes the dreaming possibility as “internally indistinguishable” or “subjectively indistinguishable” and suggests that a dream may be content identical to waking life (14–16). For instance, he states the problem posed by dreaming to be “how then can I not arbitrarily take myself to be awake, when I cannot distinguish my state internally from that of a realistic dream?” (16). It seems, then, that only an externalist would be moved by the thought that there is a (externalist) sense of distinguishability in terms of which one can distinguish being awake and dreaming. But, if that is right, one wonders whether by focusing on the notion of discriminability, Sosa has assuaged any worry that would not already be answered by the point that dreaming does not undermine safety.1
In addition to arguing that one can distinguish the states of being awake and being asleep, Sosa argues that one is “automatically, rationally committed to supposing that one is not just dreaming, whenever one inquires at all” (20). There are three options with respect to the claim that one is not just dreaming: disbelief, suspension, and belief. Sosa argues that it is a consequence of the imagination model of dreaming that, if I am able to form the belief that p, or even suspend judgement on the matter of p, then I am awake, for, when I am dreaming, I am unable to do either of these things. As a result, “disbelieving is defective, since self-defeating, for I know that if I take that option I will be wrong. Suspending is also defective… since I know, about a particular alternative option (belief), that I am epistemically better off if I take that other option, since I will thereby avail myself of a correct answer to my question which I fail to do if I only suspend judgement.” (19). So, he concludes, the rational thing to do is to believe that I am awake.
Try to believe awake
Try to suspend whether awake
Lack true belief
Fail through dreaming
Illusion of true belief
Illusion of suspension
Try to escape by train
Try to escape by swimming
Warm and cosy
Cold and miserable
Do not manage
Sosa might reply that it follows from his imagination model that attempting to believe oneself awake while dreaming has no negative epistemic value. After all, it seems that there is no epistemic disvalue to engaging in a daydream which involves falsehoods, for example that, instead of sitting down in the office writing a philosophy paper one is instead walking on a beach in Brazil. However, this move brings into sharper focus the question of whether we should adopt the imagination model of dreaming. To the extent that the imagination model involves the assumption that there is no epistemic disvalue in having dreams part of whose content is false, the imagination model seems more philosophically contentious.
However, on the traditional model of dreaming, the cogito is reflectively defensible in a further sense: I cannot be under the illusion of judging that I am thinking something. On the traditional model, dreaming is compatible with my judging and affirming many things. In particular, if it is part of the content of my dream that I judge that I think that p then I do judge that I think that p. Further, the judgement is self-verifying if one understands the occurrence of “think” to refer not specifically to belief, but instead to a thinner notion of thinking which includes not only believing and judging but also hypothesising and imagining. On this thinner notion, in judging that I think that p I thereby make it the case that I am thinking that p. By contrast, on the imagination model, it could be part of the content of my dream that I judge myself to be awake even when I am in fact asleep and so not in a position to make any judgement or affirm anything. Thus, on the imagination model, I can be in a state subjectively indistinguishable from one in which I judge that I am awake, even though I am asleep and so not judging anything at all. As a result, it may seem that, at least relative to how things subjectively seem to me, it is arbitrary and irrational to take it that I am not dreaming.
“The impossibility of being affirmed falsely is thought to help give the cogito a special status, which we can reflectively see that it has. The claim that one is not now just dreaming, being equally impossible to affirm falsely, must have an equally high epistemic status, equally defensible reflectively. For it seems to share with the cogito its pragmatic safety, and its epistemically favourable features more generally, such as a high degree of self-intimation” (17).
I have been focusing on whether to answer dreaming scepticism, Sosa need do more than answer the safety challenge. I suggested that further elaboration is required only on an internalist approach. But, an internalist would not be satisfied by the idea that there is an external sense in which one can distinguish being awake from dreaming, and would argue that, from one’s subjective perspective, it is irrational and arbitrary to take oneself to be awake and not dreaming. So it is not clear whether Sosa’s further elaboration would satisfy those motivated to see the need for it. Now let us examine Sosa’s second response, the idea that knowledge requires aptness not safety. This is supposed to provide a reply to dreaming scepticism even if the imagination model is false. So, from now on, I will leave aside the imagination model and assume the traditional model on which dreaming involves forming false beliefs.
3 Knowledge requires aptness not safety
Sosa argues that knowledge does not require safety but only aptness, where a belief is apt if it is true through one’s epistemic virtue or competence. He argues that, even on the traditional model, dreaming does not undermine aptness and so does not undermine knowledge. The belief that p may be true because of one’s epistemic competence even if it is not safe, since there is a nearby situation in which one would believe that p falsely either through reduced competence, or abnormal conditions for the exercise of the competence. Applying this idea to the case of dreaming, Sosa argues that the “dreamer’s reduced or lost competence may blind him to such features of his experience, features that would enable him to distinguish dreaming from perceiving. Sleep might render one’s conditions abnormal and inadequate for the exercise of perceptual faculties.” 30.
Sosa raises the following problem for his view, that the case of the Kaleidoscope perceiver (Kate) provides an example of an apt belief which is not knowledge. Kate is in fact looking at a red surface in normal lighting conditions and comes to believe truly on the basis of her visual competence that she is looking at a red surface. So her belief is apt. However, her belief is not safe since there is a jokester in the wings who might easily make it the case that she is looking at a white surface so illuminated with red light that she would falsely take it to be a red surface. Intuitively, Kate’s belief is not a case of knowledge. Sosa deals with this apparent counterexample by arguing that Kate’s belief does constitute knowledge. He suggests that we have the erroneous intuition that Kate lacks knowledge since although she has animal knowledge she lacks reflective knowledge. We may question whether Kate really has animal knowledge. However, it is hard to do so merely by appealing to the intuition that she lacks knowledge given that Sosa offers an error theory of this intuition. However, we may put pressure on Sosa’s account of Kate’s epistemic position by comparing Kate with the fake barn subject. The fake barn subject is in area rife with fake barns, although she is unaware of this. She happens to be looking at a real barn in good conditions and, on the basis of her perceptual abilities, forms the true belief that it is a barn. On the standard analysis, her true belief is not knowledge since it is not safe: too easily would she falsely believe that there is barn in front of her. However, it may seem that her belief is still apt for it is true through her perceptual competence. If this is right, then the fake barn subject, like Kate, would count as having animal but not reflective knowledge. So, one might wonder if Sosa’s account has counterintuitive consequences for the fake barn scenario. (A similar query is raised in Kornblith, forthcoming.)
Setting aside this first problem, Sosa raises a further problem for his view. Since a full reply to the sceptic requires both animal and reflective knowledge, Sosa must show how ordinary perceivers differ from Kate in possessing not only animal but also reflective knowledge. Let us consider, then, an ordinary subject, Joe, who like Kate, has animal knowledge that the surface in front of him is red. Sosa needs to show that Joe but not Kate has reflective knowledge. He tries to do so by arguing that, unlike Kate, Joe knows that his belief that he is looking at a red surface is apt. But given the similarities between Joe and Kate, one may wonder how they differ with respect to this second order belief. Both are in fact looking at a red surface in good perceptual conditions and come to believe that the surface is red by using their perceptual competence. Let us call this “the good scenario.” For each of them, there is a nearby bad scenario. For Kate, there is a nearby scenario in which the jokester so alters the ambient lighting that she would mistakenly judge a white surface to be red. For Joe, there is a nearby scenario in which he is dreaming and forms the false judgement that there is a red surface in front of him. In good, Joe and Kate both truly believe that their belief that the surface is red is apt. Further, Sosa grants that, for each, the relevant second order belief is formed as a result of an epistemic meta-competence. So, within Sosa’s framework, Joe’s second order belief could amount to knowledge whereas Kate’s does not, only if Joe’s, but not Kate’s, belief is true via the relevant competence. Sosa explains that there are two ways in which a true belief might be formed otherwise than by a relevant competence. “For any correct belief that p, the correctness of that belief is attributable to a competence only if it derives from the exercise of that competence in appropriate conditions for its exercise, and that exercise in those conditions would not then too easily have issued a false belief” (33). Of the kaleidoscope perceiver, Sosa says he “does not aptly presume his object level perceptual belief to be apt. Any meta-competence in view through which he might get it right in so presuming, seems one that either: (a) is exercised in its normal, minimal conditions (“no apparent sign to the contrary”), but might too easily have been exercised to the effect of a false presumption, given the jokester; or else (b) is not exercised in its normal conditions, since the very presence of the jokester already spoils the conditions. By contrast, the ordinary believer can aptly apprehend the aptness of his object level perceptual belief. For, he can get it right in so presuming through a meta-competence exercised in its appropriate normal conditions. The relevant meta-competence is a default competence of taking it for granted that conditions are appropriately normal, absent some specific sign to the contrary. When asleep and dreaming we exercise no such competence…” (111).
In Good, for Kate not Joe, conditions for the meta-competence are inappropriate,
although, in Good, Kate has the relevant meta-competence and is in appropriate conditions, too easily would exercise of that competence in those conditions have issued in false belief. But, this is not the case for Joe.
Let us first consider whether there is a sense of “normal” or “appropriate” in which, for Joe but not Kate, conditions are normal for the relevant meta-competence. To get a sense of what the normal conditions for the relevant meta-competence might be, it is useful to compare this meta-competence with a first order competence. Joe and Kate both form the first order belief that the surface is red on the basis of their perceptual competence. The normal conditions for perceptual competence presumably include such things as one’s eyes being open and that nothing funny is going on with the lighting, etc. With respect to these conditions, in Good, Joe and Kate are in the same situation: their eyes are open and nothing funny is going on with the lighting, etc. Both Joe and Kate form the second order belief that their belief that the surface is red is apt. This belief is formed by a meta-competence which Sosa specifies as “a default competence of taking it for granted that conditions are appropriately normal, absent some specific sign to the contrary”. This suggests that normal conditions for this meta-competence are that there are no signs to the contrary. On this understanding of normal conditions, in Good, for both Kate and Joe, conditions are appropriate. For, there are no signs to the contrary for either of them. Alternatively one might suggest that conditions for the relevant meta-competence are normal if there is no nearby scenario indistinguishable from the actual scenario in which one would judge falsely that the relevant first order belief is apt. But, on the traditional model of dreaming, in Good, Joe would not then count as being in normal conditions for the relevant meta-competence. For, there is a nearby situation in which he is dreaming in which he falsely judges that his visually based beliefs about the colours of objects around him are apt. Lastly, one might suggest that conditions are normal when the relevant meta-competence can explain the truth of the relevant second order belief. Such a proposed definition of normalcy is unhelpful when we are trying to use the notion of normalcy to establish whether the relevant second order belief is true via the relevant meta-competence. So, it is hard to fill out the notion of normalcy or appropriateness in the required way.
Let us then consider the suggestion that the following claim is true of Kate but not Joe: in Good, Kate has the relevant meta-competence and is in appropriate conditions, but too easily would exercise of that competence in those conditions have issued in false belief. One could establish this claim in two distinct ways. First, one could show that, in the bad scenario, Kate but not Joe retains the relevant meta-competence. Alternatively, even if in the bad scenario, Joe and Kate retain the relevant meta-competence, one could argue that, in that scenario, only Kate is in appropriate conditions for the relevant competence (111). It is difficult to flesh out the second suggestion since in both bad scenarios, the subject seems to be in inappropriate conditions for the exercise of the relevant meta-competence. Just as dreaming seems to constitute an inappropriate condition for telling the colours of nearby surfaces by vision, so does having a jokester manipulate the light. So it seems that Sosa must use the first option, the suggestion that, in the bad scenario, Joe but not Kate loses the relevant meta-competence. Now, it is not at first intuitively obvious that one loses competences when dreaming. For instance, when dreaming I do not lose my competence to play the violin, ride a bicycle, or use a computer. Of course, when dreaming, we may make false estimates of our competences. But, when the jokester is playing with the light, Kate would make a false estimate of her competence in telling colours by vision. So, it is hard to distinguish Kate and Joe in this way. A more intuitive thought seems to be that, although I retain my competences when dreaming, dreaming interferes with the exercise of them. We have already rejected one way to make out this idea, namely that when dreaming, one is in inappropriate conditions for the exercise of one’s competence. A different way to fill out the idea is to appeal to the imagination model to claim that, when asleep, one cannot form beliefs. But, Sosa intended his aptness reply to be independent of the imagination model. An alternative suggestion made by Sosa is that, in the bad scenario for the dreamer, but not the kaleidoscope perceiver, there would be “signs to the contrary” (111). Now, it is beyond the scope of this paper to determine whether there are signs that one is dreaming when one is. However, at this point in the dialectic, it is odd for Sosa to make the suggestion. If there were tell-tale signs of dreaming, Sosa could have provided a much quicker response to dreaming scepticism, without appeal to the imagination model or the notion of aptness. Dreaming scepticism seems philosophically powerful precisely because it is supposed that a dream may be subjectively indistinguishable from being awake. If there are telltale signs that one is dreaming, then dreaming scepticism poses much less of a philosophical problem.
In conclusion, the best options for defending Sosa’s safety account may be rather different than he himself takes them to be. He takes it that he has two independent replies to the problem of dreaming scepticism, the imagination model and the claim that knowledge requires aptness not safety. However, we have seen that unless Sosa rejects the widely held assumption that dreaming may be subjectively indistinguishable from being awake, the best way to defend the aptness idea is by appeal to the imagination model. So, the imagination model may play a more central role in Sosa’s safety account that he himself thinks.
Thanks for useful comments and discussion from Jonathan Ichikawa, Duncan Pritchard, Jason Stanley, as well as colleagues at St Andrews, especially Martin Smith, Elia Zardini, Crispin Wright.