Posted by: jadamcarter | 14 May, 2009

Radical Scepticism, Closure and Robust Knowledge

Hi folks, I’m attaching here a rough draft of a paper I’ve written which touches upon some themes from Duncan’s “Radical Scepticism, Epistemic Luck and Epistemic Value.” Comments welcome! Below is a link to the paper–Cheers, Adam,%20closure%20and%20robust%20knowledge.pdf



  1. Okay, first let me see whether I can summarise your paper, in order to highlight any misunderstandings (as I’ve only properly read it through once). So, you’re defending the claim that we know ~SH whilst also trying to compensate for neo-Mooreanism’s main cost (i.e. that the belief in ~SH is only classed as knowledge because of lax requirements for knowledge)–you want to do this by maintaining that our beliefs in everyday propositions actually aspire to something more epistemically robust. First, you look to Sosa to see whether we might say that knowledge of ~SH is animal whilst knowledge of everyday propositions is reflective, but see that attempting to use this to vindicate neo-Mooreanism falls flat. This is because Sosa is committed to the claim that your apt belief in ~SH cannot aspire to reflective knowledge–this leaves us vulnerable to radical scepticism about reflective knowledge, by way of a highly plausible closure principle (CPK+) according to which grades of knowledge are closed over competent deduction (and therefore according to which knowledge of everyday propositions cannot aspire to reflective knowledge either). Next, you investigate whether other standard formulations of virtue epistemology might help us preserve low-grade knowledge of ~SH without thereby implying that we lack more robust knowledge of everyday propositions. However, claiming that only the latter is knowledge consisting in cognitive achievement also leads to a sceptical argument based on CPK+, concluding that knowledge of everyday propositions cannot aspire to this specifically valuable sort.

    Unwilling to accept epistemic value scepticism, you reformulate CPK+ so that the value-centred element is made explicit (without assumptions about what generates the relevant value). Here, you argue that the principle fails if ‘knows+’ is read as representing ‘possesses knowledge valuable to a certain extent’, because there are cases in which we competently deduce knowledge of very little epistemic value from knowledge of comparatively high epistemic value. Instead, ‘knows+’ has to represent ‘knows in virtue of meeting some epistemic criterion’, and (given that it clearly cannot be the criterion for reflective knowledge or cognitive achievement) you want to see whether it might be something satisfied by our knowledge of ~SH as *well* as that of everyday propositions (rather than just by knowledge of the latter). It seems the only way to do this whilst still maintaining that everyday knowledge is more epistemically robust is to avoid claiming that the EC consists in multiple criteria which can be satisfied or not satisfied (more of which are satisfied by knowledge of everyday propositions than of ~SH), and instead say that there is merely one EC which itself can be satisfied more or less robustly (and which is satisfied more robustly by knowledge of everyday propositions than of ~SH).

    Your suggested criterion is your formulation of RVE, according to which a belief counts as knowledge iff it’s creditable to the agent’s cognitive abilities that their belief is true (where it’s creditable when it depends on their ability *moreso* than luck, and it depends on ability moreso than luck ‘just in case worlds in which S’s abilities employed in the actual world couldn’t easily have led S to a false belief about whether or not p are closer to the actual world than words where, holding these abilities fixed, S forms a false belief about whether or not p’). So the most robust knowledge will be had when worlds in which S’s abilities employed in the actual world couldn’t easily have led S to a false belief about whether p are very close, whilst worlds where S forms a false belief about whether or not p are very far off (and the least robust knowledge is had when these two sets of worlds are as close as is possible while the former is still slightly closer than the latter).

    You show that our knowledge of ~SH depends on ability moreso than luck by explaining that nearby worlds will be ones where the competence whereby we take conditions to be normal couldn’t easily have led to a false belief about whether or not ~SH (because worlds where you *could* easily have believed falsely are far-off ones in which, e.g., demons are arsing around in the vicinity). To show that knowledge of ~SH depends only *marginally* moreso on ability than luck (and is thus low-grade), you highlight that worlds where one forms a false belief about whether or not ~SH are not as incredibly far off as we might initially suppose–falsity of the belief can owe to, for example, drug use or poor philosophical insight causing you to think it wouldn’t be abnormal for SH to obtain. Finally, you claim that knowledge of everyday propositions is especially robust because worlds where we couldn’t easily be mistaken about, e.g., the lights having gone off (a belief owing to a highly accurate visual competence) are close to the actual world–holding fixed our skills, we can only be easily mistaken about this in worlds in which we are duped in some bizarre way (e.g. by black holograms, the sinister possibility of which I must admit I sometimes find myself having deeply troubling nightmares about). Those are far off, so our everyday beliefs depend significantly more on ability than luck. So, in sum: your account allows us to know ~SH, explains why this is low-grade, avoids denying closure, and avoids radical scepticism about high-grade knowledge.

    I have some minor comments. Firstly, just in terms of structure rather than philosophical content, I think the introduction to the paper may benefit from a more explicit indication of the paper’s direction, given the number of dense arguments involved (and the large amount of ground you cover). A few extra sentences briefly indicating the order in which you’ll be treating possible ways to solve the neo-Moorean puzzle (and maybe one just gesturing roughly at your preferred solution) might make a difference to the ease with which what follows can be digested. Alternatively, perhaps giving the sections titles rather than just numbers would accomplish much the same. This stuff is pretty unimportant, however, especially given that it’s a draft paper!

    Secondly, I’m wondering a bit about the Caesar counterexample to the construal of knows+ as ‘possesses knowledge valuable to a certain extent’. I’m not sure about, this, but there seems to be something unsatisfying about the fact the example involves deducing the knowledge that *someone killed Caesar* from the knowledge that *Person X killed Caesar*. Although it’s true that this would be a competent deduction, I’m not sure that we can’t more strongly motivate a rejection of the initial reading of CPK+. This is because you outright state that the deduced belief is something you already knew, and so therefore a belief you had already formed. Two preliminary thoughts about this: (i) I find it hard to imagine all that many circumstances under which a deduction of an already-known belief would take place, and (ii) It seems as though you could only sensibly form the belief that *someone killed Caesar* once during this process of inquiry into Caesar’s death, and you have already done this on your way to finding out who killed Caesar. Deducing something you already know seems odd (i.e. coming to know the same proposition twice, obviously excluding cases of memory loss), in the sense that it seems to somehow involve taking a step backwards in terms of reasoning. However, I think you’re absolutely right to claim that there are cases in which we competently deduce knowledge of very little epistemic value from knowledge of comparatively high epistemic value–I just think you could use a more convincing example of a belief in a pointless truth deduced from a belief in an epistemically valuable truth. One initial suggestion might just be a slight adjustment of the case you have, making it the case that the epistemic agent knows that Caesar was stabbed to death and is deeply curious about the identity of Caesar’s killer and the area in which Caesar was stabbed. You could then use as your counterexample the seemingly competent deduction of the new belief *Caesar bled* from the belief *Person X killed Caesar by stabbing him in the heart*. This seems like a straightforward case of an agent deducing new knowledge possessing little epistemic value from knowledge possessing significant epistemic value to her, and the deduction doesn’t involve the same odd feature of coming to know something that was already known prior to even beginning the inquiry. I’m not saying that your counterexample doesn’t work, rather just that a different one (featuring deduced knowledge of a previously-not-explicitly-known belief) would possibly make the point you want to make more forcefully, by showing that the reading of knows+ as ‘possesses knowledge valuable to a certain extent’ actually fails in pretty commonplace/ordinary examples of competent deduction, when it’s not merely an already-known belief that is produced by the deduction process.

    I had another comment (regarding your argument against the claim that we only form a false belief about whether or not ~SH in far off worlds), but what I have written is so poorly formulated that it don’t want to air it right now! I liked the paper–I’m sympathetic to what you’re doing with neo-Mooreanism, and also sympathetic to your brand of RVE more generally. I enjoyed seeing it applied to problems not covered in your thesis.


%d bloggers like this: