Friday, March 03, 2006
Could a priori knowledge be subject-sensitive?
(Cross-posted at Arche. Apologies for overlap with my previous post here.)
Here's a question that arose out of reading Thomas Crisp's recent Analysis note, 'Hawthorne on Knowledge and Practical Reasoning'.
According to Delia Graff's theory of vagueness ('Shifting Sands', 2000), contextualism gets something important wrong, and something important right. Contra the contextualist, vague predicates like 'heap' and 'bald' express the same property throughout their respective suitably constructed Sorites series. Yet changes in the practical interests of a subject making judgements with such predicates determine shifts in their extensions (shifts not attributable to changes in the relevant comparison class, etc), and the resulting picture of the semantics of such expressions i. invalidates Sorites reasoning, ii. explains why we find such reasoning intuitively compelling, and iii. does not require us to give up classical logic or semantics. So the contextualist has something like the right idea, but the hope is Graff's theory delivers the same package whilst neatly avoiding some familiar problems with contextualist approaches.
Now, while the extension of a vague predicate is sensitive to the practical interests of the relevant subject, there are restrictions on such boundary-shifting; just as admissible sharpenings for a supervaluationist must respect the clear polar cases, Graff suggests that on her view the interests of a subject can never determine that the boundary shifts so that, to take an extreme case, a man with no hairs on his head is in the anti-extension of 'bald'. There's something very reasonable about this requirement, but last summer I remember Carrie, Robbie, and other people around the Vagueness project wondering whether Graff really needs or is entitled to it.
Crisp's note suggested to me that there might be a similar issue for interest-relative invariantism (IRI) about knowledge, as defended in Hawthorne and Stanley's recent monographs. According to IRI, contextualism about knowledge gets something important wrong, and something important right. Contra the contextualist, the property expressed in knowledge ascriptions is invariant. Yet changes in the practical interests of a subject determine shifts in whether or not their true beliefs count as knowledge (even keeping their evidence fixed). So whether or not a subject's true belief counts as knowledge depends on non-epistemic (at least as traditionally conceived) features of their circumstances, such as how high the stakes are regarding the possibility that they're wrong.
The parallel question to the one we might want to ask the boundary-shifter about vagueness is this; could it ever be the case that the consequences of being wrong about the truth of a proposition are so disasterous that knowledge attributions are defeated, despite the proposition enjoying some (putatively) priviledged epistemic status (apriority or analyticity, for example)? Crisp's paper suggests that without some Chisholming, Hawthorne's formulation of IRI does indeed allow this possibility. Is there a neat way to avoid such a committment, or is it just a consequence of IRI (unless we do some special-pleading regarding a priori knowledge)? If it is a genuine committment, is that a particuarly bad result? There certainly seems something somewhat counter-intuitive about it, but are there materials for an objection?
Here's a question that arose out of reading Thomas Crisp's recent Analysis note, 'Hawthorne on Knowledge and Practical Reasoning'.
According to Delia Graff's theory of vagueness ('Shifting Sands', 2000), contextualism gets something important wrong, and something important right. Contra the contextualist, vague predicates like 'heap' and 'bald' express the same property throughout their respective suitably constructed Sorites series. Yet changes in the practical interests of a subject making judgements with such predicates determine shifts in their extensions (shifts not attributable to changes in the relevant comparison class, etc), and the resulting picture of the semantics of such expressions i. invalidates Sorites reasoning, ii. explains why we find such reasoning intuitively compelling, and iii. does not require us to give up classical logic or semantics. So the contextualist has something like the right idea, but the hope is Graff's theory delivers the same package whilst neatly avoiding some familiar problems with contextualist approaches.
Now, while the extension of a vague predicate is sensitive to the practical interests of the relevant subject, there are restrictions on such boundary-shifting; just as admissible sharpenings for a supervaluationist must respect the clear polar cases, Graff suggests that on her view the interests of a subject can never determine that the boundary shifts so that, to take an extreme case, a man with no hairs on his head is in the anti-extension of 'bald'. There's something very reasonable about this requirement, but last summer I remember Carrie, Robbie, and other people around the Vagueness project wondering whether Graff really needs or is entitled to it.
Crisp's note suggested to me that there might be a similar issue for interest-relative invariantism (IRI) about knowledge, as defended in Hawthorne and Stanley's recent monographs. According to IRI, contextualism about knowledge gets something important wrong, and something important right. Contra the contextualist, the property expressed in knowledge ascriptions is invariant. Yet changes in the practical interests of a subject determine shifts in whether or not their true beliefs count as knowledge (even keeping their evidence fixed). So whether or not a subject's true belief counts as knowledge depends on non-epistemic (at least as traditionally conceived) features of their circumstances, such as how high the stakes are regarding the possibility that they're wrong.
The parallel question to the one we might want to ask the boundary-shifter about vagueness is this; could it ever be the case that the consequences of being wrong about the truth of a proposition are so disasterous that knowledge attributions are defeated, despite the proposition enjoying some (putatively) priviledged epistemic status (apriority or analyticity, for example)? Crisp's paper suggests that without some Chisholming, Hawthorne's formulation of IRI does indeed allow this possibility. Is there a neat way to avoid such a committment, or is it just a consequence of IRI (unless we do some special-pleading regarding a priori knowledge)? If it is a genuine committment, is that a particuarly bad result? There certainly seems something somewhat counter-intuitive about it, but are there materials for an objection?
Labels: Epistemology, Vagueness
# posted by Aidan @ 1:08 PM 
