Doris and Stich (2005) 'As a Matter of Fact: Empirical Perspectives on Ethics' puts forward the case for experimental philosophy. Below, I reproduce my comments from a past (off-blog) discussion.
* * *
Insofar as philosophers are concerned with non-contingent matters, it seems that a priori analysis should suffice. Consider the internalist's question: is amoralism possible? We already have the challenge from Hume's imagined "sensible knave" -- what difference do real life psychopaths make? In either case, it seems like a question for conceptual analysis: given such and such a scenario, how are we to describe it? (Does the knave/psychopath really form moral judgments, or merely schmoral ones?)
Now, a central argument of the paper criticizes conceptual analysis on the grounds that empirical work (presumably: the "vignette" method favoured by the new "experimental philosophy" movement) is required to uncover *real* folk concepts. But this doesn't do justice to the normative element of analysis. They write:
Smith can reply that responses like those Nichols reports would not be part of the maximally consistent set of platitudes that people would endorse after due reflection. But this too is an empirical claim... (p.125)
How is this an empirical claim? What people would conclude on ideal reflection depends on what propositions are maximally coherent, etc. There's no experiment we can do to pin down what this is; any amount of actual reflection by third parties can always be rebuffed as insufficient to reach the ideal end-point -- "those participants," one may claim, "have not undergone *due* reflection." Maybe they've reasoned badly. The only way we can judge this is to engage in normative reasoning ourselves, and see whether the participants' answers correspond to what we've already determined to be true from the armchair!
Similar issues arise for the problem of persistent moral disagreement. The authors write:
the argument from disagreement cannot be evaluated by a priori philosophical means alone; what's needed, as Loeb observes, is 'a great deal of further empirical research into the circumstances and beliefs of various cultures'.
But I can't see how that would help, if in the end we can only judge others' rationality according to the substantive conclusions that they reach.
Besides, we should (in principle if not in practice) be able to tell from the armchair whether one position or another is rationally necessitated. Simply imagine all the conceivable cultural disagreements, and the difficulty of adjudicating between them. (Which ones obtain in the actual world seems quite irrelevant.) If we ultimately find a conclusion to be rationally necessitated after all, and others disagree (without providing any new reasons, since - ex hypothesi - we've already considered them all in reaching our previous conclusion), then that simply shows that they haven't engaged in fully ideal reasoning yet.
Matters are different in practice, of course, due to our own fallibility. Arguably, our credence in philosophical claims should be informed by the empirical (meta-)evidence provided by others' judgments of the issue. But can we ever hope to scientifically measure the rationality of their judgments, on purely procedural grounds (i.e. without begging the substantive question at hand)? If not, we may find that the real adjudicating work must still be done from the armchair.