Today my research group in Groningen (with the illustrious online participation of Tony Booth, beaming in from the UK) held a seminar session where we discussed Fabienne Peter’s 2013 paper ‘The procedural epistemic value of deliberation’. It is a very interesting paper, which defends the view that deliberation has not only epistemic value (as opposed to ‘merely’ ethical, practical value), but also that it has procedural epistemic value (as the title suggests), as opposed to ‘merely’ instrumental value. I'll argue here that I agree with the thesis, but not for the reasons offered by Peter in her paper.
The paper begins with the following observation:
An important question one can ask about collective deliberation is whether it increases or decreases the accuracy of the beliefs of the participants. But this instrumental approach, which only looks at the outcome of deliberation, does not exhaustively account for the epistemic value that deliberation might have. (Peters 2013, 1253)
One way to spell out this idea is the following: suppose there were two knowledge-producing procedures with the exact same accuracy, i.e. which would produce the same amount of true beliefs and avoid the same amount of false beliefs. Moreover, procedure D involves deliberation, while procedure O relies entirely on an oracle, for example. If we can show that procedure D is superior to procedure O on purely epistemic grounds, then we can establish that deliberation has procedural epistemic value, rather than merely instrumental epistemic value (i.e. increase accuracy).
Peter resorts to Dawall’s second-person standpoint in ethics to argue for this thesis, in particular thorough the notion of mutual accountability between epistemic peers. She defines an epistemic peer as someone who is equally likely as you to make a mistake in a particular situation. (The notion of an epistemic peer is of course an idealization, as it presupposes someone with the exact same access to evidence, the exact same cognitive capacities etc.) The question then becomes how to respond to disagreement between two epistemic peers – disagreement of the kind that may lead to deliberation. One idea she defends in the paper, and which I find quite appealing, is that the mere fact that an epistemic peer disagrees with you on a given matter, despite the fact that you have access to the same epistemic resources (including evidence), should give you reason enough to revisit your own beliefs. This follows simply from the fact that we are fallible epistemic agents, and so it would be irresponsible not to reconsider one’s reasons for holding a specific belief when confronted with the opposite belief coming from someone with equal epistemic standing (even if only to maintain your original beliefs after inspection).
After discussing the paper, I came to the conclusion that I very much agree with the main claim – that deliberation has procedural epistemic value – and that the second-person standpoint is imminently suitable for discussions in (social) epistemology. However, I was unconvinced by Peter's strategy to argue for the thesis. As I see it, the main problem with it seems to be that she still takes accuracy to be “the only intrinsic epistemic value” (p. 1264). In this case, it becomes difficult to argue for the fact that deliberation has epistemic value that is not instrumental, if accuracy is the only intrinsic epistemic value: if it is epistemic, it is not procedural; if it is procedural and thus not instrumental, then it is not epistemic. In other words, being procedural and epistemic at once seems to become an impossible place in the logical space of possibilities once accuracy is viewed as the only (intrinsic) epistemic value.
I want to propose another way in which deliberation can be understood as having intrinsic epistemic procedural value. Let us go back to my two knowledge-producing procedures D and O, D based on deliberation and O based on an oracle (one can even add a Cassandra-like twist to the second one and stipulate that the oracle method is highly accurate, but no one will believe it). They may well be equally accurate, but only one of them will satisfy what I take to be another key intrinsic epistemic value (along with accuracy): transferability. This is a concept I borrow from a paper by Kenny Easwaran, which he introduces to offer a diagnosis of what is wrong with probabilistic proofs in mathematics: they are not reproducible by – transferrable to – another agent, as there is no guarantee that the original proponent of the proof did not cherry-pick the witnesses:
… the basic idea is that a proof must be such that a relevant expert will become convinced of the truth of the conclusion of the proof just by consideration of each of the steps in the proof. With non-transferable proofs, something extra beyond just the steps in the proof is needed—in the case of probabilistic proofs, this extra component is a knowledge of the process by which the proof was generated, and in particular that the supposedly random steps really were random.
Transferability is arguably one of the key concepts in science across the board: in the empirical sciences it becomes the desideratum of replicability; in mathematics it is the desideratum that a mathematical proof be reproducible by a minimally competent opponent (as I’ve argued before). In philosophy and other fields, transferability takes the form of arguments, reasons, which can be offered for one’s views so as to (hopefully!) convince others of their cogency. More generally, transferability pertains to offering to another agent the reasons one has to maintain certain views; it is thus an inherently second-person notion, in the spirit of Rawls’ notion of justification. (Naturally, by focusing on transferability, I am explicitly adopting a social conception of (scientific) knowledge.)
And so, I want to suggest that procedure D, which involves deliberation, will produce knowledge that is not only accurate (or as accurate as in procedure O), but which is moreover transferable in a way that knowledge produced by O is not. This is so precisely for procedural reasons, i.e. because along the way arguments for each of the sides of the dispute are produced which can then be offered to transfer the piece of knowledge in question to other agents. And thus, if transferability is indeed a purely epistemic value (which I admit is a claim some people may want to dispute), then by increasing transferability, deliberation will also increase epistemic value, even if not increasing accuracy per se (though it may well also increase accuracy).
These are all very incipient thoughts, which I began to gestate only a few hours ago, and so I’d be interested in hearing what people have to say – in other words, let us deliberate!
(And to tie this discussion to the current debates on Brian Leiter’s behavior: one of the moves he often resorts to when faced with people who disagree with him is to disqualify them as epistemic peers by questioning their credentials and credibility. For example, when Carolyn Dicey Jennings presented an alternative ranking based on placement data, he basically suggested that she did not know what she was talking about: “This does raise a serious question about her judgment”. This is a typical silencing maneuver: if you are not my epistemic peer, I have no reason to take your disagreement with me seriously at all. Recall also my post on the ‘This is nonsense’ objection.)
Recent Comments