A Problem for Duncanfactuals

Duncan Pritchard's definition of luck is anchored in a modal metric modulo a method (a little alliteration for you there).

(L1) If, an event is lucky, then it is an event that occurs in the actual world but which does not occur in a wide class of the nearest possible worlds where the relevant initial conditions for that event are the same as in the actual world. (p. 128)

He then defines a knowledge-threatening notion of luck:

Veritic EL: It is a matter of luck that the agent's belief is true.

He comments: "the agent's belief is true in the actual world, but...in a wide class of nearby possible worlds in which the relevant initial conditions are the same as in the actual world--and this will mean...that the agent at the very least forms the same belief in the same way as in the actual world..." (146, emphasis added).

I think the holding fixed part needs to be made explicit, so I define a "Duncanfactual" functor "D-->" as follows.

D--> =df In a sufficiently wide class of worlds in which S believes p via M (where M is the method by which S comes to believe p in the actual world), S's belief that p is false.

When you seek to know that p, you want the Duncanfactual to be false, for if it's true, then, chances are, you don't know, since you're subject to Veritic Epistemic Luck.

Now I want to raise a concern for Duncanfactuals which Jon Kvanvig raises for other Safety-based theories in his Value of Knowledge book. In a discussion of Sosa Safety beginning on p. 134 of that book, Jon points out that the mere *truth* of the counterfactuals isn't really relevant, for they could be made true or false by a demon in such a way that their truth really had no connection to my cognition.

The reply he considers is that what's really doing the work is the *explanation* of why the counterfactuals turn out why they do. This might go well for virtue theorists like Sosa and Greco, since their views are really views about factors in cognitive agents which explain the truth of the relevant counterfactuals. (Jon then goes on to give counterexamples to Sosa Safety, counterexamples to which Duncan responds and about which I'll comment in another post.)

The point I want to note here is that Duncan explicitly eschews having explanatory factors in the analysis of knowledge.

P. 192: "For whilest the safety-based theorist might no doubt wish to tell an explanatory story about how creatures such a ourselves come to have safe true beliefs that makes essential reference to the cognitive faculties and the epistemic virtues, she does not define knowledge in terms of these cognitive traits as the agent reliabilist does."

I'm not sure this is a good move, because I'm not sure a non-explanatory definition counts as an analysis. Mere extentional equivalence does not a good definition make as Socrates' example of "featherless bipeds" makes clear. When Earl Conee goes through various analyses of knowledge in a survey of epistemology course, he writes on the board for each one "S knows p if and only if and because..." The first time I saw that I didn't understand what was going on grammatically.

The idea is that in an *analysis* of an evaluative concept the "iff" part isn't enough. To distinguish between the analysandum and the analyisans there needs to be some asymmetry. One asymmetry would be that one side possesses an evaluative term and the other does not. But this asymmetry is, I take it, in part because the non-evaluative concepts shed light on the evaluative ones. I don't think it's *just* a matter of honoring a supervenience thesis.

I think there are ways of avoiding this problem, by going functionalist for example, but I don't think that's what Duncan wants to do. It seems to me at this point, then, that there's something important missing in Duncan's definition.