Discussion of Forthcoming Term Paper

For my proseminar paper, I'm thinking of drawing parallels between the generality problem for reliabilist theories of justification and the problem of action description in Kantian ethics, and I'm going to suggest that proposed solutions to the latter can be adapted to address the former.

Briefly, the generality problem for reliabilism is this: according to that theory, a belief-forming process confers justification on a belief token that is produced by that process only insofar as the process is reliable. But the degree to which a belief-forming process can be described as reliable seems to depend in part on the level of generality used in describing the process. For example, take the case of a belief "There is a tree outside my window" that is the result of a process of visual sense perception (for simplicity's sake I'll only consider the visual aspect of the seeing, and exclude other relevant processes such as the application of the concept 'tree'). If we describe this process as "visual sense perception" simpliciter, then it seems as though the process is not very reliable (compare e.g. my seeing the tree at night and without the aid of my glasses with my seeing the tree in the day with my glasses on), but if we describe the process in a very fine-grained way (perhaps so fine as to ensure that only that particular act of seeing fulfills the description) then the reliability of the process will be much higher. The problem is to give a criterion for determining whether and to what degree a process is reliable in a non-ad hoc way.

Briefly, the problem of action description in Kantian ethics is this: if we try to act in accordance with the categorical imperative (specifically, the universal law formulation), then whether an action is permissible seems to depend in some respect on the level of generality used to describe it. Take the paradigmatic case of the person who is questioned by the Gestapo and who lies about hiding Jews: if we describe the action as "lying" simpliciter, then it seems clearly wrong, since we could not consistently will that everyone lie. But, if we describe the action as "lying to an evil person in order to save the lives of innocent people", then it seems like we could consistently will such an action (I'm ignoring some potentially salient features of the example for the sake of simplicity, but I think the general idea is clear).

In both cases, we are presented with the problem of determining the appropriate level of generality to use in describing an action, as well as the implications that such a description has on justification (epistemic in one case and moral in the other). Thus, I think it is reasonable to suppose that progress in finding a solution to one problem can be applied to finding a solution to the other.