Approximating epistemic ideals

Idealized norms and requirements are common in epistemology. However, ordinary agents like us are not capable of meeting such ideals. So, why care about them? Why not simply care about what makes agents like us rational, justified, and the like? A tentative response to this worry roughly goes as follows: Epistemic ideals are relevant because we are in a position to approximate them (Carr 2022; Staffel 2019, 2024; Zynda 1996). In this paper, I analyze three approximation theses. I argue that they raise worries. I pay special attention to the thesis that non-ideal rationality is an approximation of ideal rationality. I argue that this view can’t account for the fact that ideal and non-ideal agents do not always optimize exactly the same thing (or function). Think, for instance, of Simon’s (1976) distinction between substantive and procedural rationality. For Simon, ideally rational agents optimize an outcome directly (like expected accuracy), while boundedly rational agents are more concerned with the optimization of procedures (like heuristics). This dual aspect of optimization causes trouble for approximation claims.