Abstract: What makes a belief-desire interpretation of an agent the correct interpretation? One tradition says that the correct interpretation must (means-end) rationalize the agent’s dispositions to choose, and depict them as responding in structurally rational ways to their evidence. Formalizing: we insist that according to the correct interpretation, their choices must (so far as possible) maximize expected utility, and their degrees of belief update over time by conditionalizing on the evidence they receive.
Many (including me) have argued that even in ideal conditions, the “structural rationality” constraints just given leave correct interpretation underdetermined. So some supplementary criterion must be added. One familiar idea is “truth-maximizing charity”: all else equal, the correct interpretation gives the target true beliefs. In my recent book, I defended the rival proposal that the correct interpretation must make the agent as substantively rational as possible (e.g. as justified, as moral, as prudent as can be).
I relate these models of the metaphysics of interpretation (what correct interpretation is) to models of the epistemology of interpretation (what beliefs and desires one should attribute to others, based on limited information). I sketch a formal model of “self-similarity mindreading” and relate it to a formal model of truth-maximizing metaphysics of interpretation---using recent work on gradational accuracy to make the connection. I sketch a formal model of “simulationist mindreading” and relate it to a formal model of substantive rationality-maximizing metaphysics of interpretation—using work on the connection between normative judgements and planning states to make the connection.