Abstract

In the philosophy of mind, the controversy between realists and antirealists often concerns the logical form of sentences embedded in attitude reports. Antirealists believe that such sentences refer to psychological states; realists believe that they refer to situations or states of the world. In this essay, it is shown how these two modes of semantic representation are associated with different approaches to the computational modeling of cognitive processes. I put forward a normative account of methodology in artificial intelligence that reconciles realism and antirealism, and suggest that my account sheds light on the problem of representing the logical form of attitude reports