Three challenges to Chalmers on computational implementation

The notion of computational implementation is foundational to modern scientific practice, and in particular, to explanation in cognitive science. However, there is remarkably little in the way of theoretical understanding of what computational implementation involves. In a series of papers, David Chalmers has given one of our most influential and thorough accounts of computational implementation (Chalmers 1995, 1996, 2012). In this paper, I do three things. First, I outline three important desiderata that an adequate account of computational implementation should meet. Second, I analyse Chalmers’ theory of computational implementation and how it attempts to meet these desiderata. Third, I argue that despite its virtues, Chalmers’ account has three shortcomings. I argue that Chalmers’ account is (i) not sufficiently general; (ii) leaves certain key relations unclear; (iii) does not block the triviality arguments.