The Сhurch-Turing thesis gives us a definite answer to the question What is computation? only as far as convenient mathematical abstractions and idealisations are assumed; generally, it does not allow one to say whether a given physical process qualifies as computation or not. The author defends a view according to which the symbolic character of computation is an essential feature of this concept. A concept of naturalised computation cannot reduce to a distinguished class of natural processes, which qualify as models of computation in some sense, but should also include some notions of epistemic agency and symbolic representation. Since representation is a basic cognitive function, and since symbolic and linguistic representations play a major role in human societies, a naturalised account of computation should not bypass cognitive and social sciences. An important role of philosophy with respect to exact sciences is to provide these sciences with new ideas, some of which may later receive a scientific justification and bring about important scientific developments. However, at least since Kant philosophy also plays a critical role with respect to the sciences. The ongoing expansion of digital computational technologies into all areas of human life leads us to the situation where the metaphors of “artificial intelligence” and “world as computer” become to appear self-evident. At this point, a philosophical critic of computational approaches in science becomes particularly pertinent.