Real algorithms : a defense of cognitivism
Part of : Philosophical inquiry ; Vol.XX, No.3-4, 1998, pages 41-58
Issue:
Pages:
41-58
Author:
Abstract:
The sciences are exclusively concerned with intrinsic features of reality such as mass and shape, not with observer-relative features such as goodness and beauty. Hence, in order for cognitivism to be a scientific hypothesis, it must be coherent to attribute intrinsic computational features to some unconscious processes. Hence cognitivism requires that some unconscious processes be intrinsically algorithmic, but on the face of it all processes are algorithmic in the trivial sense of behaving in agreement with some algorithm or other. This either threatens to make algorithmicity observer-relative or to make all processes algorithmic. Neither possibility is acceptable to the cognitivist.In order to avoid pan-algorithmicity and observer-relative algorithmicity, one could appeal to teleology. The only truly algorithmic processes, on this proposal, are those which occur in order to achieve the relevant algorithm’s end-state. However, even though this may avoid the Scylla of panalgorithmicity, it falls to the Charybdis of observer-relative algorithmicity - at least if Searle is correct. For Searle claims that such teleological notions as function imply value judgments and, hence, do not correspond to intrinsic features of reality.Bedau’s thought experiments support Searle’s claim, but one must bear in mind that Searle and Bedau are analyzing our pre-scientific folk concept of function. That normative concept could conceivably be replaced by a purely descriptive concept capable of all the explanatory work of the old. In fact, Wright’s theory of functions, which was originally meant as an analysis of our vernacular concept, serves quite well as a scientifically acceptable replacement for that vernacular, normative concept. Wright’s non-normative construal of functions, what I call "functions(d)," can be used as an objective criterion for distinguishing algorithmic from non-algorithmic processes.But appeal to intrinsic algorithmicity is only a partial response to Searle. In order for a process to be intrinsically computational, it must not only be intrinsically algorithmic but intrinsically semantic as well. Therefore, a naturalization of semantic content is also essential for the cognitivist project. In this paper, I have only addresssed the issue of naturalizing algorithmicity, leaving the naturalization of content to others.
Subject:
Subject (LC):
