Toward Computational Animal Cognition in Behavioral Informatics

Abstract

After watching a videotape of an animal's "continuous" behavioral process, how would you tell a story on how its decisions are made and what information has been indeed used? Furthermore, based on the videotape, how would you generate a "realistic" animation film in your computer with a rather short computer program? To answer the above questions, I would show the necessity of working on computational animal cognition to decode the spatial-temporal patterns in information processing underlying animal cognitive processing. Along my argument, a series of analyses would be given based on increasing detailed behavioral data recorded in our laboratory. Also cognitive issues concerned and studied by my group together with related statistical problems would be discussed. In this talk, partial answers to the above questions are given by a brand new way of building algorithmic behavioral model derived via sparse coding on coupled digital sequences encoded from continuous time recording data. At the end the computational implementation of inductive systems of Holland, Holyoak, Nisbett and Thagard (1989) would be briefly mentioned.