Hidden Variables in Knowledge Representation

It is a truism that knowledge representation is about choosing the right level of granularity. Granularity is intimately connected with abstraction, component interaction, historical dependency, and similar systems-theoretic properties. As every choice must elide something, unless the purpose to which a representation is carefully restricted, there will be predictive, explanatory and similar inadequacies. In order to understand and analyze such inadequacies a useful approach is to examine the role of what might be collectively called “hidden variables”, the entities that are elided. These can range from properties omitted from a representation to those which have been aggregated into an abstract property. The conflation of finite-memory dependencies into Markov processes is another example. This talk will survey the effects of hidden variables in knowledge representation and indicate techniques for reasoning about them.

Authors and Affiliations

1. Artificial Intelligence Laboratory, School of Computer Science and Engineering, University of New South Wales, Sydney, NSW, 2052

   Norman Y. Foo

Editors and Affiliations

1. School of Information Science and Technology Department of Information and Communication Engineering, University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo, 113-8656, Japan

   Mitsuru Ishizuka

2. School of Information Technology Knowledge Representation and Reasoning Unit (KRRU) Faculty of Engineering and Information Technology, Griffith University, PMB 50 Gold Coast Mail Centre, Queensland, 9726, Australia

   Abdul Sattar

Reprints and permissions

© 2002 Springer-Verlag Berlin Heidelberg

Policies and ethics