Considerations in designing a cybernetic simple 'learning' model; and an overview of the problem of modelling learning

Abstract

Learning is viewed as a central feature of living systems and must be manifested in any artifact that claims to exhibit general intelligence. The central aims of the thesis are twofold: (1) - To review and critically assess the empirical and theoretical aspects of learning as have been addressed in a multitude of disciplines, with the aim of extracting fundamental features and elements. (2) - To develop a more systematic approach to the cybernetic modelling of learning than has been achieved hitherto. In pursuit of aim (1) above the following discussions are included: Historical and Philosophical backgrounds; Natural learning, both physiological and psychological aspects; Hierarchies of learning identified in the evolutionary, functional and developmental senses; An extensive section on the general problem of modelling of learning and the formal tools, is included as a link between aims (1) and (2). Following this a systematic and historically oriented study of cybernetic and other related approaches to the problem of modelling of learning is presented. This then leads to the development of a state-of-the-art general purpose experimental cybernetic learning model. The programming and use of this model is also fully described, including an elaborate scheme for the manifestation of simple learning.