Luck, Michael Mordechai; (1993) Motivated Inductive Discovery. Doctoral thesis (Ph.D), University of London.

Text Luck_10091005_thesis.pdf Download (6MB)

Abstract

Research in machine discovery to date has tended to concentrate on the replication of particular episodes in the history of science, and more recently on the extraction of regularities from large databases. In this respect, current models of induction and discovery concentrate solely on the acquisition of knowledge, and lack the flexibility of reasoning that is necessary in a real-world changing environment. Against this backdrop, this dissertation addresses inductive reasoning, specifically based around the scientific discovery paradigm. A framework for inductive reasoning is presented which includes the six stages of prediction, experimentation, observation, evaluation, revision and selection. Within this framework, different kinds of inductive reasoning can be reduced to the same basic component processes. The difference between the various kinds of reasoning arises not through the use of different mechanisms, but through the influence of motivations which bias the application of these mechanisms accordingly. Also within this framework, a model and its implementation as a computer program, the MID system, have been developed, concentrating primarily on the internal stages of the framework, prediction, evaluation, revision and selection. The role of motivations in allowing reasoning for both knowledge and action is investigated and implemented in the program. By choosing different internal models of motivation for reasoning systems, different behaviours can be achieved from the same basic architecture. The MID system reasons in simple physical domains, both for knowledge and for action. It demonstrates how a basic mechanism can be used to provide an effective means for reasoning in a variety of contexts, and also how a simple motivational representation can be used as an effective control strategy.

Download activity - last month

Export as JSONXMLCSV

Download activity - last 12 months

Export as CSVXMLJSON

Downloads by country - last 12 months

Export as JSONXMLCSV

Archive Staff Only

View Item