Discussion papers

CPM-99-51 - 15 March 1999

The Cost of Rational Agency

Scott Moss

Abstract

The rational agency assumption limits systems to domains of application that have never been observed. Moreover, representing agents as being rational in the sense of maximising utility subject to some well specified constraints renders software systems virtiually unscalable. These properties of the rational agency assumption are shown to be unnecessary in representations or analogies of markets. The demonstration starts with an analysis of how the rational agency assumption limits the applicability and scalability of the IBM information filetering economy. An unrestricted specification of the information filtering economy is developed from an analysis of the properties of markets as systems and the implementation of a model based on intelligent agents. This extended information filtering economy modelis used to test the analytical results on the scope for agents to act as intermediaries between human users and information sources.

Accessible as HTML

Downloads