Using Genetic Algorithms in Economic Modelling: The Many-Agents Approach

Abstract

The theory of economies with many agents was developed in the sixties, stimulated by two papers of Aumann (Aumann 1964, Aumann 1966). The mathematical properties of this approach have been analysed by Dierker (Dierker 1974). According to Dierker the theory of economies with many agents allows removing both the convexity and the rationality hypothesis in economics. The idea of replacing the rationality hypothesis by the theorem of “bounded rationality”, has been the center of more recent work in economics (Schelling 1978, Arthur 1991). As it is possible to map the framework of the theory of economies with many agents directly onto PBE-automata, Principle-based engineering offers interesting routes to explore in the area of economics; especially in Economic Modelling. Principle-Based Engineering (PBE) is the process of designing a system through the definition of interacting elements where each element is an example of a principle (Addis 1991). Higher order systems can be constructed from principles that have, in their turn, been engineered as principle-based sub-systems. The behaviour of a principle-based system (PBS) is “emergent” through the resultant interaction of the elements and with an ‘environment’.

Using the PBE approach, economic theorist’s might directly model the behavior of many economic agents with partially conflicting interests and therefore gain insights in the results of the interdependencies of these agents for complete markets or market segments etc. This type of work is currently being actively researched at Brainware in the framework of the ESPRIT project PAPAGENA.