Neighbourhood, Geometry and Initial Conditional Effects on Fairness and Agent Longevity in Large-Scale SugarScape Models

Abstract

The agent-based SugarScape model of Epstein and Axtell has attracted interest and use in a number of sociological contexts. The model is usually configured on a regular square lattice of 51x51 cells. However, in this paper we explore how the model scales to much larger lattices and how it behaves with a range of different neighbourhood connectivity sizes and geometries including hexagonal, triangular, Moore and radial-proximity neighbourhoods. We compare the popular twin sugar peak initial model configuration with that of a flat sugar-field and use a number of metrics including the Gini coefficient of income distribution and other agent statistics to discuss the quantitative potential for this model on large scale systems. We focus on use of a single agent species type interacting solely with the spatial sugar-field. We report on how initial conditions dramatically affect the inequalities in agent opportunity and how this drives the model to longer agent population survival times.