Models for the growth of runaway systems such as cancerous systems can be incorporated into animat agent systems. Large systems can be used to simulate macroscopic emergent spatial patterns of growth and behaviour. We report on experiments involving animats with a runaway success factor introduced into a spatial agent system that would otherwise maintain a dynamical equilibrium around stable mean population values. The "selfish" runaway element can bring the whole population system to collapse using microscopic rules that might easily arise by chance mutation. We compare our predator-prey model with traditional epidemic and growth models such as the Eden model and SIR model and find our model exhibits a growth exponent of $\sim 0.65$, intermediate between these two other models, on a square mesh.
Keywords: stochastic modelling; agent based models; artificially intelligent animats.
Full Document Text: PDF version.