Robot predators in virtual ecologies: the importance of memory in mimicry studies

Abstract

An important means of investigating gains and losses to prey caused by mimicry is through mathematical or computer constructs which represent and explore limited aspects of mimicry situations. Such studies use virtual predators which are usually simple automata, ‘robots’ that, through simple rules, vary virtual attack rates on virtual insect prey. In this paper I consider the effect of variations in predator memory and learning on mimicry dynamics. When there is mimicry between unequally noxious prey, the way that memory is modelled is shown to be crucial. If forgetting rates are fixed, an increase in the density of the least defended prey produces monotonic gains or losses in protection. However, if forgetting rate is inversely related in some way to degree of noxiousness of the prey then attack rates initially rise with the density of the least defended prey, reach a cusp and then fall. I show that the generation of this highly unconventional up–down result appears to be independent of variations in learning rate. This work shows how sensitive the predictions of virtual predators may be to relatively small changes in behavioural rules.