Are Artificial Mutation Biases Unnatural?

Whilst the rate at which mutations occur in artiicial evolutionary systems has received considerable attention, there has been little analysis of the mutation operators themselves. Here attention is drawn to the possibility that inherent biases within such operators might arte-factually aaect the direction of evolutionary change. Biases associated with several mutation operators are detailed and attempts to alleviate them are discussed. Natural evolution is then shown to be subject to analogous mutation \biases". These tendencies are explicable in terms of (i) selection pressure for low mutation rates, and (ii) selection pressure to avoid parenting non-viable oospring. It is concluded that attempts to eradicate mutation biases from artiicial evolutionary systems may lead to evolutionary dynamics that are more unnatural, rather than less. Only through increased awareness of the character of mutation biases, and analyses of our models' sensitivity to them, can we guard against artefactual results. This paper explores the potential for artefactual evolutionary simulation results to derive from biases inherent within the artiicial mutation operators they employ. As an example, consider a recent coevolutionary simulation model which has suggested that signals exhibiting complex symmetry could evolve merely as a side eeect of selection for distinctiveness 1]. The model involved multicoloured, composite patterns coevolving with simple artiicial neural networks under a mutualist selection regime. Networks which were able to discriminate signal patterns from distractor patterns were favoured, as were discriminable signal patterns. This discrimination had to be achieved despite patterns being presented in various orientations and positions on each net-work's artiicial \retina". After a period of simulated coevolution, the authors report that networks were able to distinguish signals from distractors almost perfectly, and that the coevolved signal patterns displayed \marked symme-tries" (p. 171). The authors note that, like many natural displays, the evolved signals consisted of \purer, brighter colours" (p. 171) than average signals. An evolutionary-functional account for the evolved symmetry was proposed { symmetrical signals persist because they are invariant under the various transformations involved in their presentation, and hence easier to discriminate from distractors. The evolved signals' bold coloration was explained as the result of selection pressure to diverge from random distractor patterns.