Eye on the prize

Abstract

Interactive Evolutionary Computation (IEC) has been applied to art and design problems where the fitness of an individual is at least partially subjective. Applications usually present a population from which the preferred individuals are selected before the usual evolutionary operations are performed to produce the next generation. Large population sizes and numbers of generations impose significant demands on the user. This paper proposes that selecting by means of eye movements could reduce user fatigue without sacrificing quality of fitness assessment. In the first experiment, an eye-tracker is used to capture fixations and confirm the reliability of such a measure of attention as a fitness driver for subjective evaluation such as aesthetic preference. In a second experiment, the robustness and efficiency of this technique is investigated for varying population sizes, presentation durations and levels of fitness sampling. The results and their consequences for future IEC applications are discussed.