Hawks, Doves and Lifetime Reproductive Fitness Philip Hingston 1 and Luigi Barone 2 1 School of Computer and Information Science, Edith Cowan University, Mt Lawley, 6020, Australia p.hingston@ecu.edu.au 2 School of Computer Science and Software Engineering, The University of Western Australia, Crawley, 6009, Australia luigi@csse.uwa.edu.au Abstract. Evolutionary game theory is an adaptation of classical game theory used to analyse and explain observed natural phenomena where organisms choose between alternative behavioural strategies. Recent analyses and simula- tions of evolutionary games have shown that implementation choices such as population size and selection method can have unexpected effects on the course of evolution. In this paper, we propose a new evolutionary interpretation of such games that uses a more biologically faithful selection scheme, in which se- lection choices and population size emerge from the interactions of the players and their environment. Using the well-known Hawks and Doves game as an ex- ample, we show that the resulting models are also tractable, easily simulated, and flexible. 1 Introduction Evolutionary game theory ([1]) is an adaptation of classical game theory used to ana- lyse and explain observed natural phenomena where organisms choose between alter- native behavioural strategies. In recent years, simulation studies based on co- evolutionary algorithms have been applied to such scenarios. One feature of these simulations is an explicit fitness value for each organism, which is used in the fitness proportionate selection mechanism of the co-evolutionary algorithm. In nature, there is no such explicit fitness value. Individual organisms are born, interact with each other and with their environment, reproduce and die, and their lifetime reproductive fitness emerges out of this complex system - it depends upon how long the organism lives, how frequently it reproduces, and the reproductive suc- cess of its offspring. These in turn depend on the behaviour of the organism, as well as on factors outside its control, such as the number of other organisms in the population and their behaviours. The same phenotype could have different reproductive success at different times, depending on such factors. Standard analytical methods also posit an explicit fitness value, and a selection mechanism based on this fitness value, as is used in the simulation models (giving rise to the so-called replicator dynamics). At first it might seem that using an explicit fitness value is a reasonable simplification that facilitates analysis. If that were the