I consider agent-based simulation as a tool for formalising explicitly our models. I think abstract simulations with relatively few parameters are great platforms for exploring the consequences of our theories and for building quantifiable expectations that we can compare with real-world empirical data. Currently I’m working on a model of human metapopulation dynamics for my doctoral thesis where I explore the consequences of group fission and fusion. I have recently developed, with Mark Lake, a simulation of cultural transmission with frequency dependent adaptive traits [5]. I use different platform for developing my models, mostly in R, NetLogo and RePast.

Top: Parameter space showing the different effects of increased reliance in model biased transmission (parameter z). The model assumes that each trait has a normally distributed yield with variance v, and new traits are adopted through innovation (with rate μ) and model biassed transmission. The box-plots shows the distribution of the terminal median trait among the agents, indicating how the rate of cultural evolution (i.e. values in the y-axis) is slowed down when the dependence on model based transmission(parameter z) is too high or too low. (from). [5]