In a paper recently published in PLoS ONE, we proposed that model breaking could be attempted by looking for the different behaviours the model can exhibit, and in particular for the unexpected ones. Those unexpected behaviours, if they differ from what is being observed in nature, reveal weaknesses of the model and give the modeller an opportunity to revise it. By reiterating this process, we can build models in which we have a better confidence. The unexpected behaviours can also constitute predictions of the model that can be tested experimentally.
I studied cognitive science and artificial intelligence in France. I did a Ph.D. on the topic of emergence in complex systems models and simulation, and continued with a post-doc where I used evolutionary algorithms to design complex systems simulation experiments.
Entries by cherel