When modelling a particular (real-world) target system many agent-based models are potentially relevant. These can vary is many different ways. Rather than merely trying out a few of the more plausible models experimentally and choosing between them (or just make a subjective choice of one!) an alternative is to specify a set of models (eg an abstraction/aggregation hierarchy) and use a computer to search through this set of models for the best model in some defined sense. What relevant algorithms have been proposed?