1) Can be valuable for exploring scenarios and situations for which we do not have, and perhaps cannot expect to obtain, real data 2) Can be valuable for summarizing our current state of knowledge and generating predictions in which the connection between current knowledge, assumptions and predictions is explicit and clear 3) In order to be valuable in these ways, a model does not have to be a full and perfect description of the real world it seeks to mimic, all models incorporate approximations1 4) caution is therefore always necessary – all conclusions and predictions are provisional and can be no better than the knowledge and assumptions on which they are based – but applied cautiously they can be useful 5) a model is inevitably applied with much more confidence once it has received support from real sets of data