The argument of this paper is that, in the present state of knowledge, it is not good science to base integrated assessment models on any social theory and it is not appropriate to evaluate integrated assessments in relation to conventional positivist precepts. Instead, means should be sought to produce well validated descriptions of social and natural systems. A powerful means of validation turns on the application of agent based models. The agents are self-contained computer programs that interact with one another and can be designed and implemented to describe the behaviour and modes of interaction of observed social entities (individuals or groups). Reported models and simulation experiments demonstrate a procedure for the qualitative and statistical validation of agent based integrated assessment models
