The use of computational models has undergone some dramatic epistemic shifts. Originally conceived for heuristic purposes, modelling has become a core epistemic technology for prediction. However, currently, we observe a shift from the predictive use of models to an instructive one, in particular in the field of policy making, and this shift comes along with the turn from forecasting to backcasting. While forecasting poses “what will happen?”-questions, backcasting asks questions in the form of “if we want to achieve a particular goal, what actions must we take to get there?” It is this turn to backcasting which enables backward chaining of possible futures based on normative scenarios. Originating from energy pathway modelling, backcasting has become an integral part of environmental modelling and, in particular, of climate modelling. This paper explores these epistemic shifts and turns in detail and hints to an upcoming participative use of models.