Scenarios are not (deterministic or probabilistic) predictions but explorations either of possible futures (forecasting) or of necessary policies to achieve a future considered as desirable (backcasting). They are usually understood to consist of a narrative illustrated by a computer simulation using parameters derived from the storyline and quantifying (parts of) the scenario. Unfortunately today, predictions, projections, probabilistic forecasts and more are labelled scenarios, while some scenarios like economic growth forecasts are presented as (probabilistic) predictions.
A system theory based analysis shows that the way of deriving plausible or predictable futures depends on the kind of system described, which in turn significantly influences the recommendations derived from the scenarios for policy measures to be taken to achieve an externally defined, desired state of the system. For these recommendations to be a solid guidance, it is necessary that the system under analysis (the reality), the mental model used to describe it, e.g. in the narrative (the perceived reality) and the computer model used for the simulation runs (the virtual reality) have a common level of complexity. If not so, a confusion of probability, uncertainty and ignorance emerges, and the resulting scenarios including the recommendations derived can be deeply flawed, misguiding or at best be meaningless for the policy implementation process.
The results show, that only a (co-) evolutionary model of reality exhibits an adequate level of complexity, and the meaningful scenarios are descriptions of their path dependent development trajectories. In such systems, however, predictions are not possible and surprises can occur any time. Thus decisions must be taken under uncertainty, and better understanding the character of the system and the kinds of uncertainty helps improving decision making.