Sensitivity auditing: an important ingredient in the evaluation of models and metrics

Mathematical models that draw inference on real systems often feed into the process of policy making. Crit- ical aspects of this science-policy interface must include: (i) acknowledge the limitations of the mathematical models, (ii) refrain from over-interpretation, and (iii) evaluate the robustness of the conclusions. Sensitivity auditing - a novel extension of sensitivity analysis, is shown here to be up to the tasks (i-iii) above. Sensitivity auditing proposes a seven-point checklist for the scrutiny of policy-relevant modelling activities and metrics (quantifications, in general). In this contribution, we review six real-world applications of sensitivity auditing to the modelling process of a range of socio-environmental systems. The examples show the flexibility of sen- sitivity auditing to illuminate crucial and hitherto neglected aspects of the working of models at the science, policy, and society nexus.