Read on Twitter
#Modelling is a social activity meant to inform our understanding, not to support predetermined agendas. Out today on @nature our comment on how to make models serve society https://www.nature.com/articles/d41586-020-01812-9
Here is a walkthrough of the 5 principles we propose to #ModelResponsibly
Here is a walkthrough of the 5 principles we propose to #ModelResponsibly
1) #MindTheAssumptions
Assumptions that are reasonable in one situation can become nonsensical in another. All assumptions shall thus be assessed via #UncertaintyAnalysis and #GlobalSensitivityAnalysis – i.e. allowing all that is uncertain to vary simultaneously.
Assumptions that are reasonable in one situation can become nonsensical in another. All assumptions shall thus be assessed via #UncertaintyAnalysis and #GlobalSensitivityAnalysis – i.e. allowing all that is uncertain to vary simultaneously.
2) #MindTheHubris
#Accuracy is often a U-shaped function of model #complexity. Including more parameters reduces the error only up to a certain point. For at least two reasons (i) uncertainty builds up, in a cascade effect, (ii) the model loses flexibility (i.e. overfitting).
#Accuracy is often a U-shaped function of model #complexity. Including more parameters reduces the error only up to a certain point. For at least two reasons (i) uncertainty builds up, in a cascade effect, (ii) the model loses flexibility (i.e. overfitting).
3) #MindTheFraming
The technique is never neutral. Interests, disciplinary orientations and biases of the modellers will always be reflected in the model. As the choice of the tools will influence the outcome, the embedded normative values shall be explicitly mentioned.
The technique is never neutral. Interests, disciplinary orientations and biases of the modellers will always be reflected in the model. As the choice of the tools will influence the outcome, the embedded normative values shall be explicitly mentioned.
4) #MindTheConsequences
Excessive regard for producing numbers can push a discipline away from being roughly right towards being precisely wrong. #SpuriousPrecision adds to a false sense of certainty. The communication of a model is, at least, as important as its results.
Excessive regard for producing numbers can push a discipline away from being roughly right towards being precisely wrong. #SpuriousPrecision adds to a false sense of certainty. The communication of a model is, at least, as important as its results.
5) #MindTheUnknowns
Self-awareness of ignorance is a virtue; acknowledging limits of technical advice shall be a duty. Neglecting uncertainties could offer politicians the chance to abdicate accountability. See e.g. #AnthonyFauci's "there is no number-answer to your question".
Self-awareness of ignorance is a virtue; acknowledging limits of technical advice shall be a duty. Neglecting uncertainties could offer politicians the chance to abdicate accountability. See e.g. #AnthonyFauci's "there is no number-answer to your question".
Thanks to @AndreaSaltelli for leading this collective effort and to 21 co-authors @emme_dif @slopiano @arnaldpuy @GabrieleBammer @EricaCharters @ProfJohnKay @IsmaelRafols @Jeroen_vdSluijs @RogerPielkeJr @learnfromerror @philipbstark @JeromeRavetz #ModelResponsibly
