How good is a model?

Pandemic politics highlight how predictions need to be transparent and humble to invite insight, not blame.

One wonders why this “admonition” would be necessary, unless of course science and scientists themselves are often driven by ideology, money, or other self-serving interests to come to certain conclusions.

While this commentary is squarely aimed at the COVID-19 issue, the complaints found therein can be applied to many other arenas where modelling is relied upon, from climate change to evolution. Imagine that the disease models were mostly wrong because we could observe and compare with the real world in real time; what about the scenarios from the distant past that we can’t observe or replicate now?

One might also claim the reason the models were so wrong on COVID-19 is because they are the products of the mentality and approach that all other models are borne from. This is how science has approached other problems and nobody questioned them before, so how could they get it so wrong now? I would argue many other models are wrong, but because we can’t test them satisfactorily nobody important bothers to try.

The claim to fame for modelling is weather prediction, which is funny since the weather people are wrong half the time even today. Perhaps the most important takeaway fact is that models can only take into account a limited set of variables. The more variables, or parameters, that have to interact the more uncertain the predictions become until they are useless. That’s why they can never truly represent reality, much less when those models can’t be evaluated and compared to the real world in real time.

Oh well, at least someone is saying something…

Leave a Reply