Why don’t scientists go into politics?

Given that almost no active national-level polician has any meaningful training in science (even at an undergraduate level, but certaintly not as a research scientist), it’s remarkable how so much of policy is claimed to be ‘science based’.

Scientists are trained to understand natural phenomena. Human behaviour, at least in aggregate, is outside of their sphere of competance. Social sciences are not sciences, and the dismal science may be dismal, but it is not science.

Science often develops models. Newton proposed a mathematical model of the relationship between the force on an object, and its acceleration (although some would argue that he simply refined the definition of ‘force’: science often proceeds by redefining common natural language words in a way that non-scientists do not appreciate). Models are useful for testing hypotheses. If a model is constructed which predicts measurable behavior, and that prediction matches experiment, the model is useful in establishing the assumptions that have gone into constructing it. It does not prove the assumptions, because scientific hypothesis can never be definitively proved, only disproved. Complex models, once calibrated, can be useful for practical purposes, like allowing engineers to design bridges that will not fall down under most plausible scenarios. Simple models, however, also have an important place in science. They allow us to gain an insight into the underlying physical processes, and allow us to have an intuitive understanding of the world in a broader sense. Newton’s Laws are not correct (Einstein and Schrodinger proved that) but they are much more useful for a practical understanding of the world we live in than the more complex (but more complete) later models.

Policy decisions need to be informed by an intuitive understanding of the very complex system in which we live as humans. This is why, for all the benefits of DSGE models in economics, highly regarded commentators like Paul Krugman will use something like an IS-LM model to formulate their policy recommendations. I seriously doubt if better policy will come out of better modelling, because the better, and more complex models are so difficult to understand, and because what they gain in realism they lose in being more highly parametrized (and therefore needing more statistics).

In the current crisis, the progress of the disease is too dependent on social norms to be easily modelled and even if modelling were possible, the main problem is the political one of understanding how to trade off deaths against economic pain. Although we already do this in the National Institute for Clinical Excellence (the body that decides whether or not a new drug gives enough extra life to be worth the price demanded by the drug company), politicians are understandably unwilling to talk openly about the monetary of a saved life.

Anyway, in my view it’s about time that politicians stopped hiding behind the lab coats of the scientists and made some decisions. Complex models are not suited to guiding policy. We should listen to scientists but come to our own conclusions, just as we do with economists.