George Box, the British statistician, said the famous line “all models are wrong, but some are useful”. It is fairly easy to see the truth of this statement in the natural sciences, where our models are most assuredly wrong technically, but still manage to yield great utility in practice. This is because in the natural sciences, the models which are wrong (e.g. the theories of relativity, which we know cannot be fully reconciled with the theories of quantum mechanics) are still approximately correct in their contexts, which is all we need for them to be useful, to make good enough predictions. In the social sciences however, it’s often not obvious how wrong our models are, and that calls into question their usefulness.

Taleb, the author of The Black Swan and Antifragile, has made this point extensively in his works. He has harshly criticized academics in economics, finance, and politics for being overconfident in their models, pretending that they can make useful predictions when they are nothing more than easy-to-peddle lies. An example that he uses to illustrate the absurdity of these models is one where a university professor, upon observing that birds frequently sit outside the window during lectures and fly away afterwards, concludes that the lectures have taught the birds how to fly. According to Taleb, most economists make up fancy theories about how the world works backed by elaborate mathematical equations, yet their predictions still fail spectacularly in practice. These models only give their users a false sense of security and control, while yielding no better results than fairy tales. Worse actually, as sometimes they even cause harm.

This is an extremely skeptical perspective, to call out on so many extremely intelligent and accomplished people, along with their equally impressive looking models. I am naturally a highly skeptical person, and as I recognize the downsides of too much doubt, I try not to be too radical in my skepticism. But as hard as I try here, I find that for the most part, I agree with Taleb on this matter. Systems that are as complex as a nation’s economy or political dynamics are beyond our capability to model accurately. In the social sciences, unlike in most areas of the natural sciences, with the chaotic and huge impacts of second order (and beyond) effects, even approximately correct models do not give us approximately correct predictions. And we don’t even know if our models are approximately correct to begin with!

There is a sentiment from Taleb that we should simply abandon our models, since they are neither correct nor useful, and even harmful at times. But are models only useful for their ability to make predictions?

I think that models, even when they are wrong and non-predictive, help us work in groups. Just like any other collectively shared myths (e.g. the concept of money), a set of models, when adopted and shared by a group, helps align visions and goals for the group. When faced with a problem, such as a growing unemployment rate, we need a prevailing shared model in order for us to reach enough consensus to act. The model we use will be wrong, and the action we take will most likely not have the intended consequences, but at least we have the ability to do something collectively.

Here I must mention the concept of iatrogenesis, which refers to the harm done by the ones trying to help. Taleb discusses this at length in Antifragile, criticizing the people who are too eager to perform interventions in various affairs, based on their wrong models, often making things worse. I think this argument is fair, as there are certainly many cases where acting on our inevitably bad models to intervene in some complex system is bad. But at the same time, there are other, less interventional situations, where we are actively trying to create something new. We need shared models, regardless of their correctness, in order to come together and work collectively in these cases.

So instead of sitting around and letting the wind take us wherever it decides to take us, having models allows us to wander (randomly) collectively. Often it will result in us going in circles, or even backwards, but occasionally one group will triumph, and that is enough for us to slowly make progress forward, by our own actions. If nothing else, shared models give us the ability to actively work towards something together, which in turn gives us collective and individual meaning.

When we recognize the incorrectness and uselessness of our models, whether we should abandon the models completely reminds me of the following question: is it better to walk with the group in the wrong direction, or to walk alone in the right direction? I don’t think there is a right or wrong answer, and I have a feeling that Taleb would choose (and arguably have chosen) the latter. But many others will gladly go with the former.