Forecasting Unpredictable Social Events?

In 2016, Brexit and Trump victories dealt a heavy blow to the predictive capacity of most analysts, experts and pollsters who had not anticipated them. These outcomes were unforeseen events that were not on the list of possibilities. The same type of sudden and unforeseen phenomenon occurred with the Yellow Vests movement in France and the massive rejection in the Algerian streets of a fifth consecutive term of the incumbent President. At the same time, certain events that are just as unexpected at one point in time may become predictable by the polls at another time later before the election deadline, even if they are still considered abnormal. The massive election of a comedian featuring the President of Ukraine in a television series as President of Ukraine is a perfect illustration of this case. Using a model of opinion dynamics, I show that these two types of “abnormalˮ phenomena are produced by the deployment of a non- linearity in the temporal evolution of the distributions of opinion of the various competing choices. The discriminating difference between the two, being the moment of the emergence of the break in the dynamics of opinion with respect to the voting date. It means that, in principle, a knowledge of the equations governing opinion dynamics makes it possible to predict what seems unforeseeable. However, in some cases, even with the right equations, the prediction becomes random due to an incompressible uncertainty in the evaluation of certain parameters.