Chaire "International Francqui Professor" - From Social Analysis to Political Influence: Reinventing Statistical Modeling

Ernesto San Martin has focused both his research and development projects on the statistical modeling of social phenomena, particularly in education, political science, survey analysis, and public policy evaluation. Additionally, he seeks the ontogenesis of concepts currently used in probabilistic, statistical, and natural philosophy work developed from the 17th century onwards. Throughout these efforts, he has refined the fundamental epistemological aspect of statistical modeling according to his view: namely, that what can be learned from the data is different from what one seeks to learn.When these two objectives do not align, identification problems arise—issues that occupy much of his work.

Ernesto San Martin develops this work with the aim of interacting with public policymakers and politicians. This work has allowed him to understand the difference between the faculty relevant to scientific research (the rational faculty) and the faculty pertinent to political action (the aesthetic faculty). For this reason, his recent research focuses on reformulating the statistical modeling of social phenomena in a way that gives politics its due prominence, with scientific knowledge being just one of the many aspects that statesmen and women may take into account.

Working on the modeling of social phenomena has given me the opportunity to observe how sociologists and political scientists, among others, use statistical methods as a fundamental tool for analysis. Certainly, one possible reaction to these observations would be to correct   the improper uses of these quantitative methods, as well as to respond with "more flexible models" to the challenging causal questions that sooner or later arise in the understanding of social phenomena. This type of critique can already be traced to Condorcet's warning in his Réflexions sur l’arithmétique politique: “[…] et je vois encore des gens de bon sens que dès qu’il voient des chiffres dans un livre politique se sentent tout prêts à croire comme par une espèce de talisman." (I maintain the original spelling).

But instead of, or in addition to, challenging social scientists about their use of statistical methods, I have also had the opportunity to be challenged by conceptual and descriptive analyses of what the social and political realm is. One such analysis can be found in Hannah Arendt's On Violence, published in 1969, a year marked both by the Cold War and the wave of revolutions in Latin America:

“Events, by definition, are occurrences that interrupt routine processes and routine procedures; only in a world in which nothing of importance ever happens could the futurologists' dream come true. Predictions of the future are never anything but projections of present automatic processes and procedures, that is, of occurrences that are likely to come to pass if men do not act and if nothing unexpected happens; every action, for better or worse, and every accident necessarily destroys the whole pattern in whose frame the prediction moves and where it finds its evidence […] To call such unexpected, unpredicted, and unpredictable happenings ‘random events’ […], condemning them to irrelevance […] is the oldest trick in the trade; the trick, no doubt, helps in clearing up the theory, but at the price of removing it further and further from reality. The danger is that these theories are not only plausible, because they take their evidence from actually discernible present trends, but that, because of their inner consistency, they have a hypnotic effect; they put to sleep our common sense, which is nothing else but our mental organ for perceiving, understanding, and dealing with reality and factuality.”

In this context, Statistics is nothing more than a tool that identifies social regularities, in contrast to the unpredictability of political actions. Faced with this tension, I ask: is it possible to reinvent statistical modeling so that it becomes less of a tool of regularity and more of a tool that allows us to quantify how much we don't know about phenomena?

One path towards an answer appeared when revisiting a classic, yet scarcely read and even less understood, text: Kolmogorov's Grundbegriffe der Wahrscheinlichkeitsrechnung,published in 1933. This conceptual path rests on three pillars: the world is finite, discrete, and contingent. Using these pillars, we aim to present a reinvention of statistical modeling. To illustrate our proposal, we will analyze the potential electoral fraud in the 2021 Peruvian presidential elections; the Venezuela case fall also in the same case study. These cases will serve not only to emphasize the three pillars mentioned above but also to show that an identification problem arises as a consequence of a substantive question.