Classical Probability in the Enlightenment

In the AI Story

The book traces the work of Pascal, Fermat, Huygens, Jakob Bernoulli, Condorcet, and Laplace, showing how they developed probability theory in close engagement with practical problems of commerce, law, and natural philosophy. Classical probability was not pure mathematics applied externally to practical domains; it emerged from those domains, and its founding texts repeatedly emphasize that the mathematical formulations are attempts to capture the actual reasoning patterns of prudent judges, successful merchants, and careful natural philosophers.

The book's deeper argument concerns what happened to probability in the nineteenth century. As probability became established as a mathematical discipline with its own technical apparatus, it gradually lost its connection to the practical reasoning it had originally been designed to model. The 'reasonable person' of classical probability was replaced by the 'rational agent' of mathematical decision theory — a figure who calculates rather than judges, who optimizes rather than deliberates, who follows rules rather than exercising discretion. The transformation, Daston argues, was neither inevitable nor purely progressive; it represented a specific historical choice with consequences that continue to shape how quantitative reasoning operates in contemporary domains.

The book's relevance to AI is indirect but substantial. Contemporary AI systems implement a particularly extreme version of the rational-agent model — statistical inference without discretion, optimization without deliberation, rule-following without judgment. Daston's history suggests that this represents not the natural endpoint of quantitative reasoning but a specific trajectory that departed from classical probability's original intent. The question the book raises, applied to AI, is whether the technology can be redirected toward the classical probability ideal of formalizing rather than replacing judgment — or whether the rational-agent trajectory has acquired sufficient institutional momentum to resist redirection.

The book has become a canonical reference in the history of statistics, the philosophy of probability, and the history of quantitative reasoning more broadly. Its argument that probability was originally a tool for enlightened judgment rather than a replacement for it has influenced contemporary discussions of algorithmic decision-making, risk assessment, and the relationship between quantitative methods and human expertise.

Origin

The book emerged from Daston's doctoral research at Harvard and her subsequent work at Princeton and Göttingen. It was the product of extensive archival research in seventeenth- and eighteenth-century European texts, combined with philosophical and historical synthesis at the highest level. The book won the Pfizer Prize in 1989 and established Daston's reputation as a leading figure in the history of science.

The book's method — close historical analysis combined with philosophical argument about the conceptual transformation of its subject — became Daston's characteristic approach, applied subsequently to objectivity, to rules, and to the concept of data itself. Each of her major works extends the method, demonstrating how concepts that appear timeless and natural are in fact products of specific historical developments that could have unfolded differently.

Key Ideas

Probability as formalized judgment. Classical probability emerged as a tool for modeling the reasoning of prudent judges, merchants, and natural philosophers — not as abstract mathematics.

The reasonable person, not the rational agent. Classical probability's founding figures wrote for and about practitioners exercising discretion, not for optimizers following rules.

Nineteenth-century transformation was contingent. The shift from practical reasoning to abstract mathematics was a specific historical development, not the natural progression of the field.

Rational-agent model displaced reasonable-person model. Contemporary decision theory inherits a trajectory that departed from classical probability's original intent.

AI implements the rational-agent trajectory. Machine learning represents the extreme endpoint of statistical reasoning without deliberation — a trajectory the history suggests could be different.

Debates & Critiques

Historians of statistics have debated whether Daston's argument overstates the continuity between classical probability and practical judgment, or whether classical probability was already more abstract than her reading suggests. The productive resolution is that classical probability contained both practical and abstract tendencies, with the balance between them shifting over time in ways that Daston's history makes visible. A more current debate concerns whether AI systems should be understood as continuing the rational-agent trajectory or as implementing a different kind of computation that resists the old categories.