During the languid nights of the last two summer weeks, I have been reading, with much profit, the new installment of Scott Soames’s projected five-volume The Analytic Tradition in Philosophy. Volume 3: The Struggle for Modality, focused on Quine and Kripke, has been of particular interest to me. Why should an economist care about modal logic? Much of my research has been in Bayesian econometrics and, more recently, in AI, which also relies heavily on Bayesian reasoning. This approach has spilled into my daily life. Those who know me have heard me say things like: “Event X (e.g., the outcome of an election) will occur with probability 60%”, representing my degree of belief. Thus, I have spent considerable time thinking about the strengths, but also the weaknesses, of the Bayesian approach. One weakness, at least from my perspective, is that Bayesian reasoning struggles to capture logical constraints on beliefs (e.g., closure under logical consequence) and to provide structured update rules beyond moving posteriors. Modal logic, on the other hand, can impose these constraints in a more consistent way (even if at the cost of not being quantitative). In AI, it helps formalize ideas like expressiveness, i.e., reasoning about other agents’ beliefs, as in epistemic game theory or in “forecasting the forecast of others” models, something I worked on some years ago to understand business cycles. This is why I have always been drawn to @BFraassen’s project of constructive empiricism: 🔗 even if it is a minority position among researchers. It can be loosely interpreted as an attempt to merge the best of modal logic and Bayesian reasoning: the former providing the logical skeleton of belief, the latter supplying numerical content. Soames’s excellent account of Quine and Kripke’s struggles with many of these ideas will, in my modest (and not very expert) opinion, become an outstanding reference for years to come.