Proceedings of Machine Learning Research 103:346–353, 2019 ISIPTA 2019 A Retrospective on Isaac Levi: June 30, 1930 – December 25, 2018 Teddy Seidenfeld [email protected] Philosophy and Statistics, Carnegie Mellon University, United States of America Abstract a Rabbi, who trained at the Jewish Theological Seminary Isaac Levi’s philosophy places him squarely within the in New York City. That is where Isaac was born (June tradition of American Pragmatism: the noble legacy 30, 1930). The family moved frequently, as his father was of Peirce, James, and Dewey, evidently influenced a somewhat the itinerant Rabbi: Birmingham, Alabama; by his teachers and colleagues at Columbia Univer- Auburn, New York; and in 1941 the family moved to Syd- sity, amongst whom E. Nagel and S. Morgenbesser, ney, Australia, prior to the Pearl Harbor attack. In 1942, and fellow graduate students at Columbia University, still in Australia, Levi’s father joined the US Army, be- e.g., H. E. Kyburg, Jr. and F. Schick. Important for coming the first overseas Jewish chaplain. Later in 1942, understanding Levi’s original perspective on large the rest of the family – Isaac, with his younger brother scale philosophical problems is the theme that deci- and sister, and mother – left Australia to stay with their sion theory is embedded in them all. Typical of his work, Levi’s contributions are grounded on significant maternal grandparents in Ontario, Canada. The family re- distinctions, many of which are cast with the aid of united, back in the Southern US, when his father, still a sound decision-theory. In this retrospective I review chaplain, returned from overseas in 1944 to be stationed in four salient examples of his interests, spanning Levi’s Hot Springs, Arkansas. (Isaac’s lifelong fondness for the work on 1) belief acceptance, 2) belief revision, 3) so- music of Hank Williams traces to this period.) After the cial philosophy, and 4) statistical inference. war’s end, the family moved to Detroit, Michigan, where Keywords: pragmatism, belief acceptance, belief re- Isaac graduated High School in 1947. vision, social agents Isaac had a greater commitment to Jewish religion than to Jewish culture. This matched his atypical personal ex- Isaac Levi’s philosophy places him squarely within perience living as a child in small towns in Canada, the the tradition of American Pragmatism: the noble legacy US South, and Australia, dominated by a (non-Jewish) of Peirce, James, and Dewey, evidently influenced by Protestant culture, but having a rabbi for a father. With his teachers and colleagues at Columbia University, e.g., the aspiration of following in his father’s footsteps, he at- E. Nagel and S. Morgenbesser, and fellow graduate stu- tended college at New York University and entered the dent at Columbia University, e.g., H. E. Kyburg, Jr. and Jewish Theological Seminary in preparation for the rab- F. Schick. Important for understanding Levi’s original per- binate. During his first year of studies at the Seminary he spective on large scale philosophical problems is the theme won a scholarship which he shared with another student, that decision theory is embedded in them all. Typical of Judith Rubins. They married (on Christmas Day) in 1951 his work, Levi’s contributions are grounded on significant and celebrated their 67th anniversary together on his last distinctions, many of which are cast with the aid of sound day. Together, they raised two sons, each successful in the decision-theory. In this retrospective I review four salient arts and letters, and each married. Judy and Isaac celebrated examples of his interests, spanning Levi’s work on (1) be- three grandchildren. lief acceptance,(2) belief revision,(3) social philosophy, When he entered the Seminary in New York, after High and (4) statistical inference. School, Isaac’s religious views had been grounded on a First, however, it helps to know how he came to Philoso- philosophical position that the only rationale for (Jewish) phy and, for the focus of this retrospective, to understand monotheism – the belief in a single, unified deity – was that origins of Levi’s distinctive approach – to use decision the- such a fact served as a basis for morality. But his exposure ory as a central tool in his philosophizing. Here is a bit of to Philosophy at NYU, in particular, what he learned from 1 speculative rational construction I offer for that purpose. the critics of ontological arguments for the existence of God – what he learned from formidable teachers such as 0. Early Years Paul Edwards and Sidney Hook – was the important lesson that “ought” does not follow from “is”! Levi’s parents were Canadians by birth. (Their parents had Even if reason alone could establish existence of the emigrated from Lithuania and Galicia.) Levi’s father was deity, that argument does not entail a normative code of 1. See his “Self-Profile” in Kyburg-Levi (Bogdan, 1982), which serves (Jewish) ethics. He realized that his religious views were as the basis for some of my speculative reconstruction, here. dependent, not on religious facts – dependent not on the © 2019 T. Seidenfeld. ARETROSPECTIVE ON ISAAC LEVI:JUNE 30, 1930 – DECEMBER 25, 2018 existence of a deity alone – but dependent instead on value This bio-sketch serves as a prologue to the following judgments, which are needed in order to support norma- sampling of four of Levi’s distinctive contributions. tive judgments of what one ought to do. It is this awaken- ing, I speculate, that kept him alert throughout his career identifying implicit value-theoretic aspects tacit for sound 1. Acceptance as a Cognitive Decision methodology. It kept him alert to the role of utilities, and In two early works Must the Scientist Make Value Judg- not merely coherent degrees of belief, in sound scientific ments? (1960) and On the Seriousness of Mistakes (1962), practice. And, I speculate, this separation of fact and value Levi argues (contra R. Rudner) that the values reflected in helps to explain such details in his work as Levi’s use of a statistical type-1 and type-2 errors may formalize distinct cross-product of a (convex) set of probabilities and a (con- cognitive, scientific values that are not to be conflated with vex) set of cardinal utilities: where degrees of belief and economic, ethical, or political values. Belief acceptance – values combine independently in his rule of E-admissibility. a voluntary act to adopt a new, full belief B in answer to a “Is” alone does not entail “ought”! well posed “which?” question – is Levi’s account of how The upshot of his undergraduate awakening was a con- to apply common standards of rational choice in the con- version from religious monotheism to a then popular blend text of expected cognitive utility decisions. One engages in of positivism and pragmatism as a basis for underpinning risky epistemic business when accepting a new full belief B morality; a position encouraged by others at the Seminary. – where B contains new, relevant information for the agent. These themes pointed Isaac to Philosophy at Columbia Uni- B is one potential answer to the which-question. Prior to versity and the thinking of John Dewey, who was the dean accepting B, the agent understands B might be false. That of American Pragmatism during the first half of the 20th is the core philosophic idea in his 1967 book, Gambling Century. In the 1950s, Dewey’s successor at Columbia was with Truth, which formalizes the decision-theoretic trade- the acclaimed Philosopher of Science, Ernest Nagel, who off between making an error and acquiring an informative, served as the first John Dewey Professor of Philosophy at true belief. Columbia. Nagel’s prominent students from that period in- In more detail, the structural assumptions in Gambling cluded, in addition to Levi, Patrick Suppes, Henry Kyburg, with Truth require a (finite) which-H question, that Levi and Frederic Schick. identifies with what he calls an Ultimate Partition: H = fh1;:::;hkg. The elements of H are the logically strongest Isaac’s PhD studies at Columbia (1951–57) moved him relevant answers to the question that is of interest to the away from foundationalist aspects of logical positivism and agent. This is a cognitive value judgment: finer partitions towards non-foundationalist pragmatism. One important than H do not add relevant information, and coarser parti- example that Nagel was fond of using when teaching Phi- tions than H lose relevant information regarding the ques- losophy of Science, is that observation reports, especially tion at hand as the agent identified that: namely, Which as they appear in science, are theory-laden and not theory- element of H? free. Scientific observations incorporate consequences of The algebra AH generated by H constitutes the set of an agent’s volitions, e.g., acceptance of settled background possible relevant answers to the which-H question. In Gam- assumptions, and are not merely a by-product of a passive bling, Levi uses a (precise) credal probability Q and a (pre- spectator sport. cise) cognitive, epistemic utility each defined over possible As Sidney Morgenbesser (the 2nd John Dewey Philoso- answers, A 2 AH . Levi’s novelty is in these utility func- phy Professor at Columbia) emphasized, in order to know tions. His idea is that the cognitive utility in accepting A as which inferences are legitimate it is important to identify the strongest answer to the which-H question is a convex the specifics that constitute the context of an inquiry – What combination of two epistemic goals: is the question? For Levi, the question at hand helps to fix (i) an information function, the content(A)2, and cognitive values that constrain the inquiry.