[This is a guest-post by Alistair Isaac.--ES]
The facts of Patrick Suppes’ career read like tall-tales of some 17th century polymath. In addition to philosophy, he held appointments in psychology, education, and statistics; he also made major contributions in the foundations of physics, linguistics, and the social sciences. Although best known in philosophy for his work on philosophy of science, he wrote on topics as diverse as free will, liberty, meaning in poetry, corporal punishment, hylomorphism, and virtue ethics.[1] He was also an entrepreneur, using the business of education—he taught with computers as soon as there were computers, and the internet as soon as there was an internet—to gather data for the science of education.
Suppes occupied the same office in Ventura Hall for 60 years, eventually taking over the entire building, filling it with his collaborators. When university expansion threatened it, he had Ventura Hall physically relocated rather than vacate. When the (Suppes founded) Stanford Education Program for Gifted Youth (EPGY) Online High School took off, there was no longer room for all his collaborators, so he turned architect and designed Nora Suppes Hall next door to house the OHS. In the 2000s much of Ventura Hall was dedicated to Suppes’ quest to understand computation in the brain. This project combined physics (physically modeling the impedance of the skull), engineering (he was on the committee of a PhD student who designed new skull contacts for EEG recording), statistics (literally years of subtle data analysis), the psychology of language (developing associationistic semantic models), and, of course, methodology (including an epic defense of the theoretical importance of invariants across EEG and perceptual data.[2] By 2012, he’d embarked on an ambitious new project to collect EEG signals from couples in therapy, with an eye toward understanding the neuroscience of “insight.”
Suppes’ bibliography includes almost 500 articles and books, averaging 7 per year throughout the 2000s and 4 per year through the 2010s. His last single-author, peer-reviewed paper appeared in 2014 in the Journal of Mathematical Psychology and proved the conceptual independence of our qualitative notions of probability, uncertainty, and probabilistic independence—demonstrating that no one can be defined in terms of the other two. [3] Whether this result turns out to be significant or not (opinions were divided when Suppes presented it informally to peers at Stanford), it shows Suppes’ commitment until the end to the methods he became known for: the characterization of basic concepts with qualitative axioms, and the proof of precise results about them.
Suppes is probably best known in contemporary philosophy for initiating the “semantic” approach in philosophy of science with two seminal papers from 1960.[4] On the canonical interpretation, the semantic view identifies scientific theories with the models that satisfy them, where “model” is understood in the logical sense of set-theoretic structure. As the influence of the logical positivists has waned, formal methods in philosophy have fallen out of favor, and philosophers of science have turned from physics to biology for paradigmatic examples, Suppes has grown to represent an archaic and confused age. Models in science are not set-theoretical; logical proofs tell us nothing about scientific practice; and metaphysics must not be sacrificed behind a smokescreen of empty formalisms!
This popular reading of Suppes and his project misses its aims, its pedigree, and, perhaps also, its value. Suppes is typically read as a halfhearted reaction against (or even within) logical positivism: he accepted its general project (the formalization of the epistemic question of how observation relates to theory through the techniques of logic), while merely rejecting its strict adherence to first order languages. Yet, Suppes’ project had independent motivation as a development of the pragmatist tradition he inherited from William James via John Dewey and Morris Cohen, advisors to his own advisor Ernest Nagel.
Suppes’ pragmatism shaped his project in two ways. First, it drove him to reject the very possibility of the logical positivist program succeeding in its strong form. He believed that actual, day-to-day scientific practice (whether in the lab, or at the blackboard) is too nuanced and complex to convey in language—apprenticeship is essentially required in order to gain expertise in science. Or indeed any complex activity: here Suppes loved to draw comparisons to tennis and shoe-making, and to lambast the possibility of axiomatizing those activities, or treating them in a “detailed” formal manner. If science in the lab cannot be formalized, then no amount of formal analysis can bridge the gap between observation and theory.[5]
As an aside: here also, Suppes practiced what he preached throughout his career. Those who were sucked into his projects in Ventura Hall received close hands-on training—their regular meetings with him amounting to apprenticeship in the old style. This extended even to philosophy: for instance, Suppes organized a reading group on topics of his active research for his graduate students, and roped us into designing philosophy courses for the EPGY Online High School, developing their contents in close collaboration with him.
So, criticisms of Suppes’ project as insufficiently sensitive to actual scientific practice misconstrue it in the extreme. The problem here was Suppes’ own reticence. Recognizing the subtlety of day-to-day scientific practice, he stayed silent on it in most of his publications. His understanding of apprenticeship, and the vast chasm between his formal project and the nuance of actual practice, has only been made known to those outside his circle of acquaintance very recently in the form of numerous informal interviews.[6]
But if Suppes’ formal project is not meant to connect theory to observation (as 1962’s admonishment to examine the relationship between models of theory and “models of data” has often been interpreted), what was its point? One interpretation might be conceptual clarification—and here the second strand of Suppes’ pragmatism shows itself.
Most of Suppes’ actual work with models of theories consists in proofs of representation theorems. (A representation theorem shows that all models of a theory are equivalent to some model in a distinguished subset, e.g. that any finite automaton may be represented by a stimulus-response model.) In his magnum opus, Representation and Invariance of Scientific Structures, 2002, Suppes argues that one of the points of representation is “improving our understanding of the object represented” (just as blueprints help us to better understand the layout of a building) and claims “the formal or mathematical theory of representation has as its primary goal an enrichment of the understanding.” Showing that two apparently distinct concepts (as translated into some precise form, e.g. sets of axioms or set-theoretical structures) are in a precise sense “the same” (through a representation theorem), clarifies our understanding of those concepts.
This project of conceptual clarification connects closely to the pragmatist idea that concepts should be identified with their effects on action—two concepts that do not motivate different actions, are not in fact different. This core pragmatist idea was introduced by Peirce as an elaboration of Kant’s analysis of the antinomies—apparent contradictory claims are reconciled by showing they involve reason’s attempt to employ concepts beyond their potential applicability to the phenomenal world. So it is no accident that Suppes was also fascinated with the antinomies. His argument that the third antinomy (the apparent conflict between freedom and determinism) is resolved by a mathematical result (the in principle indistinguishability of discrete observations of deterministic and probabilistic systems satisfying the ergodic property) makes sense when viewed through the lens of his formalized pragmatism.
Stepping back from Suppes’ contributions to philosophy as narrowly construed to the methodological foundations of science writ large, his most longstanding contributions will likely be in the foundations of measurement.[7] Suppes and collaborators used qualitative axioms to characterize the assumptions of various empirical measurement procedures, then proved representation theorems for numerical (more generally, geometrical) structures. These theorems legitimate our practice of representing measured quantities with numbers, and they allow for a precise characterization of the meaningfulness of claims about those quantities, namely: meaningful claims are those invariant across transformations of scale which preserve the axioms.
The measurement results exemplify the general features of Suppes’ project. First, formal methods do not bridge the gap between observation and theory completely. Once the features of observation, in this case through measurement, are formalized qualitatively, however, we can precisely bridge the gap between those qualitative assumptions and our quantitative methods through a representation theorem. Furthermore, this formal analysis sheds light on the nature of (our concept of!) the measured quantity, by providing a precise delineation between those claims about it that are meaningful and those that are not.
What Suppes proposed then was a method for doing philosophy, but not an actual overarching theory. He was not a systematic philosopher, as his work was not united by a set of coherent conclusions, but rather a dedication to the importance of particular concepts: representation, invariance, structure; as well as to the importance of detail. His most frequent remark after talks was to express “skepticism” that the speaker’s project could be carried through “in any detail.” In many respects, Suppes’ obsession with detail paralyzed him, preventing him from moving beyond an (admittedly vast) array of specific results to the development of a systematic overarching view. At the end of Representation and Invariance, he confronts this paralysis in the context of scientific reduction, apologizing for the lack of reduction results in the book with the claim that “the problem was too difficult for me,” before going on to defend a pluralistic approach to the philosophy of science, where formal results of the limited sort achievable coexist happily with more qualitative analyses of the vast gap between scientific knowledge and the complexities of the everyday world.
Suppes was “Pat” to all who knew him, and the cold mathematical approach of his most influential papers belied a warm and generous personality. Pat consistently made time for undergraduate teaching, even decades after his “retirement,” and was always generous with his graduate students. He would never talk down to you, but was also happy to explain that which you did not understand. It’s true that Pat did not “suffer fools,” but a fool to him was not one incapable of the kind of mathematical work he preferred—Pat was well aware it was not for everyone, hence his pluralism—a fool was one who refused to acknowledge when he did not understand; one who spoke from a position of ignorance he made no attempt to remedy.
Pat loved nothing more than to argue, but he was also quick to admit ignorance if you displayed knowledge he did not (yet!) possess. He was eager for his advisees to work on topics that outstripped his background as it offered an excuse for him to “learn from you.” If, however, you made a claim of which he was suspicious, he would demand you back it up with proof. More often than not, when you went to seek it out, it was Pat who knew better than you after all.
The last time I saw Pat was at the 2012 conference in honor of his 90th birthday. Multiple speakers expressed their eagerness to attend the next one, in honor of his 100th. At the time it seemed almost inevitable, that Pat was invincible, because of the sharpness of his remarks and his dynamic response to each talk. He put many of the participants decades his junior to shame with his wit and energy. Throughout, it was his argumentative nature that showed through. I once audited an undergraduate course he gave on free will in 2009; each class would start slowly, Pat had no real plan and languished at the front of the room. All it took, however, was one comment to which he objected, and he leapt into action, expounding at length on science and will, habit and explanation—class had begun.
Whatever Pat’s legacy in philosophy turns out to be, his profoundest effects will largely be invisible. His unrelenting dedication to a particular way of doing philosophy, and a bottomless commitment to pedagogy, drove him to shape tens of thousands of students. Through his mathematics textbooks, his early use of computers with gifted youth, his later innovations in online education, his six decades of undergraduate teaching, his pressure to bring particular figures to Stanford, shaping its legacy in formal philosophy, and his constant use of his own personal financial resources to build buildings, fund speakers, and start the Suppes Center for History and Philosophy of Science, Pat shaped the intellectual development of generations, in many cases without them ever realizing it. He will be missed, both by those of us who knew him, and by those of us who unknowingly have been swept into his intellectual legacy.
[1] e.g. free will: “Voluntary Motion, Biological Computation, and Free Will,” 1994; liberty: “Four Varieties of Libertatianism,” 2006; meaning in poetry: “Rhythm and Meaning in Poetry,” 2009; corporal punishment: “Some Formal Methods of Grading Principles,” 1966; hylomorphism: “Aristotle’s Concept of Matter and Its Relation to Modern Concepts of Matter,” 1974; and virtue ethics: “The Good and the Bad, the True and the False” (with Aimée Drolet), 2008
[2] “Partial Orders of Similarity Differences Invariant Between EEG-Recorded Brain and Perceptual Representations of Language” (with Marcos Perreau Guimaraes and Dik Kin Wong)
[3] How does one prove that concepts are independent? Using Padoa’s Principle of course! See Suppes (1957) Introduction to Logic.
[4] “A Comparison of the Meaning and Uses of Models in Mathematics and the Empirical Sciences” and “Models of Data” (pub. 1962)
[5] Suppes’ convictions about pragmatism and apprenticeship were also shaped by his experience as a meteorologist in World War 2. He was impressed by the superiority of the predictions of veteran weathermen over those provided by the more theory-savvy youth, such as himself. This experience also contributed to his career-spanning interest in the fundamentality of probability for our understanding of the world.
[6] For instance, Roberta Ferrario and Viola Schiaffonati’s Formal Methods and Empirical Practices: Conversations with Patrick Suppes, 2012
[7] Foundations of Measurement, 3 vols. (1971, 1989, 1990, with R. Duncan Luce, David H. Krantz, and Amos Tversky), as well as numerous earlier and later papers
Thank you for this kind eulogy. I have but one historical remark on this passage:
"Suppes is typically read as a halfhearted reaction against (or even within) logical positivism: he accepted its general project (the formalization of the epistemic question of how observation relates to theory through the techniques of logic), while merely rejecting its strict adherence to first order languages."
If this is indeed so, then Suppes rejected no aspect of logical positivism at all, since the logical positivists accepted (and in the case of Carnap, Woodger, and at least at one point even Hempel explicitly relied on) higher order logics.
Posted by: Sebastian Lutz | 12/08/2014 at 06:00 PM
Terrific obit!
Posted by: John Perry | 12/11/2014 at 07:03 PM