On Williamson, Kitcher, The Great Endarkenment (pt. II) and Meta-Expertise
Early on, in the “preface” to second (revised, 2022) edition of The Philosophy of Philosophy, Timothy Williamson notes that collaborating with the economist, Hyun Song Shin, gave him “experience of the differences in research culture between two disciplines when dealing with the same phenomena.” Crucially, he learns from the economist that the “model-building approach, on which models are assumed from the outset to involve drastic simplifications of the reality under study, so that a mere discrepancy between model and reality is not news, and just pointing it out is not considered a significant intellectual contribution. Rather, what displaces a model is a better model.” (p. xiv) Refutation by counterexamples (as in Gettier 1963) is not what economists and model-builders do (the point is echoed on p. 383 including the repetition of the contrast to Gettier). As Williamson notes later in the book (in the newly inserted section 9.3):
“When a system resists direct study, because it is so complex or hard to observe, model-building constitutes a key fall-back strategy. Studying a model often yields insight into the phenomena it models. When one model is replaced by another that captures more about how the phenomena work, science progresses. (p. 372)
Unsurprisingly, Williamson goes on to argue “Not only can philosophy make progress through model-building, it has been doing so for quite some time,” (372; emphasis in original.) I agree with him (and Laurie Paul back in the day).
Now, from the perspective of theoretical wisdom, there is little risk in switching from the method of counter-example to a modeling methodology. However, from the perspective of practical wisdom, things are not so simple. A discrepancy between model and reality may well be the difference between life and death or, closer to my own area of expertise, successful monetary policy and a systemic financial crisis. To be sure, as Bill Wimsatt always stresses (at least since the 1980s) false models may also be very useful in practice. (See also David Bloor’s The enigma of the aerofoil: Rival theories in aerodynamics, 1909–1930. Chicago.)
One important difference between the method of counter example and modeling methodology is that the former method builds a kind of stress-testing into the practice (and embraces what I like to call a Popperian ethos), while the latter risks embracing a too permissive, confirmatory bias. The important point is that there are costs and benefits to both in different contexts. (One might wish to model this, of course.) How to stress test and make abductive practices resilient and informative has been the life-long interest of my teacher, George Smith whose (rather lengthy) “Closing the Loop” is worth reading. (Sadly, there is no easy recipe.) Williamson wisely opts for mixed methods (p. xxii).
Interestingly enough, Williamson allows (even concludes a sophisticated argument with it) that it is completely legitimate to claim that where “practical consequences are objectionable, so are the theories that entail them.” (399) Let’s call this the ‘inductive risk principle.’ In embracing it, Williamson is a true heir to Carnap who also judged model choice by its practical consequences. (A note to future historians: Williamson came late to the philosophy as social justice perspective, but he arrived.) Of course, the details are going to matter because there are often quite a few steps between any theory or model and practical consequences (as Williamson could have learned from most philosopher of science of my generation and younger).
Notice, however, that above I also appealed to something like the inductive risk principle when I pointed out that the discrepancy between model and reality may well be a reason to avoid embracing it in practical affairs. Of course, as stated the inductive risk principle (even as developed by Williamson) is way too course-grained. Part of the problem is the vagueness of ‘objectionable;’ another is that not all consequences and not all people are alike in the capacity to deal with them. I learned from a recent, fascinating post by Richard Pettigrew that there is now a formal literature on “how much risk it is morally appropriate for one person to take on behalf of others (see Johanna Thoma and Lara Buchak).”*
Now, while we tend to think of David Lewis as a systematic metaphysician, Lewis himself had a keen eye toward the utility of modeling (from his earliest work on conventions onward). As I noted yesterday, in his “Academic appointments: Why ignore the advantage of being right,” Lewis himself drew on Philip Kitcher’s (1990) "The Division of Cognitive Labor," Journal of Philosophy in order to think about how we should model the norms and institutions of academic hiring (at certain important universities).
Kitcher’s paper (and his wider project) is foundational to recent modeling and simulation work on the structure of science (and projects like epistemic diversity). Kitcher himself frames his modeling exercise as follows: “it is intended to raise an important, if neglected, problem about the growth of science. Is it possible that there should be a mismatch between the demands of individual rationality and those of collective (or community) rationality?” (p. 6) In a note Kitcher cites Kuhn’s "Objectivity, Value Judgment and Theory Choice," Husain Sarkar’s A Theory of Method (Berkeley: California UP, 1983), and his “very sketchy remarks in ch. 6 of Abusing Science.”
As it happens Kitcher was misinformed on the matter; the question had already been asked, and in a very similar vocabulary, by Gordon Tullock back in (1966) The Organization of Inquiry. The possible mismatch between individual and collective rationality is an organizing principle of public choice (which Tullock helped make famous). I encountered Tullock in the late 1990s, and in homage to that book, I tend to call my own (incentive alert) approach, ‘public choice philosophy of science.’ And it shaped my own reflections on how the efficient market of ideas is often tacitly and dangerously presupposed in philosophy of science and institutional design (see here and here).
While Tullock was in contact with Popper when he wrote the book, in the book he credits Michael Polanyi with noticing this mismatch. My own view is that this possible mismatch is already theorized by Plato (recall here and here) and Adam Smith. (I don’t doubt Polanyi would agree.) I return to Plato below actually.
While public choice became very mainstream in economics, Tullock’s book (whose models are all stated without math) remained somewhat obscure, ahead of its time. That Kitcher was unfamiliar with Tullock is the effect of an imperfect market in ideas and, more interestingly, the advanced division of disciplinary cognitive labor.
Now, Kitcher’s approach posits, for the point of modeling, “a philosopher-monarch, with the prerogative of directing the course of scientific research.” (Because I am going to be critical, I do want to mention I have followed this approach myself here in a modeling exercise with Rogier de Langhe.) A key modeling assumption is that the philosopher-monarch, has “an unerring eye for detecting the objective merits of theories and complete control of the scientific workforce.” (pp. 8-9) While most of the moral and political problems with this assumption can be found in the second half of the conjunct, my interest is in the first more epistemic half.
Yesterday, I could have highlighted more that while verbally agreeing with Kitcher, Lewis weakens Kitcher’s modeling assumption in a subtle way. So for Lewis, “Each member of the department can judge, by his own lights, to what extent any given candidate holds true doctrines, and to what extent he is in error.” However, on Lewis’ view, in virtue of the fact that professional philosophers do not agree on the objective merits of theories (and know this about each other), they use their agreement over objective demerits to guide hiring decisions (in order, from risk aversion, to avoid “worst” outcomes).
In spirit, Kitcher’s philosopher-monarch echoes the 19th century utilitarian philanthropist which fancied themselves Platonic philosopher-kings designing institutions for empire and the poor. Of course, in the context of modeling the effects of the advanced division of cognitive labor, the stipulation that philosopher-monarch has “an unerring eye for detecting the objective merits of theories” may itself be worth exploring.
As it happens, in the Republic, political life is the effect of the division of labor. In fact, Plato quickly adds that the division of labor is itself the effect of cognitive differences, and even in the original model of the healthy city, Socrates inscribes non-trivial cognitive differences into his political structure (370de). Part of the drama of the Republic is the attack on the poets (see especially book 10) and democratic elites (in the ship of state analogy) both of whom malign and create confusion about expertise as such, includingthe true expertise in statecraft. (Arguably, this is also the criticism of Thrasymachus in Book 1, and other sophists elsewhere.) In fact, even without strategic others (as Anthony Downs would emphasize) who have an incentive to decrease trust on this, within the advanced division of cognitive labor, it’s not so easy to be confident that anyone has an unerring eye for detecting the objective merits of theories or models (etc.).
Now, in the Republic, Plato handles this problem (or offers an ‘existence proof’ as the economist-modelers would say) by making philosopher-monarchy a collective enterprise of a whole class, and by using all the major institutions of the polity to breed and cultivate this class. In addition, this class has a lengthy training in all the sciences, and considerable esprit de corps (itself cultivated by education, communism etc.). Short of that, and we may treat this as second-best, other means are needed; this is, in fact, exemplified by Socrates' deference to Damon. The advanced division of cognitive labor involves reliance on authority and all kinds of social cues.
As regular readers know (recall) Elijah Millgram’s The Great Endarkenment: Philosophy for an Age of Hyperspecialization convinced me that even within traditional disciplines it is likely that nobody has an unerring eye for detecting the objective merits of theories because of the effects of hyperspecialization and the development of local modeling languages and pidgins with their abstruse vocabulary. And so much the worse across disciplines. (Millgram makes a persuasive case that this is already modelled as the source of civilizational collapse in the famous Tower of Babble narrative in the version conveyed by the Hebrew Bible.) The fiction of a Philosopher-monarch cannot be used to model this because the question is itself how to model how anyone can detect the objective merits of theories within hyperspecialization outside their own particular expertise?
That is to say, we need a better model to think about this. It’s not sufficient to show that there is real expertise in a domain (like Williamson aims to establish for arm-chair philosophy in a polemic with Jonathan Weinberg). Rather it needs to be established that a kind of meta-expertise that can detect the objective merits of theories within hyperspecialization accross disciplines is even possible. Plato includes such meta-expertise in his philosophers, but for us matters are not so simple.
This question takes on more urgency when we normalize philosophy (with Williamson) “as an investigation of the same world which other sciences investigate too, and philosophical knowledge as the product of ordinary human cognitive capacities.” (p. xxiv) For Williamson anti-exceptionalism about philosophical cognition is the norm. (In virtue of the inductive risk principle I have long embrace a similar norm but call it methodological analytic egalitarianism.) The problem with anti-exceptionalism and my own methodological analytic egalitarianism is that it is very hard to see how meta-expertise can be achieved. Developing such meta-expertise is a pratically urgent matter in the context of, say, novel externalities.
As it happens what I call synthetic philosophy has a role to play in this.
To be continued…
*There is quite a bit to say on Buchak’s paper, in particular, in light of Rawls’ engagement with this very problem. But that’s for another time.