by Paul Erickson, Judy L. Klein, Lorraine Daston, Rebecca Lemov, Thomas Sturm, Michael D. Gordin
Rating: ★★★
This is an intellectual history of rationality, one of the more overloaded terms in all intellectual history. The authors insist on the phrase 'Cold War rationality' because they are attempting to bound and historicise the subject, but they failed to convince me of their bounds.
It was first and foremost a sense of unprecedented urgency that distinguished debates over rationality during the Cold War from those over similar issues raised before and after: in the minds of the participants, nothing less than the fate of humanity hinged on the answers to these questions.
Yes, because modern discussions about rationality don't reference the fate of humanity or anxiously consider certain time-frames. There's certainly not a community of discordant self-critical voices arguing urgently about how to think, and what getting that wrong might mean for the species. Ahem.
There's a tone to the book that irks me. The authors seem affronted not at their subjects' methods (imperfect, sure, but they would be the first to agree, and the authors acknowledge these were at least no worse than the alternatives) but at their intellectual arrogance in attempting to solve the problem of nuclear annihilation. As if it is better to be humbly and fearfully inactive than to save the world, safer to cower before complexity than to attempt to discern some rules, some fundamental principles to guide the most important decisions in the world. Would the authors be happier, less sniffily derisive, with scholars who wrote only indecisive analyses of particular events? Who kept tidily to their own disciplines rather than pursuing fundamentals across the behavioural sciences? Decision scientists who stick to description, and make little effort to be useful to decision-makers?
Whatever the opinion of the authors, the book describes some fascinating scientific history. I had no idea that dynamic programming had origins so closely connected to the US Air Force, for example, and seeing the cost of computation at the time really drives home how incredible advances have been. While there are a bunch of names in here that I recognise (Ellsberg, von Neumann, Morgenstern, Nash, Deutsch, Kahneman and Tversky), there are also quite a few new ones with interesting contributions. Yet it felt at times a little like the authors wanted to do a comprehensive history of RAND but lacked the energy -- everything is covered fairly quickly, and it often feels like 'gosh' factoids are being held up to distract you. For example, chapter four repeats a few times the mildly interesting fragment that anthropologists were studying Micronesian cultures at the same time as the US was blowing up some Micronesian islands with nuclear weapons tests. The author of that chapter seems fascinated that neither research programme mentions the other, despite the obvious explanation that they had essentially no bearing on one another.
I struggled a bit to understand the book's perspective on events it was describing, and especially what exactly it was criticising (if indeed that was the point -- I find myself second-guessing. Was the tone that annoyed me imagined? Should I be giving more weight to the 'almost' in the title? Are all the many scare-quotes actually just some weird authorial habit?). I'm probably not the right person to do a proper critical review of this -- I'd want to read a dissection from a modern rationalist who's reread the sequences more recently, uses more formal methods, and ideally has some prior interest in the Cold War (so, I guess approximately anyone who's regularly taken part in Petrov Day exercises). However, I can plead also that some of the things seemingly being critiqued were really all over the place. In the first content chapter, we're meant to doubt the sense in scientists... looking for rules? In the fourth, the terribly strange rationalists are... collecting social and anthropological data efficiently? The sixth sees the authors questioning whether the rules of logical conjunction make sense. I omit the second and third only because I'm genuinely not sure if the authors were criticising anything there. Sometimes what sounded like fairly sensible research methods were described in shocked detail, and I had to check a few times to see that, yes, the authors acknowledge that these methods were very successful and produced useful results and in several cases led to Nobel prizes.
I mean, I'm not in general opposed to the thesis that scientists studying rationality in the period from World War II until the 90's could have been up to some weird stuff and gotten a bunch of important findings wrong. That would be pretty much expected at this point. What I'm confused about is the book doesn't really show us any such thing, at least not consistently. About the only approach the book compellingly presents as wrong is the deterrence rationality work by Kach, but that chapter spent much more time on the work of Osgood, whose predictions it contends were better supported than Kach's, and who it seems was also studying instrumental rationality and game-theoretic methods in the context of the Cold War, just using a slightly different frame. This is hardly damning of the whole research culture.
If I had to strengthen what I think their point might be, it would go like this: it is natural in any science to form models that are simplified and controlled forms of the phenomenon being studied, to enable fundamental rules to be discerned. However reasoning about and optimisation of models is only useful insofar as they are true to the phenomenon. So if you draw a map of Europe, you might justifiably omit terrain features and just focus on the national borders. But if you intend to use the map to plan the march of an army, this map is in many ways terrible, and optimisations that make sense on the map, like marching into Italy from France, might encounter unexpected barriers in the territory (like the Alps). The study of rationality in conflict during the Cold War showed some of these problems -- simplified representations of decision-making and reasoning problems were optimised and studied in the laboratory, but the very simplification of the model meant that results from its analysis could not be relied upon in the desired application domain of international diplomacy, and there was sometimes an uncritical assumption that actors like the Soviet Union would behave according to rational incentives rather than being affected by the multifarious 'other factors' stripped out of the models made to arrive at those predictions. This is agreeable enough, and a criticism that I think modern rationalism is well aware of and tries to handle with its much heavier empirical testing focus.
Overall, I was not convinced and honestly a little puzzled by the authors, and felt there was a depth lacking to the history, but I did enjoy some of the study descriptions, and it was a nice if odd trip around a subject I like.