"It has been shown, for example, that a pollster can in principle always publish his prediction of an election result in such a form that, despite the reactions of voters to the forecast, the prediction is not falsified by those reactions. CF. Herbert A Simon"--Ernest Nagel ((1960) Structure of Science, 473, n 13)
"it was shown that it is always possible in principle to make a public prediction that will be confirmed by the event...It was shown that correct prediction requires at least some knowledge of the reaction function."--Herbert Simon (1954) 253; emphasis in original)
Simon won a Nobel in economics in 1978. (In his intellectual autobiography he recounts his debts to Carnap and his ongoing interests in philosophy of physics.) In philosophy he is probably now best remembered for his work on bounded rationality/satisficing. The work discussed by Nagel (who didn't just invent the category analytical philosophy as we know it, but was also a leading figure in the discipline in the 50s and 60s [Nagel's Structure was much cited]) above and from which I cite is part of a larger literature in which Simon played a non-trivial role as acknowledged in the key (1954) paper by Grunberg and Modigliani (who won the Nobel in 1985). Simon & Grunberg/Modigliani provide a framework for showing under what conditions social scientific predictions need not be self-refuting--a welcome result to the economics profession that was warming up to Milton Friedman's (1953) proposal that what mattered was not the realism of the assumptions in an economic model but its predictive power (Simon was critical by the way).
Nagel does not name his targets in dealing with a "difficulty confronting the social sciences, sometimes cited as the gravest one they face." (466) But if one goes to Simon's paper it's not so hard to figure out; after a nod to Aristotle, he names Frank Knight and Hayek in the first footnote. Now, there are very important differences between Hayek and the now-forgotten Knight (not the least of which is that in TJ Rawls encouraged attention to Knight (he even tells us to read the footnotes) and not so with Hayek--a judgment of relative value that the discipline has reversed), but one important commonality is that in the 1940s and 50s economists from free-marketeers like Alchian to high theory types like Arrow were extremely eager to reject their skepticism about the technocratic turn of the discipline. Nagel is certainly aware of Hayek's skepticism. (In addition, Grunberg & Modigliani also point to another forgotten economist, Vining--recall my post.)
Although there is always a genuine possibility that action based on knowledge of social processes will alter the character of those processes, that possibility can therefore often be ignored, since such action does not generally transform radically the over-all pattern of habitual social behavior. (473; emphases added)
Let's call this the Ignore-Thesis. Now Nagel's argument for the Ignore-Thesis relies on an assumption that "constraints imposed on individual social behavior by relatively stable institutions within which individuals seek to realize their aims." (472; this is a very Humean point.) One can grant Nagel's assumption and even his conclusion, but it does not follow from any of this that one knows when one may safely ignore such genuine possibility. In particular, scientific possibility is generally relative to the portfolio of models (and paradigm-preserving extensions of these) of the paradigmatic science (plus bridge principles, know-how of the expert, etc). The problem is that a paradigmatic scientist has a tendency to view her portfolio of models as necessary (because "robust," "well confirmed," etc). But it is extremely unlikely that the paradigmatic models incorporate all the reasonable modals available in discarded or non-paradigmatic models (which -- let's stipulate -- may have done better on some sub-set of questions taken to be less central to the paradigm centric discipline). [This is not a hypothetical case--something like this scenario occurred in 2007 when the modellers could not even conceive a collapse of world-trade due to a post-bubble liquidity crisis.]
Now Simon's claim in his 1954 paper is quite modest; prediction is possible if (a) we have "some knowledge of the reaction function" (that is, we need to understand how people respond to polls, including dramatically different polling environments, and the pollsters need to be able to adjust to this in real time). Moreover, a quick glance at the Grunberg&Modigliani paper reminds us that the whole approach also presupposes (b) "that prediction, if not divulged, is possible" (465; emphasis in original). [The shared language of Grunberg/Modigliani and Simon is Brouwer's Fixed Point Theorem not modality, alas.] But, of course, (a-b) are non-trivial commtments. (Simon points out that a pollster may have a hard time defending (a) in front of a Congressional Committee.) So, it is by no means obvious that Nagel is entirely entitled to his own objections to the "gravest" objections. (Moreover, there are more non-trivial commitments, but these are of more technical nature.)
Nagel has another ace up his sleeve in favor of the Ignore-Thesis. He points to "the theories of games" as providing the resources that "given a competitive action directed toward attaining a certain goal, the strategy to be followed which is in a certain sense the "best" strategy, with an outcome unaffected by information athe parties to the competition...may have concerning each other's plans." (473, n. 13) Now, the crucial part here is "given a competitive action." In the context of reflecting on Boulding (another old friend) and Knight's work, Grunberg&Modigliani point out a very surprising feature of their model: successful prediction is Pareto optimal if and only if "perfectly competitive conditions prevail throughout the economy and if the public prediction includes all the variables appearing in the expectations of the agents." (477) This means that even if social scientists get it right in their predictions and their normative theory (let's stipulate "the theory of games" fits that bill) there is then a non-negligible risk they make matters worse; everybody agrees that in no really existing economy "perfectly competitive conditions prevail."
There is a larger story of how one gets from Simon/Grunberg/Modigliani to rational expectations (Muth/Lucas/Sargent). But here I want to briefly forget the history of economics and return home to the history of Philosophy. There can be no doubt that most of the ablest and prominent of the mid-twentieth century philosophers of science aligned themselves with the economist's technocratic turn. But it is not clear that the arguments available to them warranted this. Almost certainly non-technical reasons can better account for their stance.