Karl Popper’s ‘constitution of science’ (& 3)

In the second entry of this series, I discussed one of the two big problems of Karl Popper’s ‘Proto-constitution of science’, namely, whether following in the strictest possible way Popper’s falsificationism would necessarily be the most efficient means to attain what he considers in The Logic of Scientific Discovery ¹ to be the major goal of science: the maximisation of the testability of theories. Now I will concentrate on the second problem I mentioned: why should be testability the most important goal, and perhaps the only goal, of scientific practice? As I asked in the preceding entry, what if the goals of science (and of scientists) are more diverse, more complex, and if there are more labyrinthine ways of approaching them than through an obtuse fixation with testability?

Regarding this question, it is worth remembering that Popper himself described the epistemic goal of science in more mature works in so different ways that it looks as if it had changed his mind about the problem, something that, of course, he denied (in the paper “The aim of science” ², the goal is said to be to explain everything we consider that deserves an explanation; in the book Conjectures and Refutations ³, the goal seems to have been transformed in the search for ‘interesting truths’, or in the maximisation of verisimilitude, closeness to the whole truth). It is well known that Popper said, in later writtings, that in the Logik der Forschung he had refused to mention truth as the goal of science, as he attempted to do some decades later thanks to Alfred Tarski’s ‘rehabilitation’ of the notion of truth without abandoning the idea that a more abstract notion of ‘progress’ is what distinguishes science from other activities. One may wonder how Popper’s book would have look liked if he had opted for explicitly asserting that ‘progress towards (knowledge of) the truth’ were the ultimate goal of science. Probably, his views on methodology would not have changed substantially, for the concept of verisimilitude, or approximation to the truth, was later defined by Popper in such a way that a falsificationist methodology would naturally derive from it, or at least would be coherent with it, in the following sense: theory X is at least as close to the truth than theory Y (according to Popper’s ‘qualitative’ definition) if and only if every true statement following from Y also follows from X, and every false statement following from X also follows from Y; hence, under the hypothesis that X is at least as verisimilar as Y, it follows that all verified predictions of Y will be correct predictions of X, and all verified mistakes of X will also be mistakes of Y. Stated differently, the hypothesis that X is at least as close to the truth as Y allows to make the ‘second-order prediction’ that X tested predictions would always be at least as good as those of Y. Hence, the provisional corroboration of this ‘second-order prediction’ (that all empirical successes of Y have been to this date matched by X, and all empirical failures of X known to this date are also failures of Y), would count as a corroboration of the ‘second-order conjecture’ that asserts that X is closer to the truth than Y. In Popper’s own words:

I do not suggest that the explicit introduction of the idea of verisimilitude will lead to any changes in the theory of method. On the contrary, I think that my theory of testability or corroboration by empirical tests is the proper methodological counterpart to this new metalogical idea. The only improvement is one of clarification. Thus I have often said that we prefer the theory t₂ which has passed certain severe tests to the theory t₁ which has failed these tests, because a false theory is certainly worse than one which, for all we know, may be true. To this we can now add that even after t₂ has been refuted in its turn, we can still say that it is better than t₁, for although both have been shown to be false, the fact that t₂ has withstood tests which t₁ did not pass may be a good indication that the falsity-content of t₁ exceeds that of t₂ while ists truth-content does not. Thus we may still give preference to t₂, even after its falsification, because we have reason to think that it agrees better with the facts than did t₁ (Conjectures and Refutations, p. 235).

As it is well known, Popper’s dream of supporting his falsificationist method with a formal theory of approximation to the truth received a fatal blow when it was proved that no false theory could be logically related to any other theory in exactly the way Popper’s definition of verisimilitude demanded for the former being more verisimilar than the latter, and hence, the idea that we can ‘give preference’ to a falsified theory over another because of that reason became logically untenable. At least, of course, according to Popper’s definition of verisimilitude, for perhaps we could explicate the notion of ‘a theory being closer to the truth than another’ in different terms. The question, of course, would be whether according to these alternative definitions of truthlikeness, the methodological norms that would be efficient in selecting more truthlike theories would exactly be the norms Popper defended in his ‘proto-constitution’, or some different rules. As I have argued elsewhere, instead of starting by proposing a philosophical argument about what the goal of science should be, and logically derive from that goal the methodological norms that could better serve to its attainment, we might perhaps follow a more empiricist, or abductive, procedure, examining in the first place what methodological norms real scientists actually follow, and conjecturing afterwards which goal or goals could they be trying to pursue such that the norms actually followed in scientific practice are efficient procedures for the attainment of those goals. Perhaps we end finding out that real scientists actually follow under some circumstances what we could describe as ‘verificationist’ rules, and that this can be explained not because they are ‘irrational’, or ‘dogmatic’, but just because, in some definite sense, this is the best ‘science game’ they can play, taking into account both their goals and the limited resources they have at their disposal.

The latter discussion can also be connected with a different problem. Science is (mostly) not done by algorithms, but by human people, and every human person has usually more than one goal in everything she does. Even if the maximisation of testability and the pursuit of truth (of deep, useful, explanatory and interesting truths) were the only substantial goal of ‘science’ as an institution, the people that carry out the activities in which ‘science’ consist may have, and usually will have, other goals and values as well. These other aims and values will range from diverse preferences about the epistemic qualities of the theories, models, empirical procedures, styles of thought, etc., to more practical things like good work conditions, access to facilities, and, most importantly in the case of many scientists, prestige and intellectual influence. For example, perhaps the institutional arrangement that would maximise the testability of theories would be that all science were published in a strictly anonymous way, for this would probably avoid a lot of serious biases that are due to the ‘publish or perish’ tendency or current science. But, being humans as they are, under such institutional arrangement it would be likely that not enough people would opt for pursuing a science career. Many scientists’ thirst of glory would be hardly satisfied if science worked in such a way that no result could ever be declared ‘definitively established beyond all reasonable doubt’: for a seeker of scientific glory, provisionality and fallibility is very often just something we have to pay lip service to, and ‘firm verification’ of a theory or law is an institutional goal much more valuable than ‘having passed till now all tests but who knows what will happen tomorrow’. If scientists’ real motivations are close to this description, then establishing methodological rules that at least have a little verificationist flavour could be seen as a necessary price to pay for having a powerful science system at all, and hence, as an unavoidable element in any realistic constitution of the republic of science.