Alai, Mario
(2021)
HOW SHOULD WE JUDGE CURRENT SCIENTIFIC THEORIES?
[Preprint]
Abstract
The scientific realism-antirealism debate concerns theories in general. However, as soon as the discussion focuses on the historical development of science, some issues emerge concerning how we should regard current theories in particular, as opposed to past or future ones.
Positions here range between two extremes: on the one hand a radical version of the pessimistic meta-induction (PMI) would have it that since all past theories older than 100 - 150 years have been proven radically false and rejected, also present and future theories will be rejected within 100-150 years, and science can offer us no truth (but, at most, empirical adequacy).
On the opposite extreme blunt optimists like Doppelt (2007, 2011, 2014) or Park (2017) hold that, given the astonishing qualitative and quantitative technical and methodological progress of research in the last century or so (Fahrbach 2011), current best theories are almost completely and exactly true, so that further progress, besides adding new knowledge, can at best correct minor details of present day theories.
But evidence is against both extremes: radical pessimism cannot explain why even the ancient and now rejected theories were predictively successful. But an explanation of those successes is available: those theories had some true components which (a) by themselves were sufficient to derive their predictions, and (b) have actually been located by subsequent research. Therefore past theories were not completely false; hence, even the PMI cannot show that current theories are completely false. Again, radical pessimism cannot explain the rapidly increasing rate of success of science; which instead is plainly explained by the fact that the true components of older theories are typically preserved in current ones.
On the other hand, it is a priori implausible that just now we have reached the “end of history” in scientific research, a sort of promised land of pure truths, or Peirce’s ideal limit of research, and that our science is infallible. We see the mistakes of past science, while we obviously cannot see those of present science, and this ingenerates the illusion that there are none. But the same illusion arose in the past, age after age, only to be deluded. In addition, we positively know that there are mistakes in current theories because even two of the most successful ones, quantum mechanics and relativity, are at variance with one another and are beset by unsolved riddles.
No doubt contemporary science has made astonishing empirical and methodological progresses with respect to past science. But so did also modern science with respect to medieval science. Empirical knowledge and scientific methodology had improved a lot from 1000 AD to 1700 AD, yet many wrong theories were still held at that date, and even thereafter. So, it is hard to think that any improvement of our background empirical knowledge and methods can at some point make scientists practically infallible, and it is even harder to think that this point has already been reached.
Brad Wray (2016) pointed out that, just like at any time there are unconceived alternatives to the current theories, there are also unconceived methods (and instruments), by which those theories could be overthrown, like the discovery of the astronomical telescope helped overthrowing geocentrism, the discovery of the microscope jeopardized the theories of spontaneous generation, etc. Such new methods and instruments are produced all the time by the very progress of science, so the very advancements of contemporary science allow to suppose that many new methods and instruments will be discovered also in the next future, undermining today’s theories and opening the way to new revolutions.
The arguments against both extremes are all sound, but they are not mutually incompatible! Together, therefore, they drive to a moderate intermediate position: current theories are partly true, in fact more (perhaps much more) largely true than past ones, yet they probably still include important false components which might be replaced by future revolutionary changes. Contrary to Kuhn’s view, revolutions and progress go hand in hand.
Of course current science stands on much firmer grounds than pre-20th Century science, yet there are continuities which justify some inductive inferences from past to present and future: we are still humans, basically using the same cognitive tool (reason and the five senses) and subject to the same cognitive limits; scientific method is basically the same; above all, Nature is still very complex, in fact unfathomably complex: it works in different ways at different scales or different locations in space or time. For instance, it is (roughly) deterministic at large scales, but indeterministic at small scales; the physical laws today are probably different from those a few instants after the Big Bang; entropy increases over time in the universe as a whole, but it may decrease in local areas or over short time spans; etc. (Alai 2017, 3282). Every great advancement in science has shown us unsuspected deeper and more basic layers of the structure of Nature, and we don’t know how many of those still lie ahead. Any of those discoveries brought out some basic mistake in our understanding of some of nature’s mechanisms, so spurred some kind of revolution.
Induction may be correctly applied to past science, on condition of taking a correct image of past science as a premise. If this is done, the conclusion is a more balanced judgment of current and future theories, neither completely pessimistic nor implausibly optimistic. What we observe in past science is that (1) every single theory has been found to be mistaken and replaced; yet (2) mistaken but predictively successful theories had some true components (those essential to deriving their successful predictions); (3) those components were typically preserved in replacing theories, which therefore were more largely true.
Since there is no reason to deny that this is still the trend is still, we should conclude that current theories are more largely true than earlier ones, but still partly false. Even if probably not all the content of current theories exceeding old ones is true and will be preserved in the future, still we can appreciate one by one the new pieces of information and the corrections brought up by current theories, and we see they are really many.
However, we cannot tell what in our theories is the percent of truth vs. falsity. A fortiori, therefore, we cannot tell how much more largely true than past theories they are, i.e. measure the difference in the respective percents of truth. Even less, of course, can we tell what percent of the whole truth on its particular subject a theory has gotten, since we don’t know what the whole truth is.
It might be suggested that if the assumptions which were essential in deriving novel successful predictions are most probably true, as selective realists believe, then we should be able to discriminate what is true and what is false in our theories. However, this is not the case for two reasons. First, even assumptions which have not been so essentially employed might be true: only, we don’t have the acid test of it.
Second, essential hypotheses typically feature in the derivation of novel predictions as undistinguished parts of stronger hypotheses which play the official role in the derivation; the latter are not essential, because the prediction might equally well have been derived from their essential part, even if this typically goes unnoticed. Furthermore, it can be argued that the essential content of a hypotheses often can be distinguished from its non-essential content only negatively, retrospectively and in hindsight (Alai 2021). Therefore, looking at current theories we are entitled to believe that there is some truth in them, and more precisely that there is some truth in some hypotheses which appear to have been essential to certain novel predictions; however, we cannot be certain of what exactly is true in them. If we were, that would make our heuristics much easier than they are, for in face of any experimental failure of a theory we would know precisely and in advance which parts of it should not be modified and substituted.
A further difficulty is that talk of hypotheses and “parts” of theories here is vague and rather metaphorical, and there may be different ways (none easy, anyway) to cash it out literally. For instance, suppose we formalize theories as collections of sentences, as in the classical “statement view”: then, it might be the case that 95% of the empirical atomic sentences of T are correct, that 50% of its middle-level theoretical atomic sentences are correct, but that 90% of the few very basic atomic theoretical sentences are wrong. If so, it might be matter of taste whether to call T largely true or largely false, but it would be certainly correct to call a “revolution” the substitution of T by a theory T’ which preserved most of the empirical and middle-level theoretical sentences of T while substituting 90% of its very basic theoretical sentences.
Now, nothing allows us to exclude that a number of our best theories today is in a position like T. From this and the previous arguments it follows that it is quite possible, in fact rather probable, that our science, successful and largely true as it is, will undergo a number of revolutions in the future.
Monthly Views for the past 3 years
Monthly Downloads for the past 3 years
Plum Analytics
Actions (login required)
 |
View Item |
![[feed]](http://philsci-archive.pitt.edu/style/images/feed-icon-32x32.png){kind=icon}