The Pessimistic Meta-Induction

In the ninth and final assignment in the series, we are asked to examine the concept of the pessimistic meta-induction - a philosophical argument against scientific realism that was popularised by Larry Laudan¹ in 1981. It essentially states the premise that if scientific theories that have been accepted in the past subsequently turn out to be wrong, we have no valid reason to believe that current theories are correct. In considering this question, we are asked to look at relevant contributions from a number of philosophers of science including (of course!) James Ladyman. The assignment questions are listed below, together with my answers.

In what sense is the Pessimistic Meta-Induction 'meta'? And why is it best construed as an argument for, as Ladyman puts it, atheism about unobservables, rather than agnosticism?

The term meta is a prefix that indicates that we are talking about a concept that is abstracted from another concept. As a software engineer, for example, I might refer to meta-data, which essentially means "data about the data". In database terms, for example, this would refer to information about the overall structure of the database, the size and type of the fields used in a database record, and so on.

In the context of scientific realism, or more precisely, the pessimistic meta-induction argument used by Larry Laudan against scientific realism, the term meta is used because Laudan is not referring in his argument to specific examples of theory change, but to "the evolution of science as a whole and its evolution in time".

Ladyman calls this an argument for atheism² (as opposed to agnosticism) about unobservables because it claims to show that a belief in unobservables is not justifiable, and that this has been demonstrated by the history of science. Laudan's argument is that there are a large number of theories, now abandoned, that once enjoyed both explanatory and predictive success (i.e. they were not just empirically adequate, but could also both explain phenomena and predict new phenomena).

On that basis, argues Laudan, it would have been perfectly reasonable for realists to believe that these theories were true (or approximately true) and that the unobservable entities to which they made reference did in fact exist. Most contemporary scientists, however, now believe these same theories to be inaccurate, and do not believe in the existence of the unobservables to which they refer.

We have therefore gone from a position in which we (as realists) believe in the truth of a theory and its unobservables to one in which we reject that theory and assert that its unobservables do not, in fact, exist. This is certainly an atheistic position; we are not merely undecided (as the agnostic is). We no longer believe. Laudan's argument appears to be that if we can reject a theory that was previously accepted as the truth once, there is no reason why we should not do it again. In doing so, he appeals to induction.

My interpretation would be something like this: "we have observed that in the past theories that were accepted as the truth were subsequently rejected and replaced by new theories, so (assuming that the future will be like the past) we can expect that at least some of the theories that we currently accept as the truth will eventually also be rejected and replaced by new theories." Therefore, according to Laudan's argument, it is not reasonable to claim that we should accept even our best current theories as the truth, or to believe that the unobservables to which they refer actually exist.

Why can the theoretical terms in science that allegedly refer to unobservables not be given an ostensive definition?

The term ostensive definition refers to something that is defined with reference to examples, and is often used when it is perhaps not so easy to define something in words (either because the concept in question is particularly difficult to understand, or maybe because the vocabulary required to satisfactorily explain the concept has not hitherto existed.

For example, if I wanted to define the colour "blue" to someone from another planet, I might point to a number of things that were blue in colour in order to get the idea across. It does of course rely on the ability of the person to whom I am trying to explain the concept to understand the overall context in which the explanation is given.

According to Ladyman, philosophers distinguish between the sense of a term (the ideas and description associated with the term) and its reference (the thing to which the term actually refers). Ladyman gives the example of a whale. Whales were once believed to be large fish, but are now known to be large aquatic mammals. Our sense of what a whale is has therefore changed, but we can still point to the same animal swimming around in the ocean and say "That's a whale!"³

Actually, I would probably say "Thar she blows!" and pretend to hobble around on one leg while wearing an eye patch and balancing a stuffed parrot on my shoulder . . . but that's just me. The point is, if we were only able to talk about the sense of something, and our ideas about it (i.e. the properties we supposed it to have) subsequently changed, then we would essentially end up talking about something different.

Fairly obviously, it is possible to create an ostensive definition in the case of an observable entity, because we can point to an example of such an entity and say "That's a [insert name of favourite observable entity]", and the reference can thus be "fixed", as Ladyman puts it.

The same cannot be said of unobservable entities (or maybe we should say "currently unobservable entities", if we allow that new technological discoveries may one day enable us to observe things that were hitherto unobservable). We cannot point to an electron and say "That's an electron." Hence, the term "electron", and indeed any other theoretical term used in science to refer to some unobservable entity, simply cannot have an ostensive definition.

We are thus limited, in our understanding of a term, to the sense of that term conveyed to us by the relevant current scientific theory. Somewhat unfortunately (from a realist standpoint) there are many cases where the sense of a particular entity – the atom, for example – has changed.

Then again, Putnam argues that most people have such a vague notion of the meaning of scientific terms that their concept of what the term "atom" means is not significantly affected by arguments over whether it looks like a plum pudding or is mostly empty space. Indeed, he argues that "Reference is fixed not by the description associated with a term but by the cause that lies behind the term's use [allowing for] continuity of reference across theory changes".⁴

Why does Laudan hold that successful reference of a theory's theoretical terms is not a necessary condition for the novel predicative success of the theory?

As mentioned in a previous answer, Larry Laudan's pessimistic meta-induction argument uses the history of science to undermine realist arguments (the inference to best explanation and no miracles arguments) that the truth of our best current scientific theories is warranted by their continued success in explaining observed phenomena, making novel predictions about previously unobserved phenomena, and explaining the progress of science generally.

There are of course many counter arguments from realists in response to the arguments put forward by Laudan, including the (possibly somewhat vague) claim that even theories that have been displaced by more contemporary versions were "approximately true". They have also suggested that only "mature" theories, and those enjoying novel predictive success, should be considered when making such comparisons.

In particular, many realists have focused on successful reference. A term can be considered to refer successfully "if there is something, or some things, which are picked out by it" (Laudan, obviously and probably quite rightly, argues that a theory cannot be true if this is not the case).

The effectiveness of Laudan's argument therefore rests to some extent on whether or not abandoned theoretical terms (from hitherto mature and successful theories) can be said to refer or not. If they don't, then the realist cannot really claim that they were even "approximately true", and the realist position is seriously undermined. Laudan has come up with a whole raft of abandoned but previously empirically successful theories, the theoretical terms of which, he claims, cannot refer to anything if their contemporary replacements are to be believed.

A number of the examples listed by Laudan have been dismissed by realists because they were not mature theories or because they did not enjoy novel predictive success, but there are some that qualify on both counts.

On this basis, Laudan appears to have shown that successful reference of a theory's theoretical terms is not a necessary condition for the novel predicative success of the theory, which appears to undermine both the realist claim that the explanation for empirical success is approximate truth (since successful reference of a theory's terms is a necessary condition for its approximate truth) and the no miracles argument (since it shows that a theory need not be true to be successful, and that if this is the case, the success of science generally does not depend on realism).

The realists respond to such arguments by claiming that abandoned theoretical terms do, in fact, refer (if you look hard enough and squint a bit), and that those that don't belonged to a non-essential part of the theory.

State the 'realism about what?' objection to scientific realism.

Ladyman essentially argues that (since we can't predict the future with absolute certainty) there is always the possibility that one or more of our best current theories will turn out to be in error, but that this possibility alone is not enough to threaten scientific realism. Van Fraassen⁵ has shown that it is not irrational not to believe in unobservables, but at the same time has to concede that belief in them is also not irrational. The main challenge to realism, then, appears to arise from an examination of the facts as they pertain to the history of science.

The "realism about what" objection to scientific realism is raised by Roger Jones⁶, who points out that alternative versions of physical theories have co-existed at various time (and continue to do so). Therefore, even if we believe in the truth of our best current theories, we are sometimes faced with the dilemma of choosing which truth to believe in. Classical mechanics, for example, is a mature scientific theory that has enjoyed almost unparalleled success in terms of making novel predictions.

Indeed, classical mechanics is still taught in our schools, and is still relevant for countless real-world applications. There are, however, several different versions of classical mechanics. For example, it can be formulated as a theory of forces acting at a distance between point particles, or as a theory that depends on the action of a gravitational field, or as a theory that works in accordance with the force law and laws of motion.

If classical mechanics was our current best scientific theory, we would be faced with having to choose between these apparently competing versions of it – hence the question "realism about what?". As Ladyman puts it, "each of these approaches suggests a quite different ontology and metaphysics".

Jones concludes that a physicist would therefore not be able to "articulate the 'fundamental (theoretical) furniture of the Newtonian universe'" if asked to do so. The realist could of course argue that Newtonian mechanics has been superseded by quantum mechanics and is therefore not a problem for scientific realism. The rather obvious response is that if we were to turn the clock back to the end of the nineteenth century, then the realists of the day do, in fact, have a problem.

Alan Musgrave⁷ gives a somewhat more robust response by arguing that the different formulations of classical mechanics are in fact different theories, not different versions of the same theory. He does concede, however, that they are strongly empirically equivalent, and are "more interesting than usual cases of underdetermination because they are 'embedded in the actual practice of scientists in a way that the artificial examples concocted by philosophers are not'". As Ladyman says, there is no easy solution to the problem that Jones poses.

What does a structural (scientific) realist believe?

According to Planck, "the structure of this physical world consistently moved farther and farther away from the world of sense and lost its former anthropomorphic character . . .",⁸ by which I assume he is saying that it is more and more the case that the terms used in our best current scientific theories refer to entities that are unobservable. Against this background, the two most compelling arguments, according to Ladyman, are the no miracles argument (in favour of realism) and the pessimistic meta-induction argument (against realism).

These competing viewpoints appear to have achieved a stalemate. Structural realism, introduced by John Worrall⁹ (although he attributes the concept to Poincaré), is an attempt to resolve this stalemate. Worrall uses as his example the theory change that occurred in the field of optics in the nineteenth century, from Fresnel's ether theory to Maxwell's electromagnetic field theory¹⁰. His claim is that, in the transition from Fresnel's theory to Maxwell's theory, there is "an important element of continuity" that amounted to more than just the carrying over of empirical content (although it did not constitute a carrying over of "the full theoretical content or full theoretical mechanisms".

Worrall's point is that we should not accept "full blown scientific realism, which asserts that the nature of things is correctly described by the metaphysical and physical content of our best theories". Instead, he advocates structural realism's emphasis on the "mathematical or structural content of our theories".

Worrall claims that "retention of structure across theory change" avoids the pessimistic meta-induction problem because it doesn't commit us to a belief in unobservable entities while at the same time asserting that the theory's structure alone is sufficient to describe the world, thus enabling us to make novel predictions and provide explanations without this seeming miraculous.

However, as Ladyman (more or less) says, structural realism needs a lot of work. Not just to convince the anti-realists, but probably to convince the realists as well, since it appears to require that they must concede rather a lot of ground. If I were to go along with structural realism at all, I think I prefer the way Howard Stein puts it:

"our science comes closest to comprehending 'the real', not in its account of 'substances' and their kinds, but in its account of the 'Forms' [theoretical structures] which phenomena 'imitate' [are represented by]".¹¹

This way of putting it just seems easier to understand – to me, at any rate, and represents what I would consider an acceptable compromise between two opposing viewpoints.

Notes:

1. Laurens Lauden (1941-2022) was an American philosopher of science who was highly critical of scientific realism. He expressed the pessimistic meta-induction concept in his 1981 paper "A Confutation of Convergent Realism", published in the journal Philosophy of Science (Vol. 48, No. 1, pages 19–49).
2. Ladyman describes Lauden's pessimistic meta-induction concept as an argument for what he refers to as an "atheistic scepticism" towards scientific realism (as opposed to a merely agnostic scepticism) in his 2001 paper "Science Metaphysics and Structural Realism".
3. In his book "Understanding Philosophy of Science" (2002), Ladyman uses the example of a whale to illustrate the fact that our sense of what something is should be considered distinct from a reference to that thing, i.e. our sense of a thing may change but our reference to it does not.
4. Putnam develops this argument in his essays "Explanation and Reference" and "The Meaning of 'Meaning'", published in 1973 and 1975 respectively.
5. See the article "Constructive Empiricism" in this section.
6. Roger Jones is a philosopher and physicist, and a former professor of philosophy at the University of Kentucky. His objections to scientific realism are expressed in his paper "Realism About What?" published in the journal Philosophy of Science (Vol. 58, No. 2, pp. 185–202).
7. Alan Musgrave (1940-2026) was a New Zealand philosopher of philosophy and head of the Department of Philosophy at the University of Otago for over forty years. His response to Roger Jones article "Realism About What?" (see above) is set out in his 1992 paper "Discussion: Realism About What?" published in the journal Philosophy of Science (Vol. 59, No. 4, pp. 691–697).
8. Max Planck (1858-1847) was a German theoretical physicist who formulated quantum theory and is considered to be one of the founders of modern physics. The passage quoted appears in his book "The Universe in the Light of Modern Physics" first published in 1929 and translated into English in 1931.
9. See the article "Inference to the Best Explanation" in this section.
10. Augustin-Jean Fresnel's mechanical wave theory of light, described in his work "Memoir on the Diffraction of Light" published in 1818, was based on the supposed existence of an invisible medium which he referred to as the "l'éther lumineux" (the luminous ether). His theory was widely accepted until James Clerk Maxwell laid out his electromagnetic field theory in his paper "A Dynamical Theory of the Electromagnetic Field" published in the scientific journal Philosophical Transactions of the Royal Society of London (Volume 155, 1865) and later expanded on and summarised the theory in his book "A Treatise on Electricity and Magnetism" published in 1873. Maxwell's theory explains how a self-propagating electromagnetic wave can travel through empty space without the requirement for a physical medium.
11. Howard Stein (1929-2024) was an American philosopher and historian of physics who taught at the University of Chicago. The passage quoted is taken from his 1989 paper "Yes, but... Some Skeptical Remarks on Realism and Anti-Realism".