1
How do scientific claims relate to truth?
S. Orestis Palermos and Duncan Pritchard
Scientific realism
Often we refer to claims as being âscientifically proven.â For example, that the earth revolves around the sun, that in the centre of our galaxy there is a black hole, that water is H2O, and so on. âScientifically provenâ claims are taken to constitute paradigmatic instances of knowledge and they are held to be indubitable or absolute truths. A corollary assumption is that scientific knowledge is a cumulative body of knowledge, which grows simply by adding new theories to the existing body of âscientifically provenâ theories. On this view, scientific knowledge steadily progresses towards greater levels of understanding, bringing the human intellect ever closer to the true nature of the world.
Within philosophy of science, this view is known as scientific realism. It holds that well-confirmed scientific theories are approximately true and that the aim of science is to give a literally true account of the world. It is perhaps the most widely held view of the scientific progress, with part of its appeal coming from the fact that, if true, it would elegantly explain the success of the scientific enterprise. Yet the history and philosophy of science demonstrate that arguing in its support may not be as easy as one would have hoped for.
Logical empiricism
The first philosophers of science who attempted to argue that science is a cumulative body of proven knowledge that approximates truth are known as the logical empiricists. They were a group of young intellectuals, including Philipp Frank for physics, Hans Hahn for mathematics, Otto Neurath for economics, and the philosophers Moritz Schlick (who joined the group in 1922) and Rudolf Carnap (who joined in 1926). According to their view, scientific knowledge follows the method of inductivism: in this view, scientific theories are confirmed by inductive inferences (see induction) from an increasing number of positive instances to a universally valid conclusion. For example, Newtonâs second law seems confirmed by many positive instances from the pendulum to harmonic oscillators and free fall, among others. We can think of scientific theories as sets of laws of nature. Laws of nature are sentences that express true universal generalizations, and they take the form, âFor all objects x, if Fx then Gxâ (e.g., Newtonâs second law would read as follows: if an external force acts on a body of mass m, then the body will accelerate).
The logical empiricists held that true universal generalizations are confirmed when a sufficiently large number of positive instances (and no negative instances) have been found for them. In other words, induction was at the heart of the logical empiricistsâ criterion of verification (which is why proponents of this view are sometimes known as verificationists): a claim or statement is scientific if there is a way of empirically verifying it (i.e., if there is a way of finding positive empirical instances confirming that claim or statement).
The inductive methodology, however, is problematic on two grounds. First, it is too liberal as a method for demarcating good science from pseudo-science. Political theories such as Marxism or Freudâs psychoanalysis would equally meet the requirements of inductivism. A Freudian psychoanalyst could appeal to plenty of positive instances of peopleâs dreams that can confirm the validity of Freudâs analysis of the Oedipus complex, for example. But is this sufficient to license the scientific status of Freudâs psychoanalysis? Similarly, people that read horoscopes can claim that there are positive instances in their monthly working schedule confirming the horoscopeâs warning that it is going to be a very demanding working month for Aquarians! Does this mean that horoscopes are scientific?
Second, induction, although it may be a good mechanism for drawing inferences, falls short of supporting the logical empiricistsâ claim that scientific theories and the claims they support amount to proven knowledge. The problem is that induction cannot be given a non-circular justification. This is known as the problem of induction. Induction cannot be justified via deduction, since although a large observed sample (e.g., that every swan observed so far is white) might imply that the corresponding universal claim is true (e.g., that every swan is white), it does not deductively entail it (i.e., there might nonetheless be, say, a black swan). But one cannot non-deductively justify induction, either. Induction allows us to form beliefs about unobserved matters of fact on the basis of evidence provided by past and present observations. But in order for such inferences to be rational â such that they can amount to proven knowledge in the way the logical empiricists suggested â we need the further assumption that the future will resemble the past. The problem, however, is that this further assumption is circular, in that it relies for its support on induction itself. It is not a matter of logic that the future resembles the past, after all, and our only rational basis for this claim is that the future has previously resembled the past, but of course this basis is itself an inductive reason. Therefore, given that circular reasoning does not justify, there is no way of justifying our use of induction.
Against the backdrop of these doubts against the inductive methodology, logical empiricism appeared to lose the battle of demonstrating how science might amount to a cumulative body of proven knowledge that approximates truth.
Falsificationism
Not all was lost for the idea that scientific knowledge is a cumulative body of knowledge that gets closer to truth, however. Karl Popper â undoubtedly one of the most influential philosophers of science â attempted to demonstrate how the scientific method may still be deductively valid. He argued that, despite the apparent prevalence of induction within science, the validity of scientific theories does not originate from the validity of the inductive method but instead from what he called falsificationism.
Contra the logical empiricists, Popper thought that the distinctive method of science does not consist in confirming hypotheses, but in falsifying them, looking for one singular crucial piece of negative evidence that may refute the whole theory. According to Popper, science proceeds by a method of conjectures and refutations: scientists start with bold (theoretically and experimentally unwar-ranted) conjectures about some phenomena, deduce novel undreamt-of predictions, and then go about finding potential falsifiers for those predictions. Currently accepted scientific theories have passed severe tests and have survived, without being falsified yet. If a theory does not pass severe tests, and/or if there are no sufficient or suitable potential falsifiers for it, the theory cannot be said to be scientific. The history of science is full of theories that enjoyed a relative period of empirical success until they were eventually falsified and rejected: from the caloric theory of Lavoisier (which regarded heat as an imponderable fluid) to Stahlâs phlogiston theory in the eighteenth century, and to Newtonâs ether theory. Science has grown across centuries by dismantling and rejecting previously successful theories â scientific progress is characterized and made possible by falsification.
According to Popper, falsificationism is the distinctive method of science. It is a deductive (instead of inductive) method, whereby scientists start with bold conjectures, and deduce novel predictions, which then they go about testing. If the predictions prove wrong, the conjecture is falsified and replaced with a new one. If the predictions prove correct, the conjecture is corroborated and will continue to be employed to make further predictions and pass more tests, until proven wrong. In this sense, the logical validity of scientific theories does not originate from confirmation they receive from evidence but from surviving the empirical tests that could refute them. So long as a theory survives such tests, scientists may take their theory to be valid. If it fails, they can take it to be conclusively false. According to Popper, then, science does not approximate truth by having its theories confirmed, but by rejecting theories that have been proven wrong. In this sense, science tends to truth by avoiding falsity.
This was an ingenious attempt to save scientific realism. Reality, however, turned out to be much more complex than Popperâs simple deductive scheme. In daily laboratory situations, scientists never test a scientific hypothesis or conjecture by itself. Nor can they deduce any empirical consequence out of any bold conjecture either. This problem, known as the problem of auxiliary hypotheses, is the topic of the next section.
The problem of auxiliary hypotheses
Before Popper developed falsificationism as the method of science, the French physicist and scientist Pierre Duhem had already realized that no scientific hypothesis can be tested in isolation, but only in conjunction with other main theoretical hypotheses, plus some auxiliary ones. Consider Newtonâs law of gravity. Scientists never test the hypothesis of gravitation by itself, but always in conjunction with other theoretical hypotheses H1, H2, H3 (e.g., Newtonâs three laws of motion) plus some auxiliary hypotheses A1, A2, A3 (e.g., A1 says that the mass of the sun is much bigger than the mass of other planets; A2 says that no other force apart from the gravitational one is acting on the planets; A3 reports that planetary attractions are weaker than attractions between the sun and the planets). Now, suppose we deduce from this set of main and auxiliary hypotheses some observable evidence e and we proceed to test whether e occurs or not in nature:
H & H1 & H2 & H3 & A1 & A2 & A3 â evidence e
Suppose we find that e does not occur (or that the measured value for e is not what one would expect from this set of hypotheses). This would only indicate that there must be something wrong with the whole set of hypotheses:
| |
But we do not know whether it is H or H1 or H2 or H3 or A1 or A2 or A3, or any combination of any of these main and auxiliary hypotheses, which is refuted by the negative evidence.
Given this holistic picture, the process of refutation of a hypothesis can no longer be regarded as a one-to-one comparison between the hypothesis and a piece of evidence. Instead, it takes place through a variety of direct and indirect links across the entire web of knowledge. Thus, suppose again we get a piece of negative evidence. How can we know which element of the web, which particular belief, this negative piece of evidence is going to refute? If there is leeway in the way evidence relates to any particular belief in the web, how do we know what to change in the web itself?
The problem of auxiliary hypotheses presented a serious concern against Popperâs attempt to demonstrate how scientific knowledge may progress in a deductively valid manner. While Popperâs falsificationism provided an elegant solution to the problem of induction that was facing logical empiricism, the complexity of scientific theories and their interconnectedness with auxiliary hypotheses as well as other theories makes it impossible to apply the falsificationist method in a clear-cut manner. And even if it were assumed that elaborating further on falsificationism could provide a solution out of this problem, such that scientists could direct counterevidence to isolated hypotheses and selectively refute them, the emerging field of the history of science demonstrated that falsification hardly ever occurs in practice.
The structure of scientific revolutions
Problems such as the above made apparent the limits of experimental evidence and the impossibility of the very idea of a âcrucial experimentâ, able to establish one way or another the fortunes of any theory. Against this backdrop, in 1962, the US historian and philosopher of science Thomas Kuhn offered a highly influential, radically new conception of how science grows and unfolds, in a seminal book entitled The Structure of Scientific Revolutions.
Both the logical empiricists and Popper had thought of scientific knowledge as a largely incremental affair. As scientific inquiry proceeds, and new evidence is found, our scientific knowledge accumulates (by either inductively confirming or deductively falsifying theoretical hypotheses). In this way, we gradually acquire better and better scientific knowledge. Scientific progress would be secured by the right scientific method, which would deliver theories more and more likely to be true.
However, on the basis of his historical analysis, Kuhn argued that neither logical empiricism nor falsificationism work in practice. He noted that every scientific theory is born in an ocean of counterevidence and that it is only through time that novel theories produce corroborating evidence. If Popperâs falsificationism were correct, such that theories should be rejected as soon as they are met with counter-evidence, then science as we know wouldnât have existed.
Kuhn began his career in physics and attempted to provide a more realistic picture of the scientific progress that chimed with the historical data. During a postdoctoral position at Harvard, he had the chance to study and teach a course in the history of science dedicated to Aristotelian physics. The difficulty encountered in making sense of outmoded lines of reasoning had a profound influence in the way Kuhn came to rethink scientific inquiry as a non-cumulative process of knowledge acquisition, with no distinctive (inductive or deductive) method. Most importantly, it reshaped radically Kuhnâs view of scientific progress by rescinding the link between progress and truth, understood as the ability of a theory to capture things correctly.
Instead, Kuhn suggested that science is characterized by three-stage cycles of normal science, crises, and scientific revolutions. During normal science, a scientific community works on a well-defined scientific paradigm. Although Kuhn never defined exactly the notion of âscientific paradigmâ, he thought a scientific paradigm (or what he later called a âdisciplinary matrixâ) would typically include the dominant scientific theory, the experimental and technological resources, no less than the system of values of the community at a given time (e.g., how the community may value judgements of simplicity, accuracy, plausibility, and so on). In addition, a scientific paradigm includes also what Kuhn called âexemplarsâ, i.e., concrete solutions to problems that students encounter from the early stages of their scientific education, whether in laboratories, on examinations, or at the ends of chapters in science texts (1962/1996, Postscript, p. 187). Any scientific community in periods of normal science acquires its identity by working on an accepted textbook (be it Ptolemyâs Almagest, or Newtonâs Principia) and solving well-defined problems or puzzles within a well-defined textbook tradition. No attempt to test, falsify, or refute the accepted paradigm takes place during periods of normal science.
Only when a sufficiently large number of anomalies â which cannot be done away with â accumulate does the accepted paradigm undergo a period of crisis. In periods of crises, a new paradigm may come to the fore, and the crisis resolves into a scientific revolution when the scientific community decides to abandon the old paradigm and shift consensus around the new paradigm. Kuhn stressed how theory choice in these cases is not determined by the alleged superiority of the new paradigm over the old one. The consensus-gathering process is not determined by the new paradigm being more likely to be true or correct than the old one, but by the increase in the puzzle-solving power of the new paradigm. The new paradigm should be able to solve more puzzles than the old one, and thus Kuhn redefined scientific progress in terms of increased puzzle-solving. But this shift of focus from Popperâs falsification to Kuhnâs puzzle-solving has far-reaching implications for the rationality of theory choice.
Kuhn famously claimed that scientific paradigms (say, Ptolemaic astronomy and Copernican astronomy) are incommensurable. Incommensurability meant lack of a âcommon measureâ to evaluate two paradigms â in other words, lack of a common measure for rational choice between paradigms. Different paradigms use different scientific concepts, methodology,...