How does Science undergo theory change? (That is, what do philosophers of science say about how science undergoes theory change...?)

There are a number of answers to the question, but the debate can really be split into two camps: the evolutionary model of science (illustrated by the Logical Positivists and Popper) and the revolutionary model of science (illustrated by Kuhn, and in some ways, by Feyerabend).

For the Positivists, we change theories when our theory is less highly confirmed then another competing theory. This confirmation depends upon:

i) The quality of tests used to confirm it

ii) The variety of tests used to confirm it

iii) Agreement with accepted theories

The third criterion (agreement with accepted theories) is applied simply to weed out the number of ‘crazy’ theories so that science does not have to deal with every crackpot scheme that arises. Once a theory provides enough empirical data, however, the first two criteria take over, and confirmation depends upon testability. A theory must be testable to be confirmable. The problem with the idea of confirmatory tests (particularly with ‘crucial experiments’ that decide instantly between theories) is that interpreting a test as confirmatory to a particular theory already constitutes a hypothesis about what counts as confirmation or not (one that needs to be itself confirmed, and so on). In addition to this objection, a supposedly refuted theory can always offer ad hoc defenses for itself. It seems that the three criteria above are not enough to describe all cases of theory change. Thus, the Logical Positivists introduced a fourth criterion

iv) Simplicity

Which means that when we are required to decide between two theories that are equally confirmed in all other respects, we should choose the simpler of the two. But here we run into the same sort of objections that we saw with the idea of testability. We have to decide what constitutes simplicity, and justify it, and so on.

Karl Popper, rather than rely on the notion of confirmation, provides us with the idea of falsification. He states that a change in scientific theory relies upon instances of falsification. He rejects the problems inherent in ‘confirmability’ (a few of which are stated above) and says that we only change theories when we encounter a falsifying instance (an anomaly that the theory cannot explain). Popper says that we drop a theory and move on to a better one as soon as there is a falsifying instance.

This brings us to Kuhn. Kuhn believes that Popper is wrong and I concur. Historically there are innumerable instances where theories have encountered falsifying instances but moved on to become accepted. In fact, it seems that most new theories would be contradictory when they are first born. Galileo in particular contradicted almost every accepted physical principle of the time, yet his theory became widely accepted.

Rather than as a sequence of refuted theories, Kuhn sees science as a historical process that goes something like this:

Normal Science --- Anomalies Building Up --- Crisis --- Revolution --- New Normal Science…

Kuhn thinks that though falsification or confirmation may be useful within normal science (within a single accepted paradigm) neither can explain scientific revolutions. Kuhn believes that a scientific revolution occurs when anomalies within a particular paradigm build up to such an extent that it becomes unworkable, and a new paradigm appears then a new process of normal science occurs. Notice that I said that a new paradigm appears and not refutes. This is because Kuhn sees rival paradigms as incommensurable, and the process of revolution as one paradigm dying off, rather than being ‘defeated’ by its opposing paradigm.

So what criteria do we use to decide between theories? The answer is that there are no decisive rational criteria to decide between theories, because theory-choice is not simply a matter internal to science. To assume that theory choice was an entirely rational affair would be, for Kuhn, to assume that humans are entirely rational animals, which (as history shows us) is obviously false. Instead, people like Kuhn (and more vehemently, Feyerabend) think that we do allow external (social, economic, psychological, artistic, etc, etc) factors to affect theory choice. Examples of this abound in the history of science: the rhetoric/propaganda of Galileo, the discovery of the structure of DNA, the advent of quantum theory, etc, etc.

Personally, I think that the Feyerabendian picture of theory change is more realistic. It allows us to change our criteria based on the particular situation we are in. We are not limited by a nice little rationalist box, which limits our descriptive power greatly. Rather, Feyerabend allows us to describe any particular theory change based on its historical occurence. Rather than resorting to a given set of tools for every theory change, we create tools specifically for a particular description. Accepting Popper or the Logical Positivist’s would mean either rejecting a good portion of ‘science’ as non-scientific, or simply accepting our inability to describe it. But, if allow external factors to affect the course of science, we are able to describe a much, much larger portion of the history of science.

Keep in mind that these opinions on theory change are given by historians, philosophers and sociologists of science...