Lakatos’s account of science can be understood as overcoming two problematic features of earlier accounts that we have examined. The first is the Duhem-Quine problem that beset the falsificationist theory. The problem that the Duhen-Quine problem raised for the falsificationist is that any falsifying observation applies to the entire theory-complex (the theory itself, the auxiliary assumptions, idealizations, etc.) and not the theory alone. The falsificationist is thus left without any rational way to decide between which part of the theory-complex to abandon. According to Lakatos, not all parts of science are on the same level, some are more central than others, and so when faced with some falsifying instance, we ought to modify the less central features of science instead of the more central ones.
The second is the charge of relativism that applied to Kuhn’s account of science through paradigm-shifts. According to Kuhn, there is no rational argument that can show one paradigm is better than any other. This has the consequence of making all scientific knowledge relative to the paradigm it is in and thereby leaving the idea of scientific progress across paradigms unexplained. According to Lakatos, science does operate under paradigm-like structures (Lakatos calls them “research programs”) and there are program-neutral criteria for explaining how science does progress.
One of the core ideas of Lakatos is the notion of a “research program”.
Definition 1 (Research program) A research program is a paradigm-like structure consisting of a hard core (the unmodifiable general hypotheses, fundamental assumptions) and a protective belt (the modifiable assumptions, laws, and initial conditions).
Example 1 (Hard core)
Example 2 (Protective Belt)
Lakatos’s theory of science adds guidelines for what scientists can and cannot legitimately do with the hard core and protective belt. What scientists can do is referred to as the positive heuristic; what scientists should not do is referred to as the negative heuristic.
Example 3 (negative heuristic)
Example 4 (positive heuristic)
One final feature of Lakatos’s account concerns evaluating different research programs. In contrast to Kuhn, Lakatos thought that science was progressive and scientific knowledge was not relativized to a paradigm (or in his vocabulary a “research program”). In order to support this claim, Lakatos needs some way to evaluate different research programs. That is, he needs a way to say that one research program is better than another. The key quality of a research program is how well it leads to (produce, yield) novel predictions.
Definition 2 (novel prediction) A prediction is novel if and only if when it is made it clashes with knowledge that is commonly accepted by scientists at the time.
That is, a good (or progressive) research program is one that (i) retains coherence by dealing with falsifying observations by modifying the protective belt and (ii) produces confirmed novel predictions. A bad (or degenerative) research program is one that (i) becomes increasingly incoherent and (ii) does not produce confirmed novel predictions.
In the previous section, we saw that Lakatos was able to fend off the Duhem-Quine problem by distinguishing the hard core of science from its protective belt. In contrast to the falsificationist, where a falsifying observation opened up the possibility of any feature of the theory-complex being modified, Lakatos contends that only the protective belt of science can be modified and only in a way that is sensitive to the positive and negative heuristics.
Lakatos also fended off the charge of relativism that besets Kuhn’s theory by appealing to his notion of a progressive research program. There is, however, a problem with Lakatos’s appeal to progressive research programs.
Objection 1 (New research programs are not always accepted as the result of novel predictions.) One problem with making novel predictions the hallmark of a progressive research program is that, as Chalmers (What is this called Science?, p.129) writes, “counter examples involves situations where the worth of a research program is demonstrated by its ability to explain phenomena that at the time were already well established and familiar, and so not novel”.
Example 5 (retrograde motion) One advantage of the Copernican system is not that it generates novel predictions but it more smoothly explains certain astronomical phenomena. One example is retrograde motion of the planets. Retrograde motion was explained in the past by claiming that planets orbit the earth and undergo epicycles, whereas the Copernican account can explain retrograde motion as simply perspective effect.
Objection 2 (Modifying the hard core.) According to Lakatos, the hard core of a research program cannot be modified. But, research programs adjust the hard core at times.