This blog post explores why researchers trust specific hypotheses during the development of scientific theories and the criteria for their selection.
The philosophy of science seeks to answer questions such as: What constitutes science and what does not? What methodologies can scientists follow as science progresses? If such methodologies exist, what are they? And what mechanisms explain the history of scientific development? In modern philosophy of science, prominent attempts to address these questions include Popper’s falsificationism, Lakatos’s research programs, and Kuhn’s paradigm theory. Each offers an interpretation of science from their own perspective. While each approach successfully answers some questions, they also contain sufficient points of criticism to become subjects of debate and revision. Lakatos’s research programs are evaluated as partially resolving the problems of falsificationism while building upon that approach. However, they too possess methodological limitations. I realized that explanations of Lakatos’s research program and the limitations pointed out by Chalmers in ‘Modern Philosophy of Science’ did not include content related to researchers’ trust in theories that have failed many falsification attempts. This paper briefly explains the preceding theories, presents the possibility of modifying Lakatos’s research program, and then seeks to justify it.
The core idea in Popper’s falsificationism is that while experience cannot prove a theory to be true, it can prove it to be false. That is, even if a specific theory explains multiple observational results, the possibility remains that future observations contradict the theory. However, if observed results differ from those predicted by the theory, we can know the theory is false. Scientists accept a theory because attempts to falsify it have failed, and this acceptance is always provisional. Rather than merely confirming what is already known, evidence that something is decisively wrong provides more information for scientific progress and drives scientific advancement. According to falsificationism, scientific progress occurs when a hypothesis believed to be true is falsified, or when a hypothesis believed to be false fails to be falsified.
However, falsificationism has limitations because observations capable of falsifying a theory must themselves meet certain preconditions: the accuracy of the experiment, the reliability of the observational instruments, and the truth of the underlying theory upon which the observation is based. Since these premises themselves can be wrong, just like theories, the falsification of a theory must also be partially provisional. Furthermore, historically, even when observations occur that cannot be explained by a specific theory, they often do not lead to the theory’s falsification but are instead regarded as anomalous cases or observational errors. These are cases that cannot be explained by the “theory-rejection upon falsification” perspective advocated by falsificationism.
To address these issues, Lakatos and Kuhn historically examined the development of science and concluded that scientific progress should be understood not through individual hypotheses but through “structures.” That is, certain parts of a theory are fundamentally regarded as true by researchers; even if observations exist that do not match the theory, they are not seen as problematic. However, other parts can be modified as research progresses and observations that disagree with predictions emerge. These modifications must be independently verifiable. For example, in the case of Neptune’s discovery, when Uranus deviated from the orbit predicted by Newtonian mechanics, scientists explored the possibility that the assumption that no other celestial body existed beyond Uranus might be incorrect, rather than concluding that Newtonian mechanics itself was flawed.
Lakatos’ research program views a theory as a collection of hypotheses, designating the fundamentally true ones as the ‘hard core’ and the rest as the ‘pantheon’. Researchers follow two guidelines for modifying the pantheon to protect the hard core from falsification. One is negative discovery, which maintains the core as unfalsifiable. The other is positive discovery, which modifies the protective belt to explain observations and predict new phenomena. Kuhn’s paradigm is a concept similar to the research program, and developments within Kuhn’s period of normal science correspond almost equivalently to developments within Lakatos’s program.
The difference between Lakatos and Kuhn emerges in the shift from mainstream research programs or paradigms. Kuhn explains paradigm shifts within a historical context, incorporating interactions among groups of scientists. In contrast, Lakatos evaluates mainstream research programs based on the potential success of new predictions. Research programs capable of making new predictions grow into the mainstream, replacing programs that fail to predict. This can be seen as a loose form of falsification: a single counterexample does not necessitate the abandonment of a theory, but persistent failure to falsify leads to its abandonment.
The question is how long a research program that fails to make new predictions should be maintained. For example, the annual parallax predicted by Copernicus was not measured until the 19th century. There is also ample possibility that the protective shield of a regressive research program could be modified and transformed into a progressive one.
Ultimately, determining which research program is superior to another is only possible in hindsight. In ‘The Philosophy of Modern Science’, Chalmers points out that Lakatos’s research program has limitations when viewed as a methodology consciously followed by scientific researchers. He argues that it is necessary to distinguish between the rules of theoretical change that explain scientific progress and the norms that scientists actually follow.
Lakatos’s research programs were proposed as norms for discovery, but Chalmers sees limitations in this explanation.
Chalmers’s basis for criticism is as follows. First, Lakatos failed to provide criteria for selecting between competing research programs. Judging a research program as progressive is only possible in hindsight; it cannot serve as a criterion for a researcher choosing which program to pursue in the present. Second, it cannot be said that science developed as a result of scientists prior to Lakatos’s theory following his methodology.
Chalmers proposes modifying Lakatos’s research program by separating researcher choice from theoretical change. This prevents scientists’ choices from directly causing theoretical change, allowing theoretical change to be explained independently of researcher choice. He introduces the concept of “productivity” to express a research program’s potential for development. Since the degree of productivity of a specific research program is unknown to researchers, clear norms for theory selection cannot be provided. However, scientific progress can be explained by the assumption that programs with high productivity will gradually gain dominance within the scientific community over time.
I do not believe Lakatos’s methodology necessitates abandoning explanations of researcher choice. The limitations Chalmers identifies in Lakatos’s theory are refutable. While it is true that Lakatos failed to provide norms for theory selection and that the degree of productivity of a theory cannot be known at the time of research, these limitations can be overcome because theory selection need not be the correct choice leading to a successful research program.
Although it is a recently devised methodology, as Chalmers argues, if it can describe a common, consciously followed line of thought among scientists, it can become a methodology explaining researcher choice. If it is a guideline incorporating the scientist’s thinking for selecting successful theories, then trust in hypotheses that have withstood more attempts at falsification can be a crucial criterion for selection.
It is reasonable to trust hypotheses that have withstood multiple attempts at falsification over a longer period of research within scientific theories more than those that have not.
