Scholarly Knowledge in the Physical Sciences

the paradigm of the physical sciences, embodying the "scientific method" and self-consciously promoted by Galileo Galilei (1564–1642), has been, to a greater or lesser extent, followed in other areas of scholarship. It has three essential elements:

1. 1. Quantitative experimental input.

2. 2. Regularities observed or implied are expressed with simple mathematical models, and both input data and output theory are openly published to the general scientific community.

3. 3. Theories explaining the phenomena in question can be tested and falsified—this last piece emphasized especially by Karl Popper (1902–1994)—i.e., they can be proved wrong.

Socially, scholarship has been the province of individual investigators or small teams led by individuals, and this mode was well suited to the established method. Support for major projects (starting with Galileo and lasting until the second World War) was often provided by wealthy individuals. The overall enterprise has been phenomenally successful.

Now, to a rapidly increasing degree, funding is diffused among government, university, and private sources; and big data, often gleaned from commercial databases, is analyzed by large teams of scientists using very complicated computer models. In most cases, the complex mathematical elements of these models were designed and tested by specific noninteracting groups, so that even the leaders of large experimental projects, including the academic ones such as the Large Hadron Collider (LHC) or the Sloan Digital Sky Survey (SDSS), have a detailed understanding of only pieces of the overall program.

In addition, large-scale scientific modeling enterprises, such as the "Illustris Simulation" for the cosmological assembly of galaxy growth by merger, have been developed in which responsibility is diffused; the innumerable parameters essential to the modeling are determined by fitting to observations; and the model is judged an accurate or inaccurate representation of nature by the extent to which its output fits extant observations. In such efforts, much of the data on which the theories are based is not publicly available, and many of the computed consequences of the theories are also not published, so the empirical basis and output validity are difficult for the general scientific community to judge. Furthermore, since input parameters are adopted on the basis that the results fit observations, it is logically unsatisfactory to conclude that the theories have satisfied the test of congruence with reality when the results match observations. Still other enterprises, such as the fundamental physics based on string theories, have so many possible representations that no unique tests of validity have been found.

Thus, scholarly knowledge in the physical sciences has been increasingly departing from its (perhaps too) simple methodological origins, and, while some practitioners are aware and concerned about the changes, the vast majority of young scientists plow forward in whatever direction their mentors indicate, without much concern for the increasingly unstable foundations of their disciplines.

Some aspects of this issue, such as the growing complexity of funding and input tools, are unavoidable. But the practice of science would be improved if greater attention were given in the education of scientists to logic and to methodological issues. There is an increasing amount of attention being given to "reproducibility" of results in the sciences; this is laudable, but it approaches the issues from the endpoints only. Even if this effort bears success, it will leave the scientific enterprise still on shaky foundations, departing ever further from its enormously successful beginnings, since results can be reproducible but still empty (i.e., tautologically based on circular reasoning) or wrong (based on inconsistent algorithms or unexamined input data). More attention to the logical basis for the scientific method is required.