The Project

Coast and Geodetic Survey seismologist plotting earthquake epicenter.
Photo by NOAA on Unsplash

In history and philosophy of science, prediction has been debated since the nineteenth century. A key issue among historians and philosophers of science is whether the predictive capability of a scientific theory carries more epistemological weight in comparison to other virtues, such as coherence or explanatory power, the theory may display. ‘Predictivism’ answers this question in the affirmative and, thus, attributes a higher status to the prediction of novel phenomena than to the explanation of already known ones. Objections to predictivism have been raised on both philosophical and historical grounds. The main philosophical objection is that there is no essential difference, from an epistemological point of view, between prediction and explanation. Both confer the same empirical support to a theory. Furthermore, the history of science provides many examples where novel predictions did not play any special role in the establishment of scientific theories.

Another open issue is whether prediction carries special weight as regards the solution of practical problems. Does the predictive ability of a scientific theory or method guarantee its practical effectiveness regardless of its (lack of) explanatory power? And is it appropriate to let science policy decisions to be guided by successful predictions without an understanding of why these predictions are successful? In other words, can we trust a prediction without an accompanying explanation?

In a similar vein, computer simulations have promoted a ‘‘culture of prediction’’, where the extraction of predictions from a model counts as an important virtue, although computer simulations typically fail to represent accurately the processes that underlie the modeled phenomena. Does this preoccupation with prediction detract from the validity of the results produced by models and simulations? When we ask this question, though, we must keep in mind that in many contemporary sciences the use of mathematical and computational models is absolutely necessary for the implementation of highly abstract theories in the “real” world.

A further question is whether there are limits to what can be predicted by science. On the one hand, one might argue that, in principle, there are no unpredictable phenomena, since there is always the possibility that future scientific developments will allow the prediction of currently unpredicted phenomena. On the other hand, there might very well be phenomena that lie beyond the predictive ability of science. So, when is a phenomenon predictable and when is it unpredictable? In practice, an answer to this question is not only a matter of mathematical calculations, experimental measurements or observations. Rather it involves values, which in turn affect scientists’ attitudes towards prediction.

There are, in addition, questions that have yet to be addressed systematically, even though they are of fundamental importance: What exactly constitutes a prediction? When should a prediction be considered successful? A point of departure of the PYTHIA project is that an answer to these questions is all but given. Rather, what counts as a successful prediction is often up for grabs.

Finally, another dimension of prediction has to do with the unprecedentedly large scale of many scientific projects. In the era of Big Science, research projects often require substantial resources, huge human effort and long-term commitment. There have been cases where ambitious research programs were designed to meet a practical goal and proved to be extremely successful (e.g., the Human Genome Project); but, on the other hand, there have been programs that ended as striking failures (e.g., the “war on cancer” declared by President Nixon in the 1970s). Thus, being able to predict whether the outcome of a research program will be successful is of vital importance.

In considering the above issues regarding the perils of prediction, the project will focus on two questions: 1) how is the gap between high-level theory and predictions of particular phenomena bridged? 2) What counts as an adequate/successful prediction in different physical sciences? These questions will be investigated via case studies in four areas: seismology, high energy physics, quantum chemistry, and environmental science. A fifth case study on prediction in late 19th century physics will provide a long-term historical perspective on the topic.