Aarhus University Seal / Aarhus Universitets segl

Specialeeksamen: Martin R. Vase

Can Science Be Both Successful and False? An Anti-Realist Account of the Success of Science

21.06.2020 | Kristian Hvidtfelt Nielsen

Dato tir 30 jun
Tid 13:00 14:00
Sted D2 (1531-119)


In this thesis, I argue that scientific theories can be empirically successful while failing to latch on to the unobservable structure of the world. I aim to show that we therefore generally lack indicators of truth when it comes to unobservable entities. This claim conflicts with the position known as “scientific realism”, which I define as the view that science compels us to infer the likely truth of certain claims about the unobservable. I defend the rival anti-realist view that science compels us to believe only certain claims about the observable. The thesis is divided into three distinct papers. In the first paper I survey a common claim made by realists (including Peter Lipton, Stathis Psillos, and Philip Kitcher) that if we admit of inferences to observable entities we have to admit of inferences to unobservable entities. The reason is that there is no principled epistemic difference between these two types of inference. Since antirealists usually endorse the former type of inference they therefore have to accept, by parity of reasoning, the latter type of inference. I argue against this claim that there is in fact a principled epistemic difference, of which I gave a detailed account. In the second paper I survey the realist argument known as the “no-miracles argument”. This argument aims to show that if our theories are successful in a certain way, they are likely to be true. The aim of the paper is to give the best possible reconstruction of this argument, one that offers a challenge to anti-realists which they have not yet met. In the third paper, however, I respond to the no-miracles argument and meet the challenge introduced just before. In particular, I list certain features that make theories likely to produce novel predictions. I argue that these features are unrelated to truth and that the fact that they account for the novel-predictive success of theories implies that we cannot infer from this kind of success to the truth of a theory. In addition to these arguments I advance an “axiological” argument for antirealism in the introduction which says that if there is any room for doubt about our theories, we should not endorse realism because it will conflict with certain aims of science about which we can all agree. Since the three papers show that there is room for doubt, we should not endorse realism.