Student Colloquium

Underdetermination and patterns in data

Sarah Westh Krog

Supervisor: Samuel Schindler

Do empirical data uniquely determine the correct scientific theories? The so-called thesis of underdetermination suggests that this is not the case. But if this is correct, how do scientists pick the correct theory on the basis of the evidence? In one specific application of the underdetermination problem, it has been claimed that empirical data do not even determine data patterns (let alone theories). As the philosopher of science James McAllister has pointed out there are in principle infinitely many ways of decomposing any data set into the components of a regular pattern and a remaining noise term. The challenge now is: what are good reasons for regarding one rather than another pattern as real? McAllister concludes, radically, that there are no such reasons. In this talk I will introduce McAllister’s account as a version of the general underdetermination thesis, review McAllister’s arguments for his radical conclusion, and assess those arguments in the light of counterarguments.