Empirical evidence in conceptual engineering, or the defense of 'predictive understanding'

Abstract

The advent of digital philosophy of science—the practice of employing large textual datasets together with text mining and natural language processing tools for asking questions in philosophy of science—has dramatically shifted philosophers’ ability to account for the scientific practice, without relying on small and often arbitrary samples of the literature. The methodological issues that come with this novel approach have been traditionally discussed around the necessary choice between using these methods for discovery and exploration of new philosophical hypothesis, and the testing of existing stances. In this chapter, we propose a different path that remains hitherto underexplored in digital philosophy: namely, one offered by conceptual engineering. We provide general methodological guidelines for using digital tools in the study of scientific literature for the purpose of clarifying and honing existing philosophical concepts. To substantiate this position, we draw examples from existing research which has hinted at this approach, and apply our proposed methodology to the notion of “understanding” in the scientific practice to show how it can help researchers across subdisciplines of philosophy apply this concept usefully, including in the philosophy of artificial intelligence.