2021_Metaphilosophy_Kent.pdf (1.61 MB)
Download file

Hypothesis-Driven Digital Philosophy of Science

Download (1.61 MB)
posted on 01.06.2021, 13:02 authored by Charles PenceCharles Pence

Recent trends toward both experimental philosophy and digital philosophy seem to have begun collapsing the distinction between scientific and philosophical methodology. In at least some cases, philosophers are turning toward empirical methods, in the process having to consider how those methods should be put in dialogue with traditional philosophical questions and approaches. In this talk, I want to pick up on an element of this methodological discussion in the context of digital philosophy of science.

Digital humanities (and, by extension, digital philosophy) is, at its heart, a “big data” discipline. It is not uncommon to perform analyses on dozens or hundreds of books, or thousands or tens of thousands of journal articles, in an effort to extract generalizations about scientific practice that can, in turn, be used to ground philosophical claims (Lean, Pence, and Rivelli forthcoming). This poses a variety of interesting problems, already familiar to any well-versed data scientist. Analyses performed on any dataset of such a size will offer significant problems of noise. A number of spurious correlations are certain to be present (for a humorous visualization of the problem, see https://tylervigen.com/spurious-correlations). Temptations to simply “read off” philosophical conclusions from the data themselves should be resisted; any philosophical conclusion of merit will require significant human interpretation (boyd and Crawford 2012).

One common proactive way to approach these problems and others like them is to turn toward hypothesis-driven research. Specific, testable, empirical hypotheses about the content of the literature can be evaluated with the aid of digital tools, if we clearly know what it would mean for them to succeed or fail. What is less clear, I think, are the answers two further questions. First, how does the evidence of the success or failure of such a hypothesis lead to confirmation or disconfirmation of generalizations about scientific practice, which are in turn the ingredients we hope to use to build novel claims in the philosophy of science? And second, what are the disadvantages of such a method? How can we balance the utility of these digital approaches as tools for discovery of unexpected trends in the sciences with the risk that we will simply see the connections that we want to see?

For all that these are widely acknowledged questions, the ways in which they might bear on digital philosophy of science are less clear. I hope in this talk to take some first steps toward an analysis of these issues, with the goal of offering insight for philosophers of science that is both theoretically grounded (in an understanding of the utility of empirical generalizations about scientific practice for the philosophy of science) and practically useful (in that it can give philosophers advice about how to formulate and use such hypotheses).

p { margin-bottom: 0.25cm; direction: ltr; color: #000000; line-height: 115%; text-align: left; orphans: 2; widows: 2; background: transparent }p.western { font-family: "Times New Roman", serif; font-size: 12pt; so-language: en-US }p.cjk { font-family: "Droid Sans Fallback"; font-size: 12pt; so-language: zh-CN }p.ctl { font-family: "Droid Sans Devanagari"; font-size: 12pt; so-language: hi-IN }a:link { color: #000080; so-language: zxx; text-decoration: underline }