Big Data and Causation – Reply to Serena Galli

Accounts of induction based on a difference making logic have a long history in debates on scientific method (e.g. Bacon 1620, Bk. II; Herschel 1830, P. II Ch. VI; Mill 1843, Bk. III Ch. VIII; Mackie 1967, Appendix; Baumgartner & Graßhoff 2004). This type of induction has been called variational or variative induction (Cohen 1989; cf. Russo 2007, 2009). I respond to objections raised by Serena Galli (2023) against the specific account of variational induction elaborated in Pietsch (2016; 2022, Chs. 5,6; a formal summary is provided in 2026a) and its application to the analysis of big data and machine learning practices (Pietsch 2021; 2026b). These objections primarily concern the directedness of causation and situations of causal underdetermination. While these present genuine, non-trivial problems, I argue that variational induction can address them, provided one accepts certain - admittedly controversial - assumptions.