The Unity of Robustness: Why Agreement Across Model Reports is Just as Valuable as Agreement Among Experiments

Abstract

A number of philosophers of science have argued that there are important differences between robustness in modeling and experimental contexts, and—in particular—many of them have claimed that the former is non-confirmatory. In this paper, I argue for the opposite conclusion: robust hypotheses are confirmed under conditions that do not depend on the differences between and models and experiments—that is, the degree to which the robust hypothesis is confirmed depends on precisely the same factors in both situations. The positive argument turns on the fact that confirmation theory doesn’t recognize a difference between different sources of evidence. Most of the paper is devoted to rebutting various objections designed to show that it should. I end by explaining why philosophers of science have (often) gone wrong on this point.

Robustness and the Event Horizon Telescope: the case of the first image of M87*

“…In section 4.1, we described how Staley (2020) has argued that varying modeling assumptions to represent experimenters' ignorance is a form of robustness analysis that increases the security of the conclusion. Additionally, Dethier (2020Dethier ( , 2022 explicitly argues for a unified account of robustness across experimental and modeling contexts. This seems to be based on a view of models that treats them as 'epistemic tools' (Dethier 2021) and centers their reliability or adequacy-for-purpose over their 'truth' (see also, e.g., Morgan and Morrison (1999) and Parker (2020)).…”

“…21 However, see Dethier (2022) for discussion of why these paradigmatic examples may be misleading: Roughly, pointing to an exemplary case of robustness analysis in an experimental context (Perrin) and a particularly challenging case for robustness analysis in a modeling context (climate models) biases us against acknowledging the potential confirmatory power of robustness analyses in the latter context. Doboszewski and Elder Philosophy of Physics DOI: 10.31389/pop.74…”

“…A similar analysis of robustness is given by Dethier (2020), who defines robustness as agreement across varied sources of evidence, where "variation in evidence" is understood as variation in "auxiliary conditions"-the "physical conditions and theoretical facts that must obtain" in order for a test to provide evidence for a hypothesis (74-75). 22 Note that Dethier explicitly defends a unified analysis of robustness across appropriately varied models and experiments-a position he calls "unity" (Dethier 2022).…”

Independent evidence in multi-messenger astrophysics

“…Wimsatt also points to Peirce [1868Peirce [ ] (1936 and to Whewell's "consilience of inductions" (via Laudan (1971)) as even earlier philosophical precursors. For more recent analyses of the evidential value of robustness, see e.g., Staley (2004), Dethier (2022), Schupbach (2018), Weisberg (2006), Lloyd (2015), and Woodward (2006). 17.…”

“…Second (and perhaps controversially) I endorse a unified view of robustness that treats agreement across model reports and experimental results in the same way. This is motivated by the idea that models, like measurements, provide a basis for model-based inferences, the results of which can constitute evidence for a hypothesis (Dethier 2022;Tal 2012;Staley 2004Staley , 2020. 18 In the cases I discuss in this paper, the results are robust (or not) across variation in observational methods, messengers, and the physical mechanisms of those messengers' emission.…”