Strategies for securing evidence through model criticism

Abstract

Some accounts of evidence regard it as an objective relationship holding between data and hypotheses, perhaps mediated by a testing procedure. Mayo's error-statistical theory of evidence is an example of such an approach. Such a view leaves open the question of when an epistemic agent is justif ied in drawing an inference from such data to a hypothesis. Using Mayo's account as an illustration, I propose a framework for addressing the justification question via a relativized notion, which I designate security, meant to conceptualize practices aimed at the justification of inferences from evidence. I then show how the notion of security can be put to use by showing how two quite different theoretical approaches to model criticism in statistics can both be viewed as strategies for securing claims about statistical evidence.

Physicality, Modeling and Making in a Computational Physics Class

“…To redesign our metamodel, we need a better understanding of what modeling phenomena entails. Fortu-nately, we will in the next section draw upon impressive philosophy of science work [17][18][19][20] that provides a sophisticated view of computational models. Note that while the question of what constitutes a model is certainly very complex (see for example [29]), adopting a pragmatic approach allows us to focus instead on how physicists engage in modeling activities in the production of computational physics knowledge.…”

“…As an answer this question, in this paper we will propose a metamodel -a model of modelling processes-of knowledge production in computational physics that will then be leveraged to design and analyze projects intended to empower students to cultivate these practices. Our work here is inspired by our reading of relevant literature from philosophers of science: Kent Staley provides an epistemological perspective on experimentation grounded in scientific practice [17][18][19] while Paul Humphreys offers a nuanced perspective on simulation and more generally on computational modeling [20]. We therefore define computation as much more than just a set of methods used to numerically solve problems that cannot be solved analytically but as a dynamic set of activities that are central and necessary to the production of scientific knowledge, and that are explicitly or implicitly performed by scientists when they successfully create physics knowledge.…”

The objectivity of Subjective Bayesianism

“…Originally, this definition was meant to apply in a wide sense, e.g., to invariance of a phenomenon or inference under different experimental designs and theoretical mod- Of course, robustness analysis in statistics is more general than this particular application (e.g., Huber, 2009;Staley, 2012), but frequentist theory mostly deals with distributional robustness: deviations from the assumed sampling distribution, such as violations of Normality or homoscedasticity. It does not involve robustness with respect to expectations on the size of the observed effect (though see Mayo and Spanos, 2006).…”

Computer models and the evidence of anthropogenic climate change: An epistemology of variety-of-evidence inferences and robustness analysis

“…The literature on this subject distinguishes several notions of robust reasoning pertaining to theorems, phenomena, modes of detection (Calcott, 2011;Levins, 1966Levins, , 1993Orzack and Sober, 1993;Wimsatt, 2001), inferences, measurements, derivations, causal relationships (Woodward, 2006), parameter values, mathematical structures, representation frameworks (Weisberg & Reisman, 2008), computer models and simulations (Houkes & Vaesen, 2012;Lloyd, 2015;Muldoon, 2007;Parker, 2011). Although distinctions among these ideas are philosophically interesting, for the purposes of this paper I will focus on a general sense of robust evidential reasoning, which 1 Also see Staley (2011Staley ( , 2012. 2 Staley also links this criterion to Campbell and Fiske's (1959) criterion of "discriminant validation.…”