When scientists seek further confirmation of their results, they often attempt to duplicate the results using diverse means. To the extent that they are successful in doing so, their results are said to be robust. This paper investigates the logic of such "robustness analysis" [RA]. The most important and challenging question an account of RA can answer is what sense of evidential diversity is involved in RAs. I argue that prevailing formal explications of such diversity are unsatisfactory. I propose a (...) unified, explanatory account of diversity in RAs. The resulting account is, I argue, truer to actual cases of RA in science; moreover, this account affords us a helpful, new foothold on the logic undergirding RAs. (shrink)

Climate change fingerprint studies investigate the causes of recent climate change. I argue that these studies have much in common with Steel’s (2008) streamlined comparative process tracing, illustrating a mechanisms-based approach to extrapolation in which the mechanisms of interest are simulated rather than physically instantiated. I then explain why robustness and variety-of-evidence considerations turn out to be important for understanding the evidential value of climate change fingerprint studies.

Many philosophers have claimed that evidence for a theory is better when multiple independent tests yield the same result, i.e., when experimental results are robust. Little has been said about the grounds on which such a claim rests, however. The present essay presents an analysis of the evidential value of robustness that rests on the fallibility of assumptions about the reliability of testing procedures and a distinction between the strength of evidence and the security of an evidence claim. Robustness can (...) enhance the security of an evidence claim either by providing what I call second-order evidence, or by providing back-up evidence for a hypothesis. (shrink)

Philosophical perspectives on numerical models help us to understand concepts, but will not predict the climate in the future. Studying climate model results in isolation on the other hand may seduce us to believe what we simulate will actually happen. A model is neither correct nor wrong as such; it is simply more or less useful as a representational tool for a certain purpose. I argue that process understanding is the key to make judgments about when this tool is adequate (...) for insight relevant to certain aspects of the real world. It is only through understanding the relationships in components and variables of the climate and their representation in models, combined with understanding what our models are supposed to do, that we can make better use of them. (shrink)

In this paper, we explore the extent to which issues of simulation model validation take on novel characteristics when the models in question become particularly complex. Our central claim is that complex simulation models in general, and global models of climate in particular, face a form of confirmation holism. This holism, moreover, makes analytic understanding of complex models of climate either extremely difficult or even impossible. We argue that this supports a position we call convergence skepticism: the belief that the (...) existence of a plurality of different models making a plurality of different forecasts of future climate is likely to be a persistent feature of global climate science. (shrink)

Today's climate models are supported in a couple of ways that receive little attention from philosophers or climate scientists. In addition to standard 'model fit', wherein a model's simulation is compared to observational data, there is an additional type of confirmation available through the variety of instances of model fit. When a model performs well at fitting first one variable and then another, the probability of the model under some standard confirmation function, say, likelihood, goes up more than under each (...) individual case of fit alone. Thus, two instances of fit of distinct variables of a global climate model using distinct data sets considered collectively will provide stronger evidence for a model than either one of the instances considered individually. This has consequences for model robustness. Sets of models that produce robust results will, if their assumptions vary enough and they each are observationally sound, provide reasons to endorse common structures found in those models. Finally, independent empirical support for aspects and assumptions of the model provides an additional confirmational virtue for climate models, contrary to what is implied by some current philosophical writing on this topic. (shrink)

Most recent philosophical thought about the scientific representation of the world has focused on dyadic relationships between language-like entities and the world, particularly the semantic relationships of reference and truth. Drawing inspiration from diverse sources, I argue that we should focus on the pragmatic activity of representing, so that the basic representational relationship has the form: Scientists use models to represent aspects of the world for specific purposes. Leaving aside the terms "law" and "theory," I distinguish principles, specific conditions, models, (...) hypotheses, and generalizations. I argue that scientists use designated similarities between models and aspects of the world to form both hypotheses and generalizations. (shrink)

I introduce a new method for validating models—including stochastic models—that gets at the reliability of a model’s predictions under intervention or manipulation of its inputs and not merely at its predictive reliability under passive observation. The method is derived from philosophical work on natural kinds, and turns on comparing the dynamical symmetries of a model with those of its target, where dynamical symmetries are interventions on model variables that commute with time evolution. I demonstrate that this method succeeds in testing (...) aspects of model validity for which few other tools exist. (shrink)

Lloyd (2009) contends that climate models are confirmed by various instances of fit between their output and observational data. The present paper argues that what these instances of fit might confirm are not climate models themselves, but rather hypotheses about the adequacy of climate models for particular purposes. This required shift in thinking—from confirming climate models to confirming their adequacy-for-purpose—may sound trivial, but it is shown to complicate the evaluation of climate models considerably, both in principle and in practice.

I propose a distinct type of robustness, which I suggest can support a confirmatory role in scientific reasoning, contrary to the usual philosophical claims. In model robustness, repeated production of the empirically successful model prediction or retrodiction against a background of independentlysupported and varying model constructions, within a group of models containing a shared causal factor, may suggest how confident we can be in the causal factor and predictions/retrodictions, especially once supported by a variety of evidence framework. I present climate (...) models of greenhouse gas global warming of the 20th Century as an example, and emphasize climate scientists’ discussions of robust models and causal aspects. The account is intended as applicable to a broad array of sciences that use complex modeling techniques. (shrink)

The present paper draws on climate science and the philosophy of science in order to evaluate climate-model-based approaches to assessing climate projections. We analyze the difficulties that arise in such assessment and outline criteria of adequacy for approaches to it. In addition, we offer a critical overview of the approaches used in the IPCC working group one fourth report, including the confidence building, Bayesian and likelihood approaches. Finally, we consider approaches that do not feature in the IPCC reports, including three (...) approaches drawn from the philosophy of science. We find that all available approaches face substantial challenges, with IPCC approaches having as a primary source of difficulty their goal of providing probabilistic assessments. (shrink)

Many climate scientists have made claims that may suggest that evidence used in tuning or calibrating a climate model cannot be used to evaluate the model. By contrast, the philosophers Katie Steele and Charlotte Werndl have argued that, at least within the context of Bayesian confirmation theory, tuning is simply an instance of hypothesis testing. In this paper I argue for a weak predictivism and in support of a nuanced reading of climate scientists’ concerns about tuning: there are cases, model-tuning (...) among them, in which predictive successes are more highly confirmatory of a model than accommodation of evidence. (shrink)

This article identifies conditions under which robust predictive modeling results have special epistemic significance---related to truth, confidence, and security---and considers whether those conditions hold in the context of present-day climate modeling. The findings are disappointing. When today’s climate models agree that an interesting hypothesis about future climate change is true, it cannot be inferred---via the arguments considered here anyway---that the hypothesis is likely to be true or that scientists’ confidence in the hypothesis should be significantly increased or that a claim (...) to have evidence for the hypothesis is now more secure. (shrink)

This chapter examines the role of parameterParametercalibrationCalibration in the confirmation and validation of complex computer simulation models. I examine the question to what extent calibration data can confirm or validate the calibrated model, focusing in particular on Bayesian approaches to confirmation. I distinguish several different Bayesian approaches to confirmation and argue that complex simulation models exhibit a predictivist effect: Complex computer simulation models constitute a case in which predictive success, as opposed to the mere accommodation of evidence, provides a more (...) stringent test of the model. DataData used in tuning do not validate or confirm a model to the same extent as data successfully predicted by the model do. (shrink)