Philosophy and Climate Science
Eric Winsberg
Reviewed by Greg Lusk
Philosophy and Climate Science
Eric Winsberg
Cambridge: Cambridge University Press, 2018, £71.99/£21.99
ISBN 9781316646922/9781108164290
To some extent, it is surprising that a book like this was not published sooner. As Eric Winsberg’s introduction points out, if one were to closely examine climate science and its cognate disciplines (for example, geophysics, oceanography, meteorology), one would observe scientists wrestling with philosophical questions regarding the representational capacity of models, the relationship between data and theory, and the proper use of probabilities in decision making. Since climate change entered the public consciousness in the late 1980s, such questions have taken on increasingly significant scientific and social importance. Yet, it has only been in the last decade or so that philosophers have given climate science the attention it deserves. Philosophy and Climate Science is an attempt to remedy this (relative) lack of philosophical attention, surveying and extending the broadly philosophical literature that does exist. In this aim, the book is a significant success.
But why now? There’s obviously no single explanation for why philosophy has become suddenly interested in climate science, but reasons abound: the pressing nature of climate change and the high level of political apathy make climate science an important domain of inquiry; the academy and the humanities have sought ways to become more socially relevant; scientists, because of internal controversy and external scepticism, have struggled to maintain credibility and produce a persuasive public message. Winsberg points to warming conditions in philosophy itself, melting the icy slopes of logic on which Western philosophers took residence in an attempt to remain politically neutral during the post-war period (p. 4). This philosophical thawing is attributed (in part) to the expanding influence of feminist philosophy of science, which is spurring a growing number of projects that engage with the practice of science and are socially relevant.
In this vein, the book is ‘for the benefit of everyone who wants to come down from the icy slopes’ and who prefers material ‘presented with real, living examples of scientific practice’ (p. 5). Winsberg hopes to appeal to the curious scientist, interested members of the general public, as well as students and scholars in the philosophy of science. As we will see, some of these groups will benefit from different parts of the book more than others. Broadly, the book can be divided into three parts: the first (Chapters 2–5) discusses the methodology of climate science and provides much of the necessary background for the rest of the book, the second (Chapters 6–9) examines uncertainty and interpretations of climate hypotheses, while the third part of the book (Chapters 10–13) focuses on epistemological issues. It concludes with a brief epilogue (Chapter 14) and an extended appendix on structural model error and the hawkmoth effect.
The first part of the book gives readers approaching climate science for the first time—or approaching philosophy of climate science for the first time—the tools necessary to appreciate the complexities that arise when establishing even the best supported scientific claims. The chapters survey the methodologies involved in gathering climate data, modelling, running computer simulations, and dealing with chaos. As is typical of such stage setting, these chapters move quickly and save much of the philosophical work for later chapters.
Nonetheless, the first part of the book has its highlights, including particularly lucid discussions of the modularity of climate models and chaos in climate science. Modularity—the degree to which subcomponents can be identified, validated, and maintained independently of other components—is often seen as a way to break down the complexity of a modelled system, helping to establish a simulation’s reliability. Winsberg demonstrates that software modularity is overemphasized and often exceeds the independence in the actual system. As evidence for this point, he employs empirical work by computer scientists that dissects typical climate simulation architectures in particularly interesting and clever ways. This is, admittedly, a small section, but it is an example of how informative empirical studies of science can be for philosophical understanding, which also nicely frames Chapter 10 on verification and validation.
Chapter 5 is one of the clearer explanations of chaos I’ve encountered in non-scientific discussions of climate science: it is accessible without oversimplifying the concept. This chapter is notable for incorporating discussion of a structural form of chaos—known as the hawkmoth effect—that has captured recent philosophical attention. The hawkmoth effect is an analogue of the well-known butterfly effect that arises from structural errors in the model rather than sensitivity to initial conditions. For those who want the technical details explaining the hawkmoth effect, these can be found the appendix. This chapter and the appendix are helpful for those interested in understanding current debates—in which Winsberg is a central actor—on the extent to which the hawkmoth effect undermines the decision relevance of climate models (see Frigg et al. [2014]; Winsberg and Goodwin [2015]; Nabergall et al. [forthcoming]).).
The primary concern of the second part of the book is how to interpret the probabilistic information that climate science produces, and how that probabilistic information should be used in decision making. Here, more than elsewhere, the chapters build upon each other. Chapter 6 examines probabilistic statements made by the Intergovernmental Panel on Climate Change (IPCC) and argues that probabilities are the best way to communicate knowledge under uncertainty from experts to policy makers, and that ‘subjective credences are the only probabilities that make any sense in climate science’ (p. 87). Chapter 7 argues that we should understand the probabilities given by the IPCC as group credences that summarize the group consensus on the range of what one’s degree of belief should be. Confidence statements regarding those probabilities should be interpreted as second-order probabilities, which are best expressed qualitatively. Chapter 8 details the difficulties using such probabilities in decision making. Chapter 9 is a very good survey, and clarification, of the literature examining how scientific conclusions and evidence evaluation in climate science are (non-epistemically) value-laden, and demonstrates the inadequacies of Bayesian responses to the argument from inductive risk.
The most interesting extensions of the existing literature come in the third part of the book. These chapters focus on model assessment and skill, robustness, diversity of evidence, and social epistemology, respectively. The jewel of this section is the extended analysis of robustness that spans Chapters 11 (‘Robustness’) and 12 (‘Diversity’).
Discussions about robustness analysis in the climate science literature are often framed using an apparent disagreement between Lisa Lloyd ([2009], [2015]) on one hand and Wendy Parker ([2011]) on the other. The simplistic story would be that these views are in opposition to each other: Lloyd supports the idea that robustness analysis can be used to support climate models and hypotheses, while Parker rejects it. But as Winsberg shows, this isn’t altogether an accurate picture. Parker in fact focuses on model robustness, arguing that agreement between sets of different climate models imparts no special epistemic significance on climate hypotheses. Lloyd, on the other hand, argues that climate hypotheses can be well supported by agreement between a variety of pieces of evidence, including results from climate models. The difficulty here is that robustness analysis is often cashed out in terms of model agreement; when it’s not, the kind of robustness provided by other detection types is thought to be different in kind from model robustness. It’s not just that Parker and Lloyd just have slightly different targets, but that the conceptions of robustness they deal with cut in different directions.
Winsberg argues that there is an account of robustness that can span both the model and non-model detection procedures, and thus he ends up showing that there are compatibilities between Lloyd’s work and Parker’s. Winsberg’s chosen account of robustness, borrowed from Schupach ([2018]), has two important aspects: it is hypothesis-relative, and the detection procedures involved must be diverse enough to rule out competing explanations for particular outcomes. The debate regarding robustness in climate science now comes into stark focus. Parker, Winsberg concludes, is mostly right in that the model ensembles in climate science tend to lack the necessary diversity required for robustness analysis for most hypotheses—though for some hypotheses, in rare instances, a new method known as emergent constraint reasoning can provide model-only robust assessments. Lloyd is also vindicated: model and non-model results can be used in conjunction to provide robust support for certain crucial climate hypotheses, as Winsberg shows through an analysis of the lower bound of equilibrium climate sensitivity (that is, the equilibrium change in air surface temperature resulting from a doubling of atmospheric CO2). Winsberg’s argument is a convincing reconceptualization of robustness analysis in climate science; I expect this will become the starting point for future conversations.
As with any survey with this breadth of scope, practical constraints and structural choices prevent discussion of all the relevant material in the domain. The structure of this book emphasizes philosophy’s potential contributions to climate science, rather than climate science’s potential impact on philosophy. I find few faults with the way Philosophy and Climate Science is put together. But for those interested in exploring the topic further, it is worth pointing out gaps, and mentioning other work not highlighted.
While the first part of the book does an adequate job of setting the stage for novice readers, it often moves quickly (especially for those approaching this literature for the first time), and opportunities to connect the climate-related material to contemporary philosophical discussions go unrealized. Chapter 2 is a good example: this chapter is partially divided by different types of evidence for global warming, with some types (for example, ocean heat) getting simply two sentences and a graph. Information about the Quine/Duhem problem, data modelling, and the data/phenomena distinction are sprinkled into the chapter, with the bulk of the philosophical discussion focusing on Lloyd’s ([2012]) account of the controversy between satellite remote sensing groups over tropical tropospheric warming.
All of this material is apt, but modern discussions of measurement—which take a model-based approach (for an overview, see Tal [2013])—are quite relevant here and would have situated the issues within contemporary discussions of data production. This could have tied in nicely with philosophical work about the use of weather simulations to homogenize observational data—a process called data re-analysis (see Parker [2017])—which is scientifically and conceptually interesting, but absent from this book. These are not crucial omissions, but they would have shown just how messy the use of data in climate science can be, and how reliant climate science is on modelling.
It is also worth pointing out that in rare instances, sustained discussion of alternative views is lacking. For example, the philosophically uninitiated may walk away from the second part of the book thinking that there is a philosophical consensus that probabilities interpreted as subjective credences are the best tool for climate science and communicating with policy makers. Winsberg’s approach is indeed a good way to understand the statements of the IPCC, which is his main concern. However, at least when it comes to assessing and interpreting the results from climate models, there is ongoing philosophical debate about whether probabilistic assessments are appropriate. For example, philosophers have developed possibilistic (rather than probabilistic) interpretations of IPCC confidence statements (see Betz [2010]; Katzav et al. [2012]; Katzav [2014]), which offers a way to communicate climate impacts via the ranking of possible outcomes. Explicitly examining alternative proposals would have better contextualized the author’s position and shown its benefits.
There is, as the cliché goes, something for everyone in Philosophy and Climate Science. Scientists unfamiliar with philosophical work focused on climate science and philosophers unfamiliar with climate science will be much better able to interact in an interdisciplinary way having read this book. The seasoned philosopher of climate science will benefit from the places where Winsberg extends the conversation, particularly the sections on robustness. The book is structured in such a way that it is well suited for graduate (or honours-level undergraduate) courses and would work best paired with many of the papers central to Winsberg’s analysis, helpfully included in a ‘further reading’ section at the end of each chapter. Ultimately, this work demonstrates how fruitful practice-oriented philosophy can be for climate science.
Greg Lusk
Michigan State University
greglusk@msu.edu
References
Betz, G. [2010]: ‘What’s the Worst Case? The Methodology of Possibilistic Prediction’, Analyse & Kritik, 32, pp. 87–106.
Frigg, R., Bradley, S., Du, H. and Smith, L. A. [2014]: ‘Laplace’s Demon and the Adventures of His Apprentices’, Philosophy of Science, 81, pp. 31–59.
Katzav, J. [2014]: ‘The Epistemology of Climate Models and Some of Its Implications for Climate Science and the Philosophy of Science’, Studies in History and Philosophy of Modern Physics, 46, pp. 228–38.
Katzav, J., Dijkstra, H. A. and de Laat, A. T. J. [2012]: ‘Assessing Climate Model Projections: State of the Art and Philosophical Reflections’, Studies in History and Philosophy of Modern Physics, 43, pp. 258–76.
Lloyd, E. A. [2009]: ‘Varieties of Support and Confirmation of Climate Models’, Proceedings of the Aristotelian Society, 83, pp. S213–32.
Lloyd, E. A. [2012]: ‘The Role of “Complex” Empiricism in the Debates about Satellite Data and Climate Models’, Studies in History and Philosophy of Science Part A, 43, pp. 390–401.
Lloyd, E. A. [2015]: ‘Model Robustness as a Confirmatory Virtue: The Case of Climate Science’, Studies in History and Philosophy of Science Part A, 49, pp. 58–68.
Nabergall, L., Navas, A. and Winsberg, E. [forthcoming]: ‘An Antidote for Hawkmoths: On the Prevalence of Structural Chaos in Non-linear Modeling’, European Journal for Philosophy of Science, 9.
Parker, W. S. [2011]: ‘When Climate Models Agree: The Significance of Robust Model Predictions’, Philosophy of Science, 78, pp. 579–600.
Parker, W. S. [2017]: ‘Computer Simulation, Measurement, and Data Assimilation’, British Journal for the Philosophy of Science, 68, pp. 273–304.
Schupbach, J. N. [2018]: ‘Robustness Analysis as Explanatory Reasoning’, British Journal for the Philosophy of Science, 69, pp. 275–300.
Tal, E. [2013]: ‘Old and New Problems in Philosophy of Measurement’, Philosophy Compass, 8, pp. 1159–73.
Winsberg, E. and Goodwin, W. M. [2016]: ‘The Adventures of Climate Science in the Sweet Land of Idle Arguments’, Studies in History and Philosophy of Modern Physics, 54, pp. 9–17.