Causal Reasoning in Physics
Jenann Ismael · The Philosophical Review · 2016
In 1913 Bertrand Russell argued that contrary to what most of us presume, the causal structure of the world is not fundamental, that is, that there is no temporally asymmetric relation of causal determination, production, or influence built into the fundamental fabric of reality. His argument was that causal notions had been eliminated from physics in favor of time-symmetric, fundamental laws. Causation should be pushed aside as a relic, he said, of a bygone era. Cartwright's (1983) influential critique of Russell's position convinced many that causal eliminativism, in the form that Russell defended it, is not supportable, because causal structure plays a functional role in practical reasoning that cannot be played by nomological relations. Causal pathways identify strategic routes to bringing about ends, and a law-like relationship between A and B does not entail that A-ing is a way of bringing B about. So, for example, bad breath is correlated with tooth decay as a matter of physical law (since both arise with the presence of bacteria in the mouth), but taking breath mints is a not a strategic route to preventing tooth decay.By making it clear that what Russell identified as the physical replacements could not play the functional role of causal beliefs in practical reasoning, Cartwright renewed interest in causal relations. There have been important developments in recent decades in understanding causal relations. There has been progress in understanding what causal claims add to structures like laws and probabilities. Formalisms have been developed for representing causal claims in science. Methods of causal inference and discovery have been developed. But among physicists and philosophers of physics, the dominant position on the ontology of causal relations remains that—contrary to what philosophers and scientists took for granted for centuries—causal structure is not part of the fundamental fabric of the physical world. This is the position that Frisch calls the “anti-causalist” position. His book is intended as a counterpoint to the prevailing anticausalism. His aim is, in his own words, “to show that, contrary to what appears to be the received wisdom among philosophers of physics, causal structures play a legitimate role in physics” (21).By causal structures, Frisch means asymmetric structures represented by directed acyclic graphs (DAGs) that play the practical role that Cartwright identified, indicating strategic pathways for bringing about ends. Frisch has been a gadfly in the discussion of causation in the foundational literature in physics. This book brings together arguments that he has been advancing against particular anticausalist programs into a single, sustained critique of the anticausalist position. It is the most careful and comprehensive treatment of causal reasoning in physics to date. The first chapter is an introduction to the historical backdrop, separating the strand of discussion in the philosophical literature that he is contributing to from related philosophical projects, and setting up the discussion that follows. Chapter 2 sketches a general pragmatic account of scientific representation. The generalized pragmatism is an undercurrent running through the whole text. It tempers too strong a metaphysical reading of Frisch's conclusions. The next two chapters give a systematic survey of anticausalist arguments found in, or suggested by, the literature. These arguments point to features that make causal representations applicable in the special sciences but are purportedly absent from physics. Chapter 3 considers arguments that causal representation depends on coarse graining or a distinction between causes and background conditions. Frisch argues that these are both in fact features of representation in physics as well. Chapter 4 considers arguments that draw on the interventionist account of causal structure. According to interventionists, causal claims are to be understood in terms of their implications for the results of hypothetical interventions, and they have argued that since our fundamental theories deal with the universe as a whole, the notion of an intervention is inapplicable, so causal relations simply do not arise. Here Frisch responds that even an interventionist can make sense of causal claims applied to the universe by generalizing the notion of intervention to apply to hypothetical interventions that set the initial state of the world at some forced value.Chapter 5 addresses the most prominent and influential anticausalist argument: the contrast between the asymmetry of the causal relation and the symmetry of the fundamental laws. Frisch argues that standard patterns of reasoning even in fundamental physics depend on asymmetric causal assumptions, as a matter of practical necessity. Local inferences from known data employing only laws are too weak to draw even rudimentary conclusions, so physical inferences rely on patterns of causal reasoning. Chapters 6 and 7 examine two examples of time-asymmetric causal reasoning in physics: linear response theory and classical radiation theory. Frisch shows that there are different ways to express the causal assumptions implicit in the inferential patterns that he exhibits: as a causal Markov condition (which states that any node in a DAG is conditionally independent of its nondescendants, given its parents), or as an assumption of microscopic randomness in the initial—but not the final—state of a modeled system. Which of them is to be thought of as the “base” or “root” assumption, that is, the one that supports all of the others and explains why they hold? Frisch views them as two sides of the same coin. For him, these two assumptions are interchangeable with one another, and also with the claim that the world is causally structured.1I found the attempts to articulate arguments that physics is incompatible with causal representation and physics strained. The thought of the anticausalists was not that physics is incompatible with causal representation but that we don't find causal structures at the fundamental level of physical description. The most valuable part of the book is the demonstration that causal reasoning plays an indispensable role in many contexts, even in fundamental physics. That lesson should remain at the forefront of the discussion of the status of causal notions.In foundational discussions in philosophy of physics surrounding causation, almost everyone accepted Cartwright's observation that causal beliefs play an indispensable role in practical reasoning. The question for neo-Russellians in the wake of Cartwright's critique is what physical facts (about the world, on the one hand, and ourselves, on the other) support the formation and use of causal beliefs? Chapter 8 is a discussion of the two most influential forms of neo-Russellianism: (i) the views of Barry Loewer and David Albert, which link causal structure to the thermodynamic gradient, and (ii) the views of Huw Price and James Woodward, which link causal notions to agency and manipulation. On a neo-Russellian view, causal structure isn't eliminated but demoted from the fundamental ordering principle of the world to a network of emergent relations exploited by situated agents in strategic reasoning. Differences among neo-Russellians can be easily seen as differences in emphasis rather than substance, with Price and Woodward emphasizing the facts about us and our position in the world that make causal beliefs indispensable and give them their role in practical reasoning, and Albert and Loewer emphasizing physical asymmetries in the environment that provide the external infrastructure for causal reasoning. There is a lively and ongoing debate between Frisch and the proponents of these programs over details, but I had trouble isolating a substantive disagreement. Frisch says that the neo-Russellian view is reductive and his own is not, but there is nothing in his arguments that challenges the neo-Russellian conviction that physics has eliminated causal structure from its account of the fundamental fabric of the physical world. Frisch's defense of the indispensability of causal reasoning appeals to epistemic limitations on the part of agents and is entirely compatible with the neo-Russellian metaphysics. He does say in chapter 9, where he sums up his conclusions, that he sees no reason to think that causal structure is any less fundamental that the causal Markov condition or the initial randomness condition, but this is where his pragmatism about scientific representation becomes important. The metaphysics is explicitly deflationary, and he does not seem to be defending causal fundamentalism in the sense that the early Russell and the neo-Russellians are concerned to deny—namely, the view that the fundamental fabric of the world incorporates asymmetric relations of causal production or determination.If there is a divide between Frisch and the anticausalists, I suspect that it has to do more with the sorts of projects in which they are engaged. Frisch is doing philosophy of science. He is interested in how the world is represented in our physical theories. The book is an extended argument for the conclusion that the only way in which physics has moved beyond causal representations is in principle, and even then, only in the totalistic models of a closed universe. It remains as true as it ever was that, in practice, the aim of a large part of science, including physics, is the investigation of the causal skeleton of the world. Anticausalists like Loewer, Albert, Russell, and Price, by contrast, are engaged in naturalistic metaphysics. They are interested in the metaphysical lessons that physics holds for us about the fundamental structure of the universe. When they claim that causes have been eliminated from physics, they mean that it is no longer recognized as part of the fundamental fabric of nature.As physics becomes increasingly abstract, concepts that are basic and indispensable in the worldview of a situated agent disappear from the fundamental level of physical description and become targets of analysis. They get recovered in a project that places agents in the foreground and looks at how human representations are structured to help them cope with a complex and changing world. Questions about whether causal relations fall on the side of the world or on the side of the agent no longer have a simple answer. We have emergent structures in the agent's environment on the side of the world, and a functional role on the side of the agent. And we would do better to ask: what is it about the way the agent couples to the world that allows those structures to play that role for agents like us?In a recent essay for Aeon magazine, Frisch (n.d.) wrote an essay about causation whose subtitle is “Either cause and effect are the very glue of the cosmos, or they are a naive illusion due to insufficient math. But which?” If we've learned anything from the discussion of causal relations since Russell's essay, it should be that physics has moved beyond this simple dichotomy, revealing a more complex story about the physical infrastructure that supports the human ability to identify and use structures in the physical environment as strategic pathways to bringing about ends.Overall, this is a terrific book. It represents the best kind of engagement between philosophy of science and metaphysics, going beyond the simplified schemas that often appear in metaphysicians’ representations of physics to real instances of physical reasoning. It also reflects a very welcome move away from the global models of a closed universe that tend to attract philosophers’ attention to the models of open subsystems that capture the bulk of the attention of the practicing physicist. The standard of discussion is very high, and debate about the status of causal structure has been considerably sharpened and significantly advanced by it. The take-home lesson of Frisch's book—that causal reasoning is just as indispensable in physics as it is in other sciences—is an important one that needs to be made.