Barbara Ehrenreich has a new piece on agency in science that makes some serious mistakes and deserves careful treatment. I wanted to like the piece because she’s coming from a perspective I find attractive, but a correction of her mistakes ultimately undermines her view. It’s important for those of us interested in issues of organization and complexity to be clear about why her position is untenable.
She opens her critique of “rationalist science” with a discussion of play that I’m quite sympathetic to, echoing some of David Graeber’s commentary in a companion article:
So maybe carnival and ecstatic rituals serve no rational purpose and have no single sociological “function.” They are just something that people do, and, judging from Neolithic rock art depicting circle and line dances, they are something that people have done for thousands of years. The best category for such undertakings may be play, or exertion for the sheer pleasure of it. If that’s the case, then we have to ask why it has been so difficult for observers, especially perhaps white bourgeois Europeans, to recognize play as a time-honored category of experience.
Ehrenreich isn’t talking about “play” in the sense of unstructured idle activity. She’s talking instead about celebration as a ritualized social event: “doing something together, something that was fun and sometimes ecstatic to the point of trance”. I think it’s important to distinguish these structured social rituals from “play”, since it’s often the case that the idle unstructured sort of play isn’t tolerated at social functions where ritualized repetition is central to the activity– try being a kid at a wedding or graduation and see how much fun you have. But that’s a minor complaint. I find the rituals of human celebration interesting too. I appreciate the article’s recognition that these behaviors don’t serve the simple function of “social cohesion” in a straightforwardly instrumental way, but work at a much deeper and more elusive level. Ehrenreich says, “I am drawn to this idea [of play as a ‘basis of reality’] as a metaphysical speculation, so long as we remember that play has no moral valence. It can be elegant, it can be rough, it can be deadly—or all those things at once.”
Like Ehrenreich, I’ve also been thinking about celebration in terms of deliberately induced social trauma of a far more tumultuous sort than the term “social cohesion” can handle. Celebrations are extreme events: excessive consumption, overstimulation, insobriety, and sheer physical damage are as much a part of a 5 year old’s pinata birthday party as they are a part of Mardi Gras. Part of the function served by these events is simply a recognition that we survived this trauma together. It provides the opportunity to experience each other in the midst of trauma, and it helps to prepare us for potentially traumatic experiences we might share in the future. Celebration doesn’t create social cohesion ex nihilio but reinforces and adjusts existing social ties by subjecting it to these extreme events, something like the way you temper metals by subjecting them to intense heat. This connects to my essays on extremism where I describe how the rhetorical dynamics at the extremes reinforce the conceptual organizational of the whole system. In both cases the system finds points of stability through engaging with the edges. I suspect there are general lessons to draw from these examples.
But this is where my sympathies with Ehrenreich’s view ends. Instead of taking seriously the lessons of organizational nonlinearity and exploring their implications, she uses the example to mount an old and tired critique of modern science and mechanical explanations. She argues that science’s difficulties in conceptualizing play are symptomatic of deeper biases within science:
I would say that the roots of our short-sightedness about play range far beyond economics, that they extend into all of Western science, and that what is at stake here is ultimately even deeper than play. For the last few hundred years, Western science has been on a mission to crush all forms of agency, which I mean in the philosophical sense as the capacity for action… The goal of science was always to replace agency—or whim or desire—with deterministic mechanisms…
She argues that rationalist science back to Descartes assumed– falsely– that all things can be explained: “Everything happens “for a reason” as part of some vast, insensate, cosmic mechanism.” The implication is that play and celebration resist any mechanical explanation and demonstrate a failure of rationalist science: we don’t do these things for any reasons whatsoever; these are just things we simply do. Consequently, play and other apparently inexplicable human behaviors reveal a failure of rational science’s attempt to give a mechanistic explanation of the world.
I agreed earlier that celebration was something that humans do, and have done for a very long time, and moreover that this activity doesn’t serve any simple social function like “social cohesion”. But Ehrenreich concludes from this that celebration serves no function and therefore cannot be explained, and this conclusion simply does not follow. Celebration serves many complex functions for many different aspects of our organizational structure. For instance, our celebrations now also serve a variety of complex economic functions, and has a serious impact on the pace and schedule of technological innovation, industrial production schedules, and so on. These economic functions are quite distinct from the psychosocial functions that play served in our evolutionary history, but they are undoubtedly part of the function of human celebration today.
The functional roles of components change over time as a system adjusts to the constraints of its environment, a fact that shouldn’t be surprising from an evolutionary perspective. In biology, the change of functional role over time is called exaptation. The first example of exaptation in Wikipedia is of bird feathers, which originally evolved for purposes of temperature regulation but ultimately served purposes of flight, swimming, mating, and all the other ways to which birds put their feathers to use. The indefinite plurality of ends to which birds might put their feathers is not evidence that bird feathers evolved without reason, or that their appearance and nature cannot be explained, and certainly isn’t grounds for doubting mechanistic explanations in the sciences more generally. Rather, it demonstrates just how complex of an environment these feathers are adapting to, and the range of constraints they are subjected to over the course of their evolutionary development. These constraints are mind-bogglingly complex and not entirely understood. But there’s little doubt among scientists that a mechanistic explanation of a feather’s features can be given in terms of evolutionary development within specific environmental constraints. Some of those features might be random or without a direct functional explanation– the so-called “spandrels” of evolution– but that’s different from saying that some features can’t be explained.
But Ehrenreich pushes all in, claiming instead that the empirical evidence is pushing science to abandon the project of mechanical explanation entirely. I quote the next passage at length because it is the source and evidence of her most serious errors:
But science itself has been changing, and not because of any philosophical unease about the paradox of “man’s” existence in an otherwise dead world. It was simply overwhelmed by new empirical evidence, starting with studies of the photoelectric effect that led, in the early twentieth century, to quantum mechanics and the shocking realization that electrons move as if by chance or, to risk the dread charge of anthropomorphism, by choice. Then came, late in the twentieth century, the equally paradigm-challenging formulations of nonlinear dynamics—or, as it is more sensationally termed, chaos theory. Put simply, very small differences in starting conditions can lead to huge differences in effects, which is why, among other things, we cannot hope to predict the weather with total confidence. Biological systems also turn out to be mathematically nonlinear, often better described by algorithms than by static (and effectively deterministic) equations. It turns out that patterns can arise out of muddle; microscopic events can synchronize to produce macroscopic effects.
Her discussion here is riddled with fundamental conceptual mistakes, of the sort that even Wikipedia is sufficient to correct. This isn’t a difference of opinion or a matter of intellectual disagreement, this is simply a case of her getting the fundamental concepts wrong. So let’s start with at the end and move in reverse order:
Error #1: Contrasting “algorithms” with “static equations”.
This distinction does not exist. In fact, you can describe all the “static equations” of mathematics in terms of algorithmic procedures and vice versa. The idea that science has somehow recently turned from the latter to the former is misleading at best.
Error #2: Contrasting “algorithms” with “effectively deterministic”
An algorithm is defined as a procedure that guarantees a definite result. Computer scientists use the term “effective method”, and it is effective in precisely the sense that the results are deterministic. Probabilistic algorithms might generate some range of results, but if it is an algorithm it is following some definite (and explainable) stepwise procedure. The results of an algorithm can be explained mechanically by appeal to that procedure. An algorithm is not an alternative to a mechanical explanation
Error #3: Contrasting “chaotic” or “nonlinear systems” with “deterministic systems”
This is the most serious error of this paragraph, and demonstrates how grievous the conceptual misunderstanding runs. It is very common to confuse “chaotic” systems with systems that are random or inexplicable, but this is a mistake. Chaotic systems are systems that develop in ways that are extremely sensitive to initial conditions. That means two systems that start off very similar might be completely different later on, so it’s really difficult to predict the developmental trajectory of any given system. But chaotic systems still develop in deterministic ways. That’s exactly why the difficulties of prediction are surprising! The opening paragraph of the wiki article on chaos theory makes this determinism clear:
Chaos theory studies the behavior of dynamical systems that are highly sensitive to initial conditions—an effect which is popularly referred to as the butterfly effect. Small differences in initial conditions (such as those due to rounding errors in numerical computation) yield widely diverging outcomes for such dynamical systems, rendering long-term prediction impossible in general.This happens even though these systems are deterministic, meaning that their future behavior is fully determined by their initial conditions, with no random elements involved.In other words, the deterministic nature of these systems does not make them predictable. This behavior is known as deterministic chaos, or simply chaos.
Complex natural systems are chaotic and nonlinear, and these nonlinearities often behave in surprising and unexpected ways, especially when we’re expecting linearity. But Ehrenreich is treating nonlinearity as a failure of mechanistic explanation itself. In fact, she’s going a step further and embracing animism as the alternative to mechanistic explanation, which brings us to the final error:
Error #4: Claiming electrons move by “chance” or by “choice”
This isn’t just said in passing in the above quoted paragraph; she reiterates this animist conclusion later in the essay, and it’s clearly a central part of her thesis and a primary motivation for her article:
Agency, in some form, is everywhere, from inchworms to electrons, and it is time for science to recognize that everything is, if only in some metaphorical sense, alive.
She makes clear that modern science has “no gods and spirits, no vital forces”, but nevertheless she is arguing that contemporary science supports animism over mechanistic explanation. But this comes at a cost; throughout the essay Ehrenreich casually equates whim and choice and agency with chance and indeterminacy.
So let’s be clear about two things. First, electrons don’t “move by chance”. They’re probabalistic, but they don’t “move by chance” in the sense that we don’t know what it will do next. In fact, because they are probabilistic (and not deterministic), their behavior isn’t even chaotic; it is highly regular and predictable, and can be harnessed to (for instance) reliably power all the electronic equipment that forms part of the fabric of our digital lives. The fact that these machines work isn’t a matter of chance, but a matter of the very predictable and regular behavior of the electron. That behavior, for what it’s worth, is explained by a theory called “quantum mechanics”, which is not “mechanical” in the sense imagined by Descartes, but is a mechanistic theory nonetheless. With quantum mechanics, the physics of virtually all the phenomena we experience everyday can be explained and understood mechanically. Which is to say that from a scientific perspective, we’ve never been more confident that there are mechanical explanations of every phenomena we experience. The idea that nonlinearities have somehow rendered the mechanistic project obsolete is a complete nonsequitur.
But second, and more importantly, chance is not synonymous with choice! Ehrenreich is looking for agency “in the gaps” where our traditional explanations fail. She seems to think that if it’s not a mechanism that explains the behavior of an electron (which it is), then it must be the whim and fancy of the electron itself, as a distinct agency with desires and drives of its own. What a poor defense of agency that treats it as nothing more than the chance fall of dice, no different in the electron than in the person! Ehrenreich’s animist argument ironically depends on a corrupted form of the deterministic explanations she’s criticizing: if the behavior of the electron is not determined mechanically (which, again, it is) then it must be determined by the choice of a free agent. I see absolutely no support for this argument or its conclusion, and it is misleading to describe this as a growing consensus among scientists.
Ehrenreich presents recent science as anti-mechanistic and amenable to her animism, but she never bothers to explain why science has mechanistic biases in the first place. The mechanicist turn in early modern science wasn’t just some arbitrary bias or nefarious proto-capitalist/individualist conspiracy. Instead, it was a very deliberate reaction to the obvious and critical failures of animism in the natural philosophies inherited from Aristotle, which had been taught uncritically for centuries by the time Descartes and his contemporaries arrive. Aristotle explained the physics of the natural world in terms of the teleological drives inherent in all things, where even the inanimate rocks had the power necessary to achieve geometrically ideal ends. Much later, Thomas Aquinas interprets Aristotle’s animism to be continuous with scripture, effectively grounding all inquiry into the natural world in religious dogma. In these conditions, Galileo is subjected to house arrest and countless others with less notoriety were persecuted and tortured for drawing conclusions from their observations of the natural that contradicted scripture.
In this hostile environment, Descartes and other early scientists turned to mechanical explanations and skeptical methodologies precisely for their demonstrable reliability and undeniable explanatory success compared to their dogmatic alternatives. Enrenreich glides entirely over this troubled history. Machines served as a paradigm of explanation in the sciences precisely because each step in a mechanical process was accounted for and the entire sequence determined the expected result. Our ability to imagine a process in mechanical terms is a large part of what makes us confident that we understand it, and not merely perpetuating convenient or comfortable dogmas. This “bias” is at the heart of scientific methodology, and perhaps is best captured by Feynman’s quote ‘What I cannot create, I do not understand‘. And this ethos at the heart of science has in no way been called into question by recent developments at the edges of our understanding. If anything, the progress we’ve made at the edges has given us more reason for confidence in our method.
Appealing to “choice” or “chance” or “agency” or some other unseen force to explain what cannot otherwise be explained is no answer at all. I’ve argued that lazy bias towards animism was precisely the sort of dogma early scientists fought so hard against, and we do a disservice to that tradition to reject its lessons so quickly. Ehrenreich is eager to adopt an animist metaphysics, if only as a metaphor, and if it pleases her then I have no room to object. But we should be clear as crystal that these conclusions are absolutely not warranted by our best science. And if Ehrenreich’s essay is not to be taken as outright hostile to science, then we should at least recognize that it isn’t doing science any favors by fumbling with our basic concepts and drawing unsupported-but-metaphorically-comforting conclusions from these misunderstandings. Lazy animism isn’t an alternative to mechanistic science, it is a retreat from it.
For what it’s worth, contemporary metaphysics has recently become enamoured again with Thomist-style “powers” ontologies, and a number of professional philosophers have come out defending something like the animism Ehrenreich is defending here. It’s precisely because these views have become trendy that I’m giving them a careful treatment here. But instead of merely being critical, let me also try to be constructive and explain what agency looks like at the edges of modern science. I agree with Ehrenreich (and with Haraway) about the importance of nonhuman agency. And Ehrenreich is correct that agency emerges from the often chaotic dynamics of complex systems. But discussing the emergent properties of nonlinear systems has an air of inexplicability and mystery that can easily be mistaken for supporting animist conclusions, so we should be clear that there’s no room for choice in this process.
An agent is a system that is disposed to behave in certain ways within a given environment. These behaviors (function, powers, properties) are defined in terms of the interactions they have with that environment. The behavior of these systems aren’t the result of internal, unseen drives, but from the many ways in which these systems interact with each other. For instance, electrons have the property of “charge”, which is just the disposition to behave in certain ways around electrons of opposite charge, and in other ways around electrons with identical charge. Electrons don’t have a choice in this matter, and their behavior isn’t random or unpredictable; a negatively charged electron will always be attracted to positive charge, in ways that are well described by the equations of electrodynamics.
Charge also appears to be a fundamental property of matter, in the sense that there don’t appear to be the result of anything more fundamental. There’s a small set of these fundamental properties (mass and spin are among them), and I find it curious that contemporary metaphysicians don’t seem particularly interested in investigating the details of these properties along side our best physicists. In any case, most of the systems we interact with aren’t fundamental in this sense, but are the result of enormously complex interactions among trillions of these fundamental particles, which compose densely layered networks of agents at many scales of analysis. Each of these scales present new opportunities to talk about coherent systems embedded in environments with distinct functional powers. And interestingly, some of these larger agents (themselves composed of agents, composed of agents, etc) have properties that can’t straightforwardly be reduced to the powers of their components. These are the so-called “emergent phenomena” of nonlinear systems that motivates Rhrenreich’s critique. All living organisms are systems of this type, but so are other complex systems, like cities and climate cycles and countless other natural systems.
To see there’s no unexplained mystery (or whim) in this process, let’s take an example from Stump’s discussion of water:
The two covalent bonds in a water molecule are polar and give the molecule its geometry. That is, the non-bonding electrons of oxygen remain closer to the oxygen atom than the shared electrons do, and they exert a stronger repulsive force against the shared electrons. As a result, the two hydrogen atoms are pushed closer together. Consequently, the molecule has a peculiar characteristic: the charge of the electrically neutral water molecule is not distributed uniformly though the molecule. Unlike either a hydrogen or an oxygen atom taken in isolation, the water molecule, which is composed of hydrogen and oxygen, is unevenly charged.
The geometry of the water molecule emerges from the interaction of its components, which results in properties not shared by those components. This gives water its lopsided geometry and other unique properties that make it a suitable environment for the development of life. So contemporary science can agree with Aristotle, at least in this case, that the whole is more than the sum of its parts. But that’s quite different from saying that the geometry of water is unexplained. In fact, we have a number of mechanical models that explain various aspects of the geometry of water, including quantum models. Although these models can be described in terms of “universal agency”, none of these models leave any room for choice or whim. Instead, like good mechanicist scientists, each of these models makes very precise and testable claims about how water will behave in various conditions and environmental constraints.
Stump takes the emergent properties of water to vindicate at least some aspects of Thomist and Aristotlean metaphysics. I’m comfortable with recognizing that modern science may have reacted too strongly to Aristotle, and that there may be aspects of these classical metaphysics that are relevant to contemporary discussions. But we shouldn’t mistake the project of updating Aristotle with a vindication of his animism or an admonishment of contemporary mechanistic science. If anything, the project of making Aristotle relevant to contemporary science requires being even more critical and careful of his views, and clearly separating the wheat from the chaff, precisely so we don’t repeat the mistakes of our predecessors.
And that means not hiding behind comfortable metaphors or compelling narrative fictions. It means being serious about science and the views it licences us to hold. And most importantly, it means taking the time to learn about what actually going on at the cutting edges of scientific inquiry. Because if there are any metaphysical work to be done, that’s where it will happen.
As W. V. O. Quine says, “Philosophy of science is philosophy enough.”