There are a lot of ways to define science. The broadest might characterize it as a systematic process for uncovering facts or explanations about the way the world works. From there, individual scientists sometimes differ over the exact features that distinguish science from other enterprises, but they all tend to accept the basic proposition that it is an empirical enterprise. The degree of agreement between theory and observation is what ultimately decides whether a scientific idea offers a good or bad explanation of natural phenomena.

This is why science is often depicted as a naturalistic enterprise. Which is true, but there are different strains of naturalism. It’s worth taking a moment to distinguish them. First, there is metaphysical or ontological naturalism. This is the view that the universe is entirely of matter or other measurable stuff, governed exclusively by natural forces. This stands in contrast to methodological naturalism. Advocates of methodological naturalism grant that the universe may be filled with or influenced by supernatural or immaterial forces, but stipulate that those are irrelevant to science.

Both ideas have their weaknesses. In The Big Picture, Sean Carroll introduces the concept of poetic naturalism as a way of getting around them. Poetic naturalism (PE) grants breathing room for concepts that don’t necessarily relate to the steely, unforgiving rudiments of physical reality. It is traditional naturalism’s less conservative, more ecumenical progeny.  PE grants room for higher order concepts like consciousness and protons in a world populated by more fundamental stuff. It even allows room for the “supernatural”, so long as it produces some measurable effect and offers some explanatory merit.

There are certainly some physicists who are considerably less generous when it comes to the reality of emergent phenomena like biological evolution and human consciousness. Though Carroll adopts a more tolerant pose, it’s by no means revolutionary. Old school naturalism was never married to the idea that the only thing science can meaningfully address are esoteric subatomatic particles like quarks and gluons. It recognizes that everything observable is made from those things, but doesn’t automatically suggest that the only substantive way to talk about the workings of reality is in terms of fundamental physics. That is a perspective that seems exclusive to provincially minded physicists.

In this, Carroll is starting from a strange place. He is introducing a new concept in order to account for how the way most skeptically minded critical thinkers – including most scientists – already think of the world. His reasons for doing so are clear. Viewed from the realm of ordinary experience – or even, for that matter, sciences like psychology or biology – the picture of the universe that comes from studying fundamental physics is extraordinarily weird. Basic concepts like time and causality begin to look less and less essential to the way things work. The physical world of ordinary experience is mostly empty space permeated by fields, sprinkled with tiny particles that simultaneously occupy every possible state.

That’s all very bizarre. On the scales most people are used to, the ordered flow of time and the causal connections between events are conceptually indispensable. But in the world of quantum physics, it seems like they are superfluous. Our best theories seem to work perfectly well without them. In this sense, they only emerge as an artifact of our particular frame of reference: relatively large, slow moving creatures inhabiting a certain spot in the universe.

According to Carroll, anything beyond the world of the subatomic – the infinitesimal, fuzzy world of point particles and force fields – is an emergent feature that is somehow less real than the elementary stuff of which it is made. Nowhere does Carroll say this outright. Instead, it is implicit as the motivating core of poetic naturalism. It is a philosophy of science that Carroll has invented as a way to avoid saying flowers and cells and eyeballs are less real than neutrinos and electromagnetism.

The basic argument at the heart of his notion of poetic naturalism is that the truth or veracity of a scientific idea is inextricably linked to its usefulness. It is a reworking of the old instrumentalist doctrine that it doesn’t matter so much whether or not a theory is true in any axiomatic or Platonic sense. That is, science needn’t hang its hat on whether or not it is about things that actually exist “out there” in the universe. The important thing is that it reliably yields accurate predictions.

Carroll’s innovation is to essentially turn instrumentalism on its head. Back in the pioneering days of quantum physics, in the first decades of the 20^th century, scientists (and philosophers) struggled to reconcile the probabilistic theories they were uncovering with the apparently deterministic world in which they lived. Unmistakably, it is puzzling. The quantum world is one of imperfect knowledge, where your ability to know one feature of a particle in great detail (say its velocity) actually impinges on the precision with which you can measure its position. Systems are described according to wave-functions, where their state is understood as an evolving probability distribution. A particle has good odds of being here and having these properties, poorer odds of being there and having those properties, and so forth. Prior to observation, they occupy a “superposition” of all possible states. This is the world of Schrodinger’s infamously dead and alive cat. It’s almost unsettlingly counter-intuitive. It also happens to be true, as illustrated by the famous double-slit experiment.

The deep peculiarity of the quantum world led to a purely utilitarian interpretation of the relationship between the emerging physics and experimental results. Though the relevant mathematics describe the behavior of subatomic particles in terms of collapsing wave functions, physicists like Neils Bohr adopted the position that one could remain agnostic about the concrete physical nature of both the particles and their wave-like nature. All that matters is that describing them in that way yields predictions that are upheld by experiment. This isn’t a view that treats wave-functions and point particles as convenient fictions. Instead, it simply says the precise physical nature of the objects of study doesn’t matter. The important thing is that the physics yields robust predictions.

This kind of agnosticism doesn’t appeal to Carroll. In his view, the suite of subatomic particles and fields described by modern quantum mechanics are the real deal. Quarks and gluons and neutrinos and photons are the raw building blocks of reality. Everything else is emergent. Carroll’s poetic naturalism is an inverted instrumentalism. Reality is built of interacting particles and fields. Higher order theories, like Darwinian evolution and plate tectonics, are just useful approximations. But since PE grants that a given theory’s truth hinges on its usefulness within a given domain of application, the fact that higher level theories aren’t directly tethered to direct reference to quantum phenomena isn’t a problem. Because Darwinism and plate tectonics work on the relevant scales, they can be thought of as just as true as quantum field theory within the appropriate domains. According to Carroll’s PE, the elegant mathematics of fundamental physics is a more fine grained description of fundamental workings of reality.

To a degree, this particular flavor of naturalism is a sympathetic view. But there is also a point at which PE’s reliance on an instrumentalist account of science begins to rob it of some its resilience. Good scientific ideas are indeed those we find most useful. Doubtlessly, this is a critical part of how science works. Yet it is also far from the whole story.

For one thing, PE remains rather sparse on the issue of what does and does not count as useful. The most obvious and objective way to evaluate a theory’s usefulness is to test how accurately the predictions it generates match observation. Carroll never spells out his conception of theoretical utility. Throughout most of The Big Picture, it seems like Carroll’s version of scientific usefulness boils down to the correspondence between prediction and observation. That’s a pragmatic – if somewhat tepid – view. However, there are points where it is clear Carroll is invoking a picture of utility that might lead us into unnecessarily murky waters.

If all Carroll means by utility is a capacity to generate reliable predictions, it would be hard to quibble. It is conceivable that we might arrive at algorithms that allow us to make good predictions about the behavior of systems by pure chance or brute trial-and-error, all the while remaining entirely agnostic about the underlying processes. In some sense, this is always the case. For most people, the most familiar scientific theories are those that deal with large objects moving at relatively low speeds over relatively small distances. It makes sense to talk in terms of solid fingers impacting a solid keyboard. At that scale, talking instead of tiny particles and fields and empty space is unwieldy. It doesn’t buy us much in the way of understanding. There isn’t a clear connection between what is going on subatomic scales and the kinds of explanations that work in the realm of evolving biological systems or dynamic geological processes. This is the case for a large swath of science, from neurology to planetary astronomy. Scientists build robust, powerful explanations for how the world works and remain agnostic about the precise mechanisms that link the behavior of the big with the behavior of the very small.

Inasmuch as utility is a measure of veracity, it is also true that higher order theories are true in a more concrete ontological sense. Genes and minerals are not just reified constructs, invented for the purpose of making predictions about how organisms change over time or how different temperatures and pressures yield different kinds of crystalline structures. They aren’t posited with a wink and nod to a deeper understanding that what is really going on is explained in terms of particle physics. They are things that actually exist, out there in the world. Evolutionary biology and geology are good theories because they make reliable predictions and consistently avoid falsification. And they make those predictions because they present accurate models of how the world actually works. This is a considerable step beyond instrumentalism. There is no obvious contradiction between thinking of cell as a fundamental component of a biological system and thinking of a cell as something that emerges from the interactions of subatomic particles. Subatomic particles are real. And so are all the higher order structures they comprise, from protons and planets to tadpoles and trees.

As a criticism of poetic naturalism, this might look a bit like trivial complaint rattling around in a big bag of pedantry. Carroll isn’t going to dispute that there really are such things as nucleotide sequences, neurons, and metamorphic rocks. PE begins to break down when it strays from the cold, decisive reaches of traditional science. In science, utility has long been understood as an important – if partial – measure of truth. But there comes a point in The Big Picture where Carroll makes moves to substantially broaden the definition of utility. It ceases to be a circumscribed instrument for talking about how much fidelity there is between theory and observation and becomes dangerously tied up in subjective preferences.

Consider a definition of instrumentalism where usefulness is defined much more widely than might be captured by agreement between prediction and observation. An idea’s utility can be more broadly construed as a measure of how well it works to achieve or justify an end. For instance, some wealthy billionaires have the aim of maintaining their wealth and accruing more. In this, they may find libertarian economic systems enormously useful. Does that make the underlying principles true? Under a broad enough definition of utility, it obviously does. The fact that extreme libertarian philosophies don’t offer good solutions to problems of third-party enforcement or public goods dilemmas is hardly a problem. Likewise, the fact that they make thoroughly erroneous assumptions about the nature of economic systems and human behavior is irrelevant.

As a groundwork for any kind of epistemology (that is, a theory of knowledge and how to go about gaining it) this seems garishly ridiculous. Carroll, I suspect, would instantly object. Yet he puts precisely this kind of lily-livered instrumentalism to use in his defense of compatibilist free will. For the unfamiliar, compatibilism is the stance that conscious agents can exercise a narrow range of agency within otherwise entirely deterministic systems. It accepts that we are built of organs that are built of cells that are built of molecules that are built of atoms that are built of subatomic particles and fields. Likewise, it accepts that our minds emerge from the meat in our skulls, itself built of physical ingredients all the way down to the subatomic realm. But is posits that consciousness, as an emergent product of otherwise physical, deterministic systems, somehow exerts some amount of sovereignty over the natural world. To put it a little more simply, in the world of compatibilism, particles bump into particles, building more and more intricate and sophisticated structures, until they pass a threshold of complexity beyond which they produce a system sufficiently complicated that it can escape that causal chain.

There are plenty of people who find the idea of strict determinism unpalatable, for a lot of different reasons. Agency is a fundamental component of humanity’s self-conception. It is tied up in religious notions of sin and damnation and salvation. Similarly, it is exerts considerable social force when it comes to legal notions of punishment and justice. It is intrinsic to the embarrassment we feel after making a mistake and the triumph we feel after accomplishing a goal. Whatever science might say, it certainly feels like we make choices. Free will is one of the founding precepts of commonplace ideas about what it means to be human.

The problem is, our best scientific understanding of the natural world leaves less and less room for it. The more we understand about how humans work as biological systems, the less space there is for the notion that we are boundlessly willful agents. Consequently, the idea of so-called libertarian free will – that human preferences and decisions are an entirely unconstrained, top-down affair – has basically been consigned to the philosophical dustbin. Very few people who take the results of science seriously believe that humans can do whatever they want, and those who do go through some brutally torturous intellectual gymnastics to get there. Science teaches us about the world as it actually is, not as we want it to be. In the domain of consciousness and identity and free will, it teaches us this: We are our brains and, to lesser but still meaningful extent, our bodies and those systems are governed by the same rules as everything else in the universe.

Now, for the sake of intellectual honesty, it is worth pointing out that science has not ruled out the possibility of free will. It has merely (if one can describe such a monumental reordering of the human worldview so flippantly) shrunk the domain in which free will can operate. A few centuries ago, it was a free range affair, able to roam wherever philosophers and theologians cared to take it. Now, it lives in an increasingly cramped paddock. The scope for specifying what free will is and what it is capable of is constantly shrinking. One can posit free will, but the move itself is extraneous and costly. The only work it might do is explain why we perceive ourselves as volitional agents. That could be satisfying on an existential level, but it’s a little strange, considering the bulk of science consists of rigorous attempts to prevent our perceptions about what the world is like from fooling us about how the world actually works. Simultaneously, arguing for free will introduces the burden of explaining why the human brain is exempt from the rules that govern all other matter. As the neurobiologist Robert Sapolsky put it, free will has become a kind of psychological god-of-the-gaps argument. It doesn’t carry any explanatory weight, but interested parties can still find room to invoke it if they wish.

In Carroll’s case, the argument for free will rests on the grounds that the concept remains useful, particularly when it comes to issues of responsibility and punishment (or reward). He doesn’t seem to care whether or not that description is compatible with more fundamental descriptions of reality, be they quantum mechanical, molecular, neurobiological, or evolutionary. Only that it maintains a domain of utility.

Conceivably, one could define wider domains in which less circumspect claims are thought to be useful. Many religious sects find the idea of libertarian free will a useful component of cosmologies of eternal reward and suffering. That reasoning is, of course, painfully circular. Humans are imbued with free will as means to earn reward or punishment. And in the absence of free will, the concepts of eternal punishment and reward become ethically unjustifiable nightmares – the workings of a capricious hand in a cruel universe.

Carroll is an atheist, so these kinds of cosmic, supernatural reward schemes don’t appeal to him. But in his defense of free will, he makes just this kind of argument. Our criminal justice system is built on penalties and punishments. If we don’t have free will, the argument goes, that system doesn’t make much sense. Well, yes and no. The idea that our actions are determined outside the realm of conscious, willful influence is perfectly compatible with the idea that behavior is sensitive to environmental inputs. In this framework, punishment can serve two ends: discouraging repeat behavior and keeping dangerous people away from the rest of us. A system sensitive to external influence can still be entirely deterministic.

However, the idea that punishment is an end unto itself or that we should endorse the idea of free will as a means of justifying the existing system of criminal justice doesn’t hold a lot of water. Modern criminal justice systems – particularly in the United States – use the concept of punishment to perverse and unjustifiable ends. We shouldn’t use free will as a prop to stabilize that old and barbaric edifice. Rather, we should use our growing understanding that behavior isn’t subject to very much volitional control as an impetus for reform. In Carroll’s view, we ought to hold onto a constrained version of free will as a means of justifying the status quo. The more enlightened (and scientifically consistent) perspective is that we ought to use our understanding of what actually shapes human behavior to build a more humane and effective criminal justice system.

Curiously, Carroll even goes so far as to hitch the idea of free will to the uncertainty of future actions. Under most conditions, we can’t predict human behavior with very much precision. Therefore, he argues, it must be subject to some sort of top-down control. Under that line of thinking, he even goes so far as to say we will have progressively less free will as our understanding of human neurobiology grows and our capacity to forecast human behavior improves. This is a step beyond instrumentalism. It straddles an uncomfortable boundary with the kinds of idealist fantasies he had earlier rejected in dealing with “quantum consciousness” and the hazy spiritualist belief that conscious precedes existence. Here, he is not just saying that we should hold onto to the idea of free will because it serves beneficent societal purposes. He is actually saying that the existential state of free will is caught up in how well we understand the human brain.

Reality is what it is, regardless of whether or not humans understand it. And that is the fatal flaw of poetic naturalism. By binding his epistemology to an excessively permissive breed of instrumentalism, Carroll is suggesting that there are cases in which truth is anchored to human reasoning. That misses a more nuanced point. Our ability to uncover truth is inextricably tied to the power of human reasoning. What is and is not true is not. Either free will exists or it doesn’t. How useful we find the concept is irrelevant.

Carroll’s use of PE to defend a constrained version of free will is damning, primarily because it is easy to dismantle. But the flaw that cripples the poetic naturalist’s conception of free also cripples his conception of everything else. It doesn’t make sense to say that everything that can’t be reduced to particle physics is only real insofar as humans find it useful, because it binds all of reality to human understanding. The fact that our understanding of vision has yet to be reduced to the equations of quantum field theory doesn’t mean that eyeballs and retinas and optic nerves don’t really exist on a fundamental level. It only means that there are gaps in our understanding. Hopefully, they will one day be filled. But it is also entirely possible that they won’t. We may never be able to explain certain higher order phenomena in terms of fundamental physics. That doesn’t mean we need to invent a new version of naturalism to account for them or that we can use the gaps in our knowledge as an excuse to entertain any brand of wishful thinking we find convenient.