Here in our daily environment, the world of people and cars and houses, we have a complete inventory of the particles and forces and interactions that are strong enough to have any noticeable effect on anything. That’s a tremendous intellectual achievement, one of which the human race can be justifiably proud.
What this book offers: a full naturalistic worldview starter kit
The Big Picture covers quite a few topics: epistemology, quantum mechanics, the origin of life, the nature of time, consciousness, free will, morality, meaning, and more. Carroll ties science and philosophy together to present a cohesive naturalistic worldview and explore its implications for some of the deep questions that everyone grapples with. I strongly disagree with some of his major conclusions - especially about consciousness, free will, and morality - but I admire the project and learned a lot from it.
Covered in this review:
Carroll summarizes “naturalism” as follows:
1. There is only one world, the natural world.
2. The world evolves according to unbroken patterns, the laws of nature.
3. The only reliable way of learning about the world is by observing it.
He calls his way of looking at reality “poetic naturalism”, which builds on that with the following:
1. There are many ways of talking about the world.
2. All good ways of talking must be consistent with one another and with the world.
3. Our purposes in the moment determine the best way of talking.
The point is to explain how one can believe that ultimately the world only consists of fundamental physical particles/fields, and yet also believe that things belonging to higher-level categories are real and should not be regarded as illusions. For instance, it makes sense to talk about tables and chairs, even if they are just collections of atoms. Trying to talk only of atoms would be unwieldy, and the ideas of “tables” and “chairs” have explanatory power. Similarly, we need the vocabularies of chemistry and biology because trying to use just the vocabulary of physics to predict the behavior of the world would be entirely impractical, even though in principle possible.
That may not be particularly revelatory, but Carroll does a good job of exploring how different vocabularies can relate to each other and the pitfalls of mixing them:
The most seductive mistake we can be drawn into when dealing with multiple stories of reality is to mix up vocabularies appropriate to different ways of talking. Someone might say, “You can’t truly want anything, you’re just a collection of atoms, and atoms don’t have wants.” It’s true that atoms don’t have wants; the idea of a “want” is not part of our best theory of atoms. There would be nothing wrong with saying “None of these atoms making up you want anything.”
But it doesn’t follow that you can’t have wants. “You” are not part of our best theory of atoms either; you are an emergent phenomenon, meaning that you are an element in a higher-level ontology that describes the world at a macroscopic level. At the level of description where it is appropriate to talk about “you,” it’s also perfectly appropriate to talk about wants and feelings and desires. Those are all real phenomena in our best understanding of human beings. You can think of yourself as an individual human being, or you can think of yourself as a collection of atoms. Just not both at the same time, at least when it comes to asking how one kind of thing interacts with another one.
He approaches the nature of consciousness from within that framework, arguing that “[c]onsciousness isn’t an illusion, but it doesn’t point to any departure from the laws of physics as we currently understand them.” (p. 351)
Regarding Frank Jackson’s well-known (within philosophy) thought experiment about “Mary the Color Scientist,” Carroll concludes that in distinguishing between learning about the color red by studying versus learning about the color red by actually seeing red, “we have merely introduced an artificial distinction between two kinds of collections of synaptic connections” (p. 353). I more or less agree; this thought experiment is not really effective in highlighting the “hard problem” of consciousness.
The book goes on to address the thought experiment of “p-zombies”: is it possible in principle to have an exact physical replica of a person, that behaves identically to them, but that is not conscious? If like me you think consciousness involves some mental properties that are not reducible to physical properties (at least, physical properties as ordinarily conceived), you may be inclined to say “yes”, but that presents a dilemma. Either those mental properties have some effect on the physical world, or they don’t. Either option has problems.
If the mental properties are merely “along for the ride” (p. 356) - an epiphenomenalist view - why believe in them at all? I’m not sure the discussion in this book would have been enough to lead me to this conclusion, but I’m pretty convinced currently that epiphenomenalism makes no sense. The whole motivation for believing in mental properties is that we seem to have a direct awareness of them. But if the mental properties aren’t playing a causal role in the formation of your belief in them, why trust that belief? (Carroll’s podcast has an episode interviewing philosopher of consciousness David Chalmers, where Chalmers mentions this issue of where our belief that there is a hard problem of consciousness comes from, calling it the “meta-hard problem”.)
So, instead, can we suggest that mental properties do influence physical things? Carroll warns that the Core Theory of quantum mechanics has already been established to a very high degree of confidence, and it simply does not leave any room for this. He expands on that objection in the chapter on panpsychism: “One of the most basic things we know about the Core Theory is exactly how many degrees of freedom each particle has” (p. 364). You can’t just tack on new properties to the particles; we’d have found them already.
Though I have no ready response, that argument isn’t enough to make me give up property dualism; the intuitive case against reductionism still seems too compelling. But, it’s possible that I’m failing to adequately appreciate the force of his argument because I don’t have a detailed enough grasp of the Core Theory. That’s not something you can reasonably expect to gain from a popular-level science book.
If the laws of nature determine the future, can we still have free will? In his chapter on the subject, Carroll takes the compatibilist view, answering yes. One thing I appreciate is the clarity with which he outlines the debate. I’ve found that some proponents of free will take issue with the term “determinism” because of the uncertainty that quantum mechanics adds to the world, but Carroll states the argument against free will in a way that highlights the irrelevance of such uncertainty:
Quantum mechanics predicts our future in terms of probabilities rather than certainties, but those probabilities themselves are absolutely fixed by the state of the universe right now. A quantum version of Laplace’s Demon could say with confidence what the probability of every future history will be, and no amount of human volition would be able to change it. There is no room for human choice, so there is no such thing as free will. We are just material objects who obey the laws of nature.
But he rejects the conclusion that there is no free will, because this argument mixes vocabularies in the way warned against earlier. Free will may be a useful concept in the same way that molecules and chairs are useful concepts, despite not belonging to a description of nature at its most fundamental level. I think he does a good job here of explaining why it still makes sense to talk about people making “choices” even if the world is deterministic - something that proponents of libertarian free will occasionally regard as hypocritical or nonsensical.
Carroll is also right to point out that the significant difference between compatibilism and hard determinism comes “when we confront the notions of blame and responsibility” (383), but I’m unsatisfied with his discussion of that subject. He seems to simply assume that if choice is (in the poetic naturalist sense) real, then assigning moral responsibility makes sense. Further, he thinks this responsibility depends on us not having the practical ability to predict people’s behavior accurately. This leads to the surprising conclusion that people could have less moral responsibility as neuroscience becomes more advanced:
If our belief in free will is predicated on the idea that “agents making choices” is part of the best theory we have of human behavior, then the existence of a better and more predictive understanding could undermine that belief. To the extent that neuroscience becomes better and better at predicting what we will do without reference to our personal volition, it will be less and less appropriate to treat people as freely acting agents. Predestination will become part of our real world.
It doesn’t seem likely, however. Most people do maintain a certain degree of volition and autonomy, not to mention a complexity of cognitive functioning that makes predicting their future actions infeasible in practice.
I think this obscures an important point. Free will is often used to justify retributive justice - punishment done not for the sake of deterrence or rehabilitation, but because the person deserves to suffer in exchange for the choices they have made. The idea that such punishment could be more or less appropriate based merely on our practical ability to predict the person’s choices seems absurd.
Carroll provides a very clear and thoughtful exposition of the argument that there cannot be objective morality. There is a gap, he says (following Hume), between “is” and “ought” that simply cannot be bridged. Science, therefore, cannot teach us morality, it can only help us more effectively pursue whatever goals we happen to have; ethical philosophy cannot teach us morality, it can only help us systematize and refine the intuitions about morality we happen to have.
He is at least willing to face the distasteful implications of this head-on:
The lack of an ultimate objective scientific grounding for morality can be worrisome. It implies that people with whom we have moral disagreements—whether it’s Hitler, the Taliban, or schoolyard bullies who beat up smaller children—aren’t wrong in the same sense that it’s wrong to deny Darwinian evolution or the expansion of the universe. We can’t do an experiment, or point to data, or construct a syllogism, or write a stinging blog post, that would persuade them of why their actions are bad. And if that’s true, why should they ever stop?
But that’s how the world is. We should recognize that our desire for an objective grounding for morality creates a cognitive bias, and should compensate by being especially skeptical of any claims in that direction.
As he points out, this conclusion is less terrifying than people sometimes imagine. Whether there are objective moral facts or not, we can still reason with people who have similar core moral values as ourselves. And whether there are objective moral facts or not, we will often have to resort to other means of conflict resolution with people who have radically different moral values than ourselves - such as enacting laws to restrain them. Still, I do worry that many people would be less motivated to follow their own moral values if they stopped believing those values had a certain objective validity.
(I remain a proponent of objective morality; see e.g. Ethical Intuitionism by Michael Huemer for a defense of that view.)
Time & Entropy
Last year I read (or at least, as Tyler Cowen recently put it in reference to a book by the notoriously inscrutable philosopher Heidegger, I “looked at every page of”) Carlo Rovelli’s The Order of Time, from which I learned rather less than I had hoped about the scientific understanding of time. One of the ideas I didn’t really grasp from that book is how entropy can help explain the apparent difference between “past” and “present,” without that difference being part of any fundamental physical laws. Carroll’s explanation of the concept was a real “aha!” moment for me.
As I understand it, the idea is: the laws of quantum mechanics make no reference to time. For every valid change from one physical state to another, the reverse change is equally valid:
Eggs could unbreak, perfume could go back into its bottle, cream and coffee could unmix. All we have to do is to imagine reversing the trajectory of every single particle of which our system (and anything it was interacting with) is made. None of these processes violates the laws of physics—it’s just that they are extraordinarily unlikely.
Why is it unlikely for a shattered egg to unbreak? Well, each particle of the shattered egg can legally move in a bunch of different ways, which means there are a huge number of possible configurations that the entire collection of particles could end up in. But from the outside, most of those configurations are equivalent; the microscopic changes don’t alter the fact that it’s still a shattered egg. That’s what it means to say the shattered egg has high entropy - entropy is “the number of different states that would be macroscopically indistinguishable from the state it is actually in” (p. 57). Only a relatively small number of configurations correspond to an unbroken egg, so if you pick one of the allowed transitions purely at random, you’ll most likely be choosing one that goes from shattered egg to shattered egg, rather than one that goes from shattered egg to unbroken egg
An unbroken egg, on the other hand, has lower entropy, since most of the possible reconfigurations of its particles would make it broken. So a randomly selected transition is overwhelmingly more likely to be from unbroken egg to broken egg than from unbroken egg to unbroken egg.
Imagine an empty universe, with a timeline that is a series of discrete instants. (Think of the timeline as having a ‘left’ direction and a ‘right’ direction, but no inherent ‘past’ or ‘future’ directions.) Plop a perfect egg somewhere, anywhere, on this timeline. Now fill in the rest of the instants on the timeline, with the only constraint being that, between every pair of instants in both directions, all the particles must move in accordance with the laws of physics. If you select at random from all the allowable states for each instant, you’ll most likely end up with a timeline where the egg decays - in both directions from wherever you placed it on the timeline. This happens without any need for the laws of physics to make reference to a direction of time. Once you’ve committed to having a low-entropy instant at a certain point in the timeline, you get progressively higher-entropy instants surrounding it.
In our universe, the low-entropy moment is the Big Bang. Given that for whatever reason, the universe was in a very low-entropy state at that point on the timeline, it follows that the universe will get progressively higher-entropy as you get further away from that point. To us, that will make it appear as if there is a law stating that entropy increases as you go in the direction we call the future, but it’s really just that entropy is increasing as you get further away from the anomalously low-entropy point.
At least, that’s my understanding of the theory. I don’t know if it’s accurate but I find it fascinating. This interpretation yields an obvious question that I don’t recall being addressed in the book: on this model, if you could see back in time to before the Big Bang, might you see the history of another universe similar to our own, just with the directions of past and future reversed?
The definition of entropy quoted above also raises questions - what are the precise criteria for “macroscopically indistinguishable”? Is it subjective? The Rovelli book I mentioned does assert that entropy is a subjective measurement, but I found its discussion of that mostly incomprehensible at the time.