Philosophy, Physics, and How It All Fits Together

Richard Marshall at 3AM magazine has been doing a series of interviews with all kinds of thinkers, especially philosophers; some recent examples include Dan Dennett, Tim Maudlin, Rebecca Kukla, Alex Rosenberg, and Craig Callender. And I’m the latest subject. Given the venue, we talk as much (or more) about philosophy than about physics, and a lot about how they fit together.


Spoiler alert: I think it’s possible to have productive grown-up interactions between philosophy and science. I guess I’m just a radical bomb-thrower at heart.

Click through if this kind of thing floats your boat:

I think emergence is absolutely central to how naturalists should think about the world, and how we should find room for higher-level concepts from tables to free will in a way compatible with the scientific image. But “weak” emergence, not strong emergence. That is simply the idea that there are multiple theories/languages/vocabularies/ontologies that we can use to usefully describe the world, each appropriate at different levels of coarse-graining and precision. I always return to the example of thermodynamics (fluids, energy, pressure, entropy) and kinetic theory (collections of atoms and molecules with individual positions and momenta). Here we have two ways of talking, each perfectly valid within a domain of applicability, but with the domain of one theory (thermodynamics) living strictly inside the domain of the other (kinetic theory). Crucially, the “emergent” higher-level theory can exhibit features that you might naively think are ruled out by the lower-level rules; in particular, thermodynamics famously has an arrow of time defined by the Second Law (entropy increases in isolated systems), whereas the microscopic rules of the lower-level theory are completely time-symmetric and arrowless.

I think this example serves as a paradigm for how we can connect the manifest image to the scientific image. Sure, there’s nothing like “free will” anywhere to be found in the ultimate laws of physics. But that’s not the only question to ask; at the higher-level description, we should ask whether our best emergent theory of human beings includes the idea that they are (in the right circumstances) rational decision-making agents with freedom of action. Until we come up with a better description of human beings, I’m perfectly happy to say that free will is “real.” It’s not to be found in the most fundamental ontology, but it’s not incompatible with it either; it’s simply a crucial part of our best higher-level vocabulary.

This entry was posted in Philosophy, Science. Bookmark the permalink.

43 Responses to Philosophy, Physics, and How It All Fits Together

  1. steven johnson says:

    Is it really plausible that “rational decision-making agents with freedom of action” describes what people generally mean by the term free will? Even skipping over the equivocation in “right circumstances,” rational ignores questions as to the possibility of irrational motives. It’s not clear how “freedom of action” relates to decision-making or what the phrase actually means. It’s true that “decision-making agents” is clear enough, but my intuition is that this applies to some animals.

    Overall, the grown-up, productive exchange of philosophy and science here seems to have been to the disadvantage of clear thinking. This doesn’t seem a bit like a radical throwing bombs. (But let me congratulate you on your boldness and bravery!) It seems like a conformist smearing balms. (And I also congratulate you on your grown-upness.)

  2. Jeff Johnson says:

    I take your point that at the emergent level of human behavior, or of our interior conscious access to our own thoughts, that individuals can decide what they want to do, can resist coercion, and have a sense of freedom that they can live as they will.

    But even if we don\’t go down to the level of biochemistry, but simply pay very close attention to our subjective conscious preferences and choices, the ultimate reasons behind our preferences and choices are, if we are careful and honest observers of our selves, usually unknown to us. We don\’t know why we like things, why we choose things, and what makes is pick one activity over another other than “it feels right” or “it seemed like the best thing”. We don\’t know why we like the taste of broccoli (or dislike it), we don\’t know why we like a certain color of clothing or style of shoe. We just do. We may be able to state reasons, but if we keep breaking down those reasons into their foundations, it usually boils down to “just because that\’s how I am”, which amounts to the way our unconscious perceptions present things to our conscious mind.

    So a careful scrutiny of our thoughts already puts the notion of “free will” into serious doubt. It is just as true to say that we make choices by processes that rely on our conscious intelligence and its many cognitive abilities, but that the objects of such processes, the values, preferences, desires, wants, emotions, concepts, memories, etc. arise from our unconscious pre-formed beyond our control, and that the nature of these objects heavily if not totally determines the outcomes of our choices. So there is a real sense, staying at the emergent level of consciousness, in which “we” are spectators going through the motions of translating unconscious impulses into a narrative that explains our choices to ourselves and others, but ignores the underlying causes of our choices just as saying “air pressure” ignores the underlying causes of air pressure. We could say the balloon is free to choose its volume and pressure, and base that on the fact that we have observed various balloons at various volumes and pressures, or seen a single balloon slowly change its volume according to either loss of internal gas or due to changing pressure or temperature in the environment. It makes some semantic sense to say “this balloon wants to be this size”, but we know this is only a metaphor. And unless we can convince ourselves that when we want something, that we actually decide to want, rather than just having wants we can\’t explain that cause us to choose what we choose, then again free will is extremely suspect, even staying at the emergent level of subjective consciousness.

    If a dog is chained to a pole, then depending on the length of its tether, it has more or less freedom to roam. But it is a “freedom” that is not really free absolutely, but rather a relatively more or less constrained latitude. But people feel inside, and traditionally assume, I think, that our choices are unconstrained, that randomness, fancy, and inexplicable quirks and vagaries of mere whimsy can turn our choices on a dime independent of any reason or cause. If we compare humans to other mammals or computers, the difference can be likened to different lengths of chain that afford us greater degrees of freedom compared to, say, the mere passive mechanical responses to purely external causes we may see in a puppet or leaf blowing in the wind. Are these differences really due to a mechanism or property called “free will”, or are these differences due to more and more complex sets of cognitive abilities, fueled by metabolism, that afford us a greater space of options when deterministically analyzing environmental inputs and deciding on a response, also known as making choices or decisions?

    In the history of humanity, during the time language evolved, we know that human beings were deterministic systems in a material universe, so language evolved to describe the actual capabilities of such deterministic systems. At the same time, the conceptual orientation of most people that ever lived probably involved an assumption of dualism, of a ghost in the machine that was responsible for the animation that distinguished live creatures from dead, or from wood and stone, or even humans from mere animals. So while “free will” was applied to the actual observed behaviors generated by the complex space of latitude afforded humans by their deterministic intelligence, the conceptual basis of the term “free will” was falsely attributed to an imaginary capacity to be an original source of will, a kind of mini-prime mover, not subject to any involuntary forces or causes whatsoever when making “free” choices.

    So compatibilism rightly abandons the false conceptual basis, and simply continues applying the term “free will” to describe the decision making capabilities of humans, which has changed little in tens of thousands of year, even if our understanding has changed. It\’s little wonder that this borrowing of traditional dualistic language can work consistently in terms of meaning applied to observed human behaviors, because the behaviors haven\’t changed, only the understanding we have of the underlying causes of the behavior.

    But this is a mere continuation of tradition. This has some value, but it is also confusing, or conceals the new understanding of the basis of human behavior in some way. It is doubtful, in my opinion, that a disinterested third party with excellent analytical skills and no knowledge of the history of the term “free will”, would be likely to independently arrive at that term as the best possible description of human volition.

    It really all does boil down to a linguistic argument over what is the best way to describe the behavior of intelligent deterministic organisms called humans. There are many ways we could do this, and using the term “free will” seems to me like a lazy continuation of tradition that confuses rather than illuminates what is going on, simply because the word does come pre-loaded with a significant amount of semantic baggage from its history. It seems to me that the most important reasons people have for retaining the term “free will” aren\’t really related to maintaining fidelity to accurate illuminating description; rather it seems like emotional professional commitment to a particular outcome, and also a paternalistic fear that people can\’t really handle letting go of their traditional comforting notions of what humans are. In other words, I don\’t see a good reason for retaining the term “free will”, but I can see why it\’s easy and convenient to do so. But the best path isn\’t always the easiest and most convenient.

  3. AI says:

    The idea that we have some sort of “free will” is in the exact same category of feel-good beliefs as all the gods and religions. It is just as ill-defined, just as superfluous, just as popular and has just as much evidence to back it up – none.

  4. John says:

    I will believe that we don\’t have free will when I can plug what I did today into an equation and then find out what I am going to be doing tomorrow or next week.

  5. vmarko says:

    @ Jeff Johnson:

    “It really all does boil down to a linguistic argument over what is the best way to describe the behavior of intelligent deterministic organisms called humans.”

    What makes you think that humans are deterministic? There are much simpler physical systems all around in nature that are not deterministic. Why would a more complex system be deterministic if a simpler system is not?

    Best, 🙂

  6. FrankL says:

    Playing around with Sean Carroll\’s last paragraph:

    Sure, there’s nothing like “God” anywhere to be found in the ultimate laws of physics. Until we come up with a better description of human beings, I’m perfectly happy to say that “God” may be real. It’s not to be found in the most fundamental ontology, but it’s not incompatible with it either; it’s simply a crucial part of our best higher-level vocabulary.

    Usual disclaimer: I am not talking about the fully anthropomorhized God of certain organized religions whose adherents have traumatized so many into to the equally illogical position of atheism.

  7. Robin Kirby says:

    Hmm, some thoughts on the higher-level concept of free will. But what about the higher-level concept of tables?

  8. Jeff Johnson says:


    I think humans are deterministic because I\’m a materialist, and I don\’t think quantum mechanical indeterminacy plays a role in the brain. I think it\’s neurons and biochemistry.

    You wrote:
    “There are much simpler physical systems all around in nature that are not deterministic”

    Like what, for example?

  9. Stephen says:

    “There are much simpler physical systems all around in nature that are not deterministic.

    Like what, for example?”

    I have a quantum true random number generator on my desk. There is your non-deterministic physical system.

    “I think humans are deterministic because I\’m a materialist, and I don\’t think quantum mechanical indeterminacy plays a role in the brain. I think it\’s neurons and biochemistry.”

    Neuroscientist Micheal Persinger (of God-helmet fame) has been publishing papers saying he has shown a macroscopic entanglement effect in humans brains. We’ll see how this develops.

  10. Jeff Johnson says:


    To claim that theism and atheism are equally illogical, you need to think that there is a symmetry between faith and atheism, and some point of balance in between that is “logical”.

    But I see a great asymmetry between belief and non-belief. Just imagine any non-falsifiable proposition for which there is absolutely no evidence, e.g the famous flying spaghetti monster, or a pot of gold at the end of a rainbow. Why is it illogical to take the default position that most things we can imagine probably do not exist, and that it is improbable that we could simply invent some object in our imagination and expect it to exist? That seems very logical to me.

    For any phenomenon, of all the explanations we can formulate by conjecture, only one is true, and it is not necessarily among our first attempts. So the idea that a conscious intentional being created everything, an idea obviously cribbed from observing how humans make things, is no more likely than millions of other possibilities, most of which may be beyond our imagining. It seems thus very logical, consistent with laws of nature and probability, to think that a conscious intentional being or intelligence being responsible for all existence is quite improbable, and thus disbelief is far more likely correct than belief.

  11. Jeff Johnson says:


    How does randomness lead to the very stable and consistent behaviors of all organisms?

    I\’m skeptical at face value that Persinger\’s claims, as you describe them, are anything but empty hype, but I\’ll see what he has to say if course.

  12. Stephen says:

    @Jeff Johnson

    I am an agnostic. Theistic arguments which fail to impress me are:
    1) Arguments from ancient texts (ugh… just ugh)
    2) Fine tuning arguments (anthropic principle defeats these)
    3) Unlikelihood of life/evolutionary adaptations arguments (unlikely != impossible)

    But if you wanted to make an empirical argument for God’s existence I could respect (respect, not believe) it would probably look like this:
    1) Prove that Near-Death Experiences are non-physical.
    2) Then since NDEers report seeing God, there is your evidence.

    Point 1 is very, very controversial and medical and neuroscience experts argue back and forth on this (I’ve read both sides of the debate). I do not have the qualifications to judge. And it does not in any case constitute proof. But it is of enough concern to me that I do not believe atheism is rationally justified yet by the current state of science.

  13. Stephen says:

    @Jeff Johnson

    There are many random behaviors present in the biological world. Fruit flies have been found to have a lot of randomness in their flight behavior, for example. Other experiments show that if you require a person to press a button and measure the force, the force exerted by the finger randomly fluctuates.
    As for Persinger’s claims, there has been evidence for quantum coherence in biological systems (published in Nature) so it’s not completely out there.

  14. Stephen says:

    @Jeff Johnson

    As another example of indeterminate behavior of a biological organism, suppose I set my quantum RNG to return a binary value of 0 or 1. 0 means I go to McDonald’s for lunch and 1 means I go to KFC. Now there is no way in the Universe for you to predict which fast food joint I go to. It’s not free will, of course, but it is completely indeterminate.

  15. Pingback: Using “weak emergence” to reconcile physics and “free will”. | Gordon's shares

  16. Jeff Johnson says:

    Having had drug experiences in the past, having been knocked unconscious, and having had some very wild dreams, I\’m sticking with Occam\’s Razor on NDE and assuming people are experiencing anomolous brain states from anesthetics, lack of oxygen, or other causes.

    Are dice random? They appear to be, but they are actually a deterministic dynamical system that is complex enough that we can\’t predict it in practice, though in theory we could predict the outcomes quite well with the right sensors and computers. Probability is math we use to describe deterministic systems that are beyond our measurement and computing abilities to solve. It doesn\’t mean, when probability is applied, that the underlying system is necessarily truly random. So the fruit flies probably aren\’t random at all, unless you can prove they are navigating by some quantum indeterminacy. Intentionally erratic evasion is not random.

    In your fast food example, you are doing nothing random. You are deciding to act on an externally generated random input. Think of the human as a system. I claim that system is deterministic in the way it interacts with its environment, which need not be deterministic for the claim to hold. There is a difference between claiming that the biological systems in an organism operate according to deterministic physics and chemistry, and claiming that the entire universe is deterministic. I don\’t claim the latter. Just as computers use quantum tunneling in their semi-conducting solid state components, the entire system as a computing machine executes instructions deterministically. This is an analogy to humans, which at the atomic level no doubt are subject to quantum laws, but the larger scale behavior of the system is still deterministic. I doubt that natural selection could work properly if organisms were random and indeterminate. In order to evolve, the brain had to conform to deterministic laws.

  17. vmarko says:

    @ Jeff Johnson:

    Ok, this is how determinism goes away, in short. On a microscopic level, you have Heisenberg\’s inequalities, which give a fundamental bound on precision for position and momenta of any particle. Therein lies true, fundamental randomness. These uncertainties are of course extremely small, and one might say that in the everyday macroscopic world, and for brain-sized physical systems they are negligible.

    However, there is the second part of the story — chaos theory. Chaos theory basically deals with propagation of error-bars for a given physical system. If the system is described by complicated-enough equations of motion, it will display “chaotic behavior”, in the sense that the uncertainties of initial conditions will grow exponentially, and after a finite amount of time outgrow the size of the system itself, rendering it fundamentally unpredictable, i.e. nondeterministic.

    The amount of “finite time” above depends on the properties of the system. You can find typical times for various systems here:

    You can see that typical times for loss of predictability vary wildly, depending on the system. The behavior of the human brain is based on electrochemical reactions and hydrodynamics, so the longest possible deterministic period of time for the brain is between 2 seconds and 5 minutes. This is just a wikipedia-estimate. More precise estimates can probably be made if one puts some effort into it, and they can make the Lyapunov time only smaller than Wikipedia values.

    This means that the evolution of the human brain becomes nondeterministic after a very short period of time, definitely shorter than, say, 5 minutes.

    As for examples of other nondeterministic physical systems, they are all around us — the behavior of fire flames, the weather, the Solar System, etc. But my pet-peeve is an incredibly simple, and yet nondeterministic mechanical system — the double pendulum:

    The most fascinating thing about this example is to demonstrate how a very simple mechanical system, which obeys Newton\’s laws of motion, can still be nondeterministic. Its motion is extremely sensitive to initial conditions, and these exist only up to Heisenberg\’s inequalities. Thus the microscopically small quantum effects can have large influence on a very big mechanical system, rendering it nondeterministic, despite Newtonian laws of motion which govern it.

    So much for determinism.

    Btw, I can see the lack of understanding of this issue arising over and over on various blogs, so I\’m thinking of writing a more elaborate text about (non)determinism in nature, and posting it somewhere. When (if) I get that done, it will hopefully cut down on these discussions of determinism and clear up various preconceptions that people have about a “Clockwork Universe” and the like.

    HTH, 🙂

  18. Stephen says:

    Thank you vmarko, I was thinking about the double pendulum example as well.
    Randomness is even a evolutionarily beneficial adaptation. If a fly is flying randomly, it would be much harder for a larger brained predator to develop any strategy, innate or otherwise, to predict its behavior and catch it.
    The brain has a lot of sub-neuronal stuff going on in synapses, glia cells, various neurochemical transmitters, etc. It would be easy to imagine some truly random chemical process creating a chaotic reaction which propagates up into macroscopic behavior, causing true randomness.

  19. David Park says:

    The following discussion is partly inspired by the book: The Lifebox, The Seashell, and the Soul: What Gnarly Computation Taught Me About Ultimate Reality, the Meaning of Life, and How to be Happy by Rudy Rucker. Despite its somewhat flippant title this is a serious book about the nature of computation and the world. Rucker’s ideas in turn are partly based on the ideas of Stephen Wolfram and his Principle of Computational Unpredictability: Most naturally occurring (and specifically Class Four) complex computations are unpredictable.

    If we think of modeling the physical world as a computation, then an important consideration comes in. Computations can be broken into different classes and one class, which encompasses most of them, has the property that there is no shortcut to the result. Even if all the input is exactly known, and the rules are exact, there is no way to predict the result. One simply has to go through the entire process.

    Think the weather, or the economy, or the future course of evolution.
    This might give one the illusion of free will. We might plan to do something tomorrow but tomorrow we ‘change our mind’ and do something different. There was no way another person or we ourselves could really anticipate what we actually finally did. Even if no novel inputs occurred.

    Now think of a pencil perfectly balanced on its tip in an enclosure with no air currents and perfectly isolated from outside disturbances. It is only buffeted by the air molecules around it and subject to gravity. A molecule hits it from one side and it starts to tip to one side but then another molecule hits it and it tips another way. It performs a kind of random walk until it gets so far from the balance point that gravity finally has its way and the pencil falls in a certain direction. It’s deterministic but unpredictable or perhaps quantum randomness enters. In any case, would you say that the pencil was an ‘agent’ and that it ‘made a decision’? Would you say the pencil had free will?

    Now think of the visual illusion produced by a drawing of a staircase of blocks. First it looks like the blocks project out one way. Then after a while it flips and they project out the opposite way. Do you have the free will to control which way they project?

    Now suppose that you have been impressively taught by your parents: When buying an electronic device never buy the warranty if the device costs less than $500 and always buy the warranty if it cost over $1000. You have been strongly programmed in such cases and your action is automatic. But now you are purchasing a device for $750. Should you buy the warranty? You don’t know. You dither. One factor pushes you one way; another factor pushes you another way. There is time pressure pushing you to make a decision. Finally you think: “What the heck, I’ll buy the warranty.” Congratulations you made a decision. But how is this different in principle than that pencil being buffeted by the air molecules? We all make lots of decisions of that category. It gives us an illusion of free will – but is it?

    If a person’s actions are in some cases predictable because ‘they always do that’ and in other case unpredictable because they involve an unpredictable, but deterministic (or maybe true quantum randomness), Class Four computation, then how does that look different than the common view of ‘free will’? How could one pin down the definition of ‘free will’ so one could experimentally determine the difference? It might be that brain cell measurements can show, at least in some cases, that the brain makes the decision before the person is aware of it. But lengthen the time span enough and the Principle of Computational Unpredictability says that brain cell measurements cannot predict the result. This looks an awful lot like ‘free will’ as an emergent phenomenon.

  20. vmarko says:

    @ David Park:

    “In any case, would you say that the pencil was an ‘agent’ and that it ‘made a decision’? Would you say the pencil had free will?”

    The concept of free will needs a more strict definition if you want to discuss its (in)equivalence with randomness. That’s certainly an important issue to resolve.

    However, if you claim that the world is deterministic, then any conceivable definition of free will would be incompatible with determinism. Therefore, the lack of determinism (both on microscopic and macroscopic scales) demonstrates that the laws of physics do leave enough “room” for the concept of free will to exist in the physical world. In mathematical terminology, lack of determinism is a necessary condition for the existence of free will, but it is not a sufficient condition.

    I would say that inherent randomness is not equivalent to free will, but it certainly is a necessary piece of it. For a plausible model of free will, you may look at the old well-known two-stage-model of free will:

    In short, it first has a “free” stage, which basically generates random possibilities for various actions of the brain (imagination stage), and after that a “will” stage, which deterministically picks one of the imagined choices over the others (deliberation stage).

    So in this model the free will is not equivalent to randomness, but utilizes a combination of random and conscious deterministic processes. Consciousness acts as an “agent” that makes a “deliberate” choice for one among many randomly-imagined possible actions. In your example, the pencil does not have free will, because it does not consciously deliberate in which of the possible directions to fall down. The human brain, on the other hand, does have a free will, because it consciously deliberates which of randomly imagined actions to perform.

    The two-stage model is of course probably not the only possible model one can construct, but it is good enough to demonstrate how one can sort out the difference between randomness and free will.

    HTH, 🙂

  21. David Park says:

    @ vmarko

    If one cannot predict the action in advance, because it would require a simulation at the microstate level faster than it occurs in the real world, then how does it look different than free will – even if the physics were perfectly deterministic with all input known?

  22. vmarko says:

    @ David Park:

    I am not sure I understood your question properly, but I’ll try to answer anyway… 🙂

    Imagine a physical system that you can (in principle) simulate. Given some initial conditions, you can run the simulation (take whatever time is necessary, it doesn’t play a role for my argument), and arrive at a certain result A. Then you repeat the simulation again, for exactly the same initial conditions, and arrive at a certain result B.

    If the system is deterministic, the two results must be the same, A=B.
    If the system is not deterministic, the two results may be different, A=/=B.

    It is assumed that the simulation of the physical system is “perfect”, in the sense that it takes into account all properties of the system. Including randomness, if present.

    So in a deterministic system there is no room for free will, in the sense that the physical system could not have behaved differently, given the absolutely identical initial circumstances.

    If you can obtain two different results on two instances of simulation, given the identical initial conditions, then the agent (either simulated or real) can in principle behave differently, which allows for the possibility of (but still does not imply) free will.

    HTH, 🙂

  23. David Park says:

    @ Marko

    The point is that you can’t do the simulation in time to anticipate the result, unless you have a machine that can reproduce the relevant part of the world (a brain with trillions of connections?) faster than it happens in the real world. So for all practical purposes an entity did something that you couldn’t predict. Does that look like free will? It sort of does, even if it is deterministic.

  24. vmarko says:

    @ David:

    Oh, now I see what you mean. Sure, I agree with you, but there are probably people who woudn’t, and they do have a valid point — what you describe is a technical problem, rather than a conceptual one. They typically assume that there “might exist” some super-intelligent alien race, with technology developed way beyond our own, etc., and they “might” have a machine which could make those simulations fast enough.

    So it is a question about the difference between possibility and feasibility.

    It is like creating a wormhole in general relativity — you “just” need to carefully create two big black holes nearby, and then move one of them away. Conceptually speaking, it’s certainly possible. Practically speaking, there is absolutely no way that can be done. 🙂

    But philosophy tends to talk more about concepts than technical solutions.

    Best, 🙂

  25. Torbjörn Larsson, OM says:

    Despite the quite blatant attempt to both poison the well of criticism and deflect analysis by fuzzily describing science models as “languages/vocabularies” which they of course aren’t*, I find basis analysis, and its implied criticism, quite effective.

    What can possibly be the useful interaction between the anti-story telling of science, which is constrained by testing, and the story telling of philosophy/theology/religion/literature and similar pursuits?

    The best interaction I know of between technology & science and literature is science fiction. Compared to that philosophy is primarily boring, and a lot less influential which should be what counts. (But philosophy is *so* boring, has no relation to the world at all. “Ontology”, really?)

    So, if anyone still like philosophy after millenniums of going round in the same circles, good for you. But I don’t see why anyone would expect others to share the liking of its moldy way of story-telling.

    Sci fi on the other hand sell quite well, I hear, which may or may not correlate to its actual influence on society.

    *A model isn’t just terminology and the rules you choose for that. It is also a way to make and test predictions, which an abstracted “language/vocabulary/ontology” is not.