Monday, February 21, 2011

Mailbag Monday: Free Will

Mailbag Monday: A weekly segment that covers readers' questions and concerns about all things Philosophy, Bro, and Philosophy Bro that don't quite fit anywhere else. Send your questions to with 'Mailbag Monday' in the subject line.
Ricky writes,
Can you do free will or determinism please?
your fellow bro
Alright, bro, I'll take a crack at it.

First of all, though, there's a more fundamental question than free will or determinism. For some reason, lots of bros seem to think that if determinism is true, free will is obviously an illusion. It's an intuitive line of reasoning - 'if the future is already determined, then of course we can't have free will. Fucking elementary, Watson.'  Hold your horses there, asshole. Not so fast.

Put a bro in a maze - but not a normal maze. This maze only has one path, from the start to the finish, and no forking paths at all. But this maze does wind in all sorts of directions, twisting and turning on the way. Now, let's put a blindfold on our intrepid hero. Basically, if this dude makes a wrong turn, he's going to slam into a wall. Now, at the end of this maze is a fucking hottie, and obviously this bro is going to get to her. He can hear her humming to herself, and he's like a bloodhound on a scent. So he starts walking forward, toward the sound, until he gets the sense that it's coming from the left - so he turns left and walks forward until it's coming from the right - and so on. Eventually, he gets to the end, she's thrilled to see him because girls love bros, and they live happily ever after until someone even hotter comes along and he peaces out.

The point is, even though this bro's path was determined - he couldn't have gone forward when he had to go left - he had no idea. He just followed his motivation and ended up where he wanted to be without ever feeling the walls the confined him. His path was determined, sure, but he also freely chose it. At that first turn, had he wanted to go right instead of left, the sound would have been coming from the right - otherwise he wouldn't have fucking chosen right. He's just trying to get to the slampiece. So, had he acted otherwise than he did, the determining laws would have been different, but not as a result of his turning right - just in conjunction. If the bro had turned right, it wouldn't have caused the walls to be different - they just would have been different the whole time. And we definitely know the bro could have turned right - after all, later on he actually does.

The thesis that free will and determinism are compatible is called compatibilism. Maybe that argument works, maybe it doesn't - honestly, I doubt I just resolved the compatibilism debate in two paragraphs, but if so then I'll be leaving and you can find me teaching kindergarten - but the point is that free will and determinism aren't as obviously incompatible as you might think right away. In fact, the problem of free will might just be the deepest fucking rabbit hole philosophers have yet dug; if you don't believe free will and determinism are compatible, then you've got your own fucking problems, champ. You have two options: either deny free will or deny determinism. 

If you deny determinism, or embrace indeterminism, you believe that the future isn't determined by the laws of physics and the way things are. Even with the present laws, right now, maybe the future could be a bunch of different ways. Some bros think, for example, that quantum physics is indeterminate. I won't commit myself to that, because I don't have a firm enough grasp of what is a complex and highly technical piece of physics to make claims one way or another (Quantum Mechanics, Relativity, and Gödel's Incompleteness Theorems are the most often abused concepts, ever) but even if that were the case, even if physics is indeterminate like that, no one wants to say that random particle decay is the source of free will. That's not any more free - particle decay is completely random, and a random choice is no more free than a coin flip.

Look at it this way: if I decide, "Yeah, I'm going to fucking bungee jump!" There has to be some way for me to get from that decision to actually bungee jumping. That will eventually require me to willingly strap a bungee cord to my legs and jump off a bridge. So a neuron fires that sends the signal from my brain to my legs that says, "TIME TO FUCKING JUMP." And that signal travels down some nerves to my muscles, my muscles contract, and I go leaping into space like a fucking boss. Fuck yeah. But everything in between making the decision to jump and me jumping is a deterministic process. It seems like we need some determinism to execute our decisions - if free will depends on indeterminism, we could never get to that end result. If making the decision is random, it's still not free - it's as good as making a decision based on a die roll.

So maybe you deny free will instead - after all, there are pretty good arguments that say that free will isn't compatible with either of two opposites. Alright, then explain this: what can we say about right or wrong? If you answered anything other than "nothing", you have some fucking explaining to do. If I told you, "You know, you really should have had sex with Marilyn Monroe while she was alive," you would look at me like I was a fucking idiot. "But, uh, she was, you know, dead, before I was born." "Yeah, well, that's sure as shit not my problem. You should also stop the genocide that's happening on Mars. HELPLESS MARTIANS ARE BEING WIPED OUT. And you're just sitting there surfing teh interwebs. Shameful." None of that makes any sense - of course it's not fair to expect you to do something that you have absolutely no capacity to do. But if free will is an illusion, then we don't have any ability whatsofuckingever to do anything except what we actually do. We could no more ask each other to act morally than we could ask each other to not be so awesome. And that's just crazy-talk. Bros are always awesome

So, there you have it. It looks like we have arguments that free will is incompatible with determinism and indeterminism, but without free will our moral judgements, all our moral judgements, whether right or wrong objectively exist, are meaningless - that is, some things are obviously wrong, or defunct in some way - some decisions are better than others. So which arguments are faulty? Where did our intuitions go wrong? That's fucking tough. Consider this: when we find a mechanism for decision-making - say, a neuron firing - if it falls under some deterministic process, determinists will say, 'aha!' and question free will. If it falls under some indeterministic process, it will appear random and indeterminists will say, 'so there!' and question free will, since by definition no process can predict or impose order on an indeterministic process. Perhaps we have no way of knowing at all what free will would look like, even if we were staring it right in the fucking face.


Further reading:

David Lewis and Peter van Inwagen, a compatibilist and an incompatibilist respectively, are probably the clearest meta-writers about free will - they do the best job of framing the debate. Van Inwagen's How to Think About the Problem of Free Will lays out the problem incredibly clearly, and serves as the basis for this post. His An Essay on Free Will is a much more in-depth treatment of the topic.

Lewis' Are We Free to Break the Laws? (pdf) lays out an excellent, though technical, case for compatibilism. 

Daniel Dennett's Elbow Room: The Varieties of Free Will Worth Wanting discusses different versions of free will with respect to determinism, and is a good technical treatment of the topic for those who want something a little deeper than a single paper.


  1. Nice -- I just spent some time reading (and reading about) Inwagen.

    There's one other position worth noting -- the "semi-compatibilist" view, which is the one that says "yeah, but whether we're determined or undetermined, we're still morally responsible."

    Here ya go:

    There are decent models for how moral responsibility could apply even in the presence of hard determinism. My favorite example is the Uruk-Hai -- those guys were BAD, right? Born bad. No choice in the matter. But killing them was still _right_ -- as they say in Texas, "they needed killin'".


  2. Using italics for relevant cues is a nice fucking touch. God bless you.

  3. I've never been sold on the idea that randomness associated with particle decay is some how incompatible with, or creates tension with, free will. It seems like at best question-begging.

    If you grant that there is free will--that is, a completely free and unpredictable source of decisions, and then try to scientifically study the source of these decisions, you'll be led to think that they're random. But they're not, (by assumption). Randomness is not the true explanation, it's just the best scientific one, because free will is completely free and unpredictable (by assumption) and therefore can't be discovered by scientific study. The perceived randomness is not the source of free will; free will is the source of the perceived randomness.

  4. Btw, don't mean to a douche, but I find your use of the term slampiece pretty offensive. I like the site, and I get that a lot of the language is meant in jest, but I think there's a line between funny and incendiary, and it's been crossed.

  5. Ben,

    If something is inherently unpredictable, then it is random. Were there, as you suppose, any hidden internal structure to such a thing, that structure would necessarily impose corresponding regularity on whatever followed from it -- this regularity would, in turn, be observable.

    This is the underlying principle behind why no random number generator creates strings of truly random numbers, which I'm sure you've heard before; there are always internal patterns to those numbers, and they are a function of the structure of the generator itself.

    You're trying to have your cake and eat it too.

  6. Melanchthon,

    The fact that something has an internal structure guide it does not imply that any patterns resulting from its behavior would be recognizable as patterns. When we say that particle decay is random, what we mean is that we cannot distinguish a pattern that allows us to predict behavior beyond a certain level of accuracy; that doesn't imply that there's no pattern. To write off this possibility is the question-begging I referred to.

    Methinks you're trying to neither have your cake nor eat it, and surely, surely something must be done about that cake.

  7. Bro, I love this shit, but I fail to understand something.

    You see, I guess I'm one of those who deny free will. However I don't think that this makes all our moral judgments meaningless. A cogwheel inside my watch doesn't have much choice which way to roll, but for sure it isn't meaningless, is it?!

    If I tell a bro I morally disapprove him beating up his children, I may plant a thought inside his head. Let's say he starts to long back to my grace or he actually starts to see the harmfulness in his behaviour, and thus he stops beating the kids. If this happens, it's not out of his own free will any more than it is out of mine, but it still means something. At least to me it does. No?

  8. "The fact that something has an internal structure guide it does not imply that any patterns resulting from its behavior would be recognizable as patterns."

    Agreed, though what's important here is that the presence of such structure necessitates that it could be recognized -- whether or not it actually is recognized is besides the point.

    This is what I feel you overlooked when you originally said that "free will is completely free and unpredictable (by assumption) and therefore can't be discovered by scientific study." (emphasis added)

  9. "the presence of such structure necessitates that it could be recognized"

    I don't see how you get here, unless you assume hard determinism a priori. How do you reject the possibility that a structure could cause events to occur in non-random ways, yet not be recognizable as a non-random actor?

    Are you assuming the existence of an omniscient being, which could detect the patterns inherent in the causal structure? If so, you're just assuming away the possibility of free will by defining an omniscient being such that free will is impossible.

    Alternately, you could try to argue empirically, but this brings us back to my earlier point that particle decay could either be random, or it could non-random, but nonetheless appear to be random because we can't understand it, but in fact be caused by some free actor.

    Or perhaps you have another approach...

  10. "How do you reject the possibility that a structure could cause events to occur in non-random ways, yet not be recognizable as a non-random actor?"

    Here's my reasoning:

    1: If events occur in non-random ways, then there is a pattern to those events.

    2: If those events can be perceived, then their pattern can be perceived as well, as by definition their pattern is the sum relation of those events to each other.

    3: Since effects are non-random if and only if their cause is non-random, if a) the events we perceive are the effects of an actor, and b) those events are non-random, then c) that actor must be non-random as well.

  11. Mel,

    I reject 1 as a false dichotomy. I'm positing that particle decay / human behavior are neither random, nor do they have patterns. I'm saying they're caused by free actors, with free meaning that they're bound neither by rules nor by randomness.

    When scientists study phenomena, they can fully separate them into predictable ones and non-predictable ones. They can call these groups patterned and random, but to do so is to define a map, not to accurately describe the territory. The framework I describe, wherein there are patterned, random, and also free actors, can also be split into predictable and non-predictable phenomena, and these phenomena can be given the labels patterned and random accordingly. We agree on the map; we disagree as to its implications regarding the territory.

  12. "I'm saying they're caused by free actors, with free meaning that they're bound neither by rules nor by randomness."

    So, you say there are three options: 1) Bound by rules, 2) Random, and 3) Free.

    You're being cruel to me! You've identified -- correctly -- that I think 'bound by rules' and 'random' are mutually exclusive and exhaustive. Hence, I clearly can't imagine a state of being that isn't one or the other of those two. But, that's exactly how you defined "free"; free is anything neither bound by rules nor random.

    You've left me without any idea of what "free" is aside from what it is not -- can you help me by offering a positive definition of freedom, instead?

  13. Mel,

    I don't know if I can help you, but I can try and re-frame:

    Consider hard-determinism, which says that all behavior is predictable based on prior events, as there are firm rules tying everything together. In a hard-deterministic world, there is no randomness. When an event is unpredictable, like particle decay, the it must be explained away by saying that while particle decay is unpredictable now, it is defined by hard rules, and some day we will be able to predict its behavior more acutely.

    Now suppose you want to construct a different model, in which random events can occur. In this second model, you allow that unpredictable events like particle decay may actually be truly random. Hard-determinists will object to this model, saying "there's no such thing as randomness", since that's the lens through which they view the universe. Furthermore, they'll challenge the very definition of randomness, since it contradicts their worldview. If random is defined as not predictable, hard-determinists will protest that this is a negative definition, and that the idea of randomness is unimaginable. They'll ask for a positive definition of randomness that makes sense within their worldview, but it's impossible to provide one because the hard-determinist lens blocks the possibility of randomness, even if the universe itself does not.

    This is the predicament I find myself in. I want to add free will into the model as a third option. I want to say that when I choose between Corn Flakes or Cheerios, my decision is based neither on a hard rule, nor is it a random decision; the choice is controlled by me, a non-random actor, but my choice is unpredictable. Within either of the two earlier frameworks, free will is impossible, just as randomness is impossible within a framework where everything is predictable. The very act of defining free will necessitates rejecting both earlier models.

    To fairly evaluate which model is best is complicated and difficult, but it's impossible so long as you embrace model 1 or 2, since to do so is to rule out the possibility of model 3.

    Free your mind, bro.

  14. Blech, that comment was too long. Sorry everyone. Sorry electrons.

  15. Thanks Bro for covering the subject! Anyone who's interested in the problem of free will can check out this textbook

  16. Ben,

    Actually, I thought that was a great response -- not too long at all.

    You argue that just as 'random' cannot be positively defined from a hard determinist perspective because the one excludes the other, 'free' cannot be positively defined from either perspective for the same reason.

    I want to challenge this by providing simple, positive definitions for both 'random' and 'determined'; I hope you take this as a challenge to do the same with 'free':

    -Determined means that from identical causes, identical effects must follow.

    -Random means that from identical causes, identical effects may follow.

    From my perspective, a false dichotomy rests between 'random' and 'free'; I see them as meaning the same thing.

  17. Whoa, dude. TLDR.

    Believe in free will, bro. Or don't. It's totally up to you.

  18. Mel,

    I'd separate random and free as follows:

    -Random causes can be modeled using a probability distribution and the model will be accurate.

    -Free causes can be modeled using a probability distribution and the model may be accurate.

    It's the difference between a known unknown and an unknown unknown.

    Consider the board game monopoly. The rules of monopoly are defined by three elements: hard rules--if you land on go to jail, you go to jail, don't collect $200; random events--dice rolls; and player decisions--buy a property, or don't.

    Dice rolls and player decisions are obviously not the same thing. But if you studied monopoly without knowing the rules, only by observing gameplay, you might think they're the same. Over time, you'd see that players roll 7 at 1/6 probability. You'd also see that players who land on Pacific Avenue mortgage a property to buy it at 1/6 probability. So you might think these are both random events. But if you modify the rules of monopoly such that player decisions are decided by dice rolls, surely you create a very different game.

    In the much more complicate game of The Universe, we don't know the rules; we can only guess at them by observing gameplay. So we don't know whether particle decay is the result of a metaphysical dice roll, or a metaphysical free choice.

  19. Interesting! I'll use your definitions to model both random and free responses to a simple choice, what to do at a fork in the road. I'll also model the determined response for contrast.

    There are three options: go left, go right, or go back. For consistency, each model will choose between paths repeatedly and under the same circumstances.

    First, the determined response: Determined means that from identical causes, identical effects must follow.

    The circumstances and hence the causes are identical, so the same choice will be made each time. Hence, one path gets a probability distribution of "1", and the other two get a distribution of "0".

    "Random causes can be modeled using a probability distribution and the model will be accurate."

    This one's easy; there are 3 options, so the probability for each possible choice is 1/3.

    Let's see how this differs from the free choice.

    "Free causes can be modeled using a probability distribution and the model may be accurate."

    The key difference between random and free here is the difference between will and may, so what does may entail?

    It means that however we model the free response, it can behave according to a different model.

    This implies that the free response cannot, after all, actually be modeled (contra the first part of its definition). Let's ignore that, though, since we can actually work around it; let's make a meta-model! We simply model the free response in every possible way (since we know it can freely move between them), and then model those models.

    This turns out to be pretty simple.

    Since we are modeling every possible model, we see that for every model that has a probability distribution of 0 turning left, another will have a distribution of 1. Likewise, for every model showing 0.1 probability of turning left, another will show 0.9 probability.

    The thing to notice about each of these pairs is that they all add to 1, and that every possible option will have an equal number of such pairs within it; this means that our meta-model will show equal probability for every possible choice, which for 3 choices we express as 1/3 probability.

    This is the same as the random distribution.

  20. Mel,

    We're using different meanings of the word model. The way I understand a model, it isn't a rule that phenomena must follow; it's an interpretation, crafted by observers of the phenomena, with the aim of predicting future behavior. A model is a map, not the territory. The limitations constraining the creation of a model are the availability of observations and the creativity of the observers.

    There may be metaphysical rules that determine a given phenomenon's behavior, but I don't take this as a given. Some phenomena may be guided by a deterministic rule--from identical causes, identical effects follow. Some may be random--from identical causes, effects follow according to a probability distribution. But I maintain there's a third possibility--that there is no rule. Some phenomena are free to behave as they choose.

    The lack of a rule doesn't prevent observers from (incorrectly) modeling the behavior. You can model how I'll behave at a fork in the road. Go ahead and try. But I maintain that I do what I want when I want, so your model, whether based on determinism or probabilities, will be inaccurate.

    In short, I take your model and I throw it on the ground. I'm not a part of your system.

  21. "We're using different meanings of the word model. The way I understand a's an interpretation...with the aim of predicting future behavior."

    I understand the same thing by a model, so we're not using different meanings of the term.

    A model can be either accurate or inaccurate; when we say that something "can be modeled," we mean that it can be modeled accurately, which means that it maps onto its territory successfully.

    Hence, I used your own definitions in good faith to refute your own position, that there is a distinction between "free" and "random".

    I would be happy if you could refute my argument in turn, since I, too, would like to believe in your notion of freedom, but it seems you won't give that to me.

  22. Mel,

    I don't equate "can be modeled" and "can be modeled accurately". I say free actions can be modeled, but the models are inaccurate.

    "It means that however we model the free response, it can behave according to a different model." -- add "or it can not behave according to any model."

    "We simply model the free response in every possible way." -- you can model the free response any way you can imagine; that doesn't mean you've modeled it accurately.

    Do you still think the refutation holds?

  23. "Do you still think the refutation holds?"


    "I say free actions can be modeled, but the models are inaccurate."

    This just seems to be self-contradictory; perhaps the definition of freedom we're using needs to be changed? I think it's what's getting us into trouble.

    If, as we agree, a model is a representation, then this seems to be saying that free actions can be represented, but those representations will not be of free actions, which is the same as saying that free actions can't be represented.

    "It means that however we model the free response, it can behave according to a different model." -- add "or it can not behave according to any model."

    This has the same problem as above, since it seems to contradict our definition of 'free'.

    I think though, that this addition isn't necessary, as its spirit is the same as the original: If something can always behave according to a different model, this is functionally identical to it not behaving according to any particular model (since it would always be able to behave by a different one).

    ""We simply model the free response in every possible way." -- you can model the free response any way you can imagine; that doesn't mean you've modeled it accurately."

    If we've created models for every possible permutation, then by definition one of them will always be accurate since otherwise the free response would have to behave in an impossible manner.

  24. Spam filter ate both, Mel. Published the first, deleted the second. carry on.

  25. Mel,

    Does the idea of an inaccurate model make sense? Suppose I model the behavior of a fair die such that it always rolls a 4. That's a model. It's a representation of the behavior of the die. Sometimes it will be right. But it's fantastically inaccurate. There are countless examples of people creating inaccurate models (flat Earth, ether, Greek gods, etc.)

    Does "Free actions can be modeled, but the models are inaccurate." still seem self-contradictory? All your criticisms follow from that point. I suspect you want to assume a priori that all phenomena are defined by rules, but to do so is begging the question.

  26. The idea of an inaccurate model does make sense, I agree. As I see it, the question is what it's a model of.

    I see a model as constructing a fictional object, what we call the representation.

    I don't think that model is 'tied' to any particular real object; rather, the fictional object can be compared to any or as many real objects as we want.

    For example, consider the model where a die always rolls a 4: When applied to a fair die, the model is woefully inaccurate. However, when the same model is applied to a loaded die, it could be completely accurate.

    It seems fair to say that, whatever we call it a model of, it is actually a model of a loaded die.

    Similarly, the model of a flat Earth is not actually a model of this planet unless this planet is flat; the intent of the author in modeling this planet is immaterial to the actuality of the model itself.

    The question of whether something is or is not a model of something else is not subjective but objective; the answer lies in the model itself and what we compare it to.

    This is why I say that "free actions can be modeled, but the models are inaccurate" is self-contradictory; the models are either of free actions or they are not, and if they are inaccurate they are models of something else, just as the model of the fair die was actually a model of a loaded die.

  27. If a model needs to be perfectly accurate in order for it to count as a model of a particular phenomenon (otherwise, as you suggest, it's actually a model of something else), then I'd hold that very few, if any, of our models are actually models of the things they're supposed to represent. There are no truly fair dice, so we have no model of real-world dice. How, then, could we possibly assign a model to human behavior, which is so vastly more complicated than a die roll?

  28. Let me re-frame my entire argument:

    1) Human-designed models of the universe are constrained by the limits of observable phenomena, and by the limits of human imagination.

    2) The universe itself is not constrained by these factors.

    3) Therefore, there may be phenomena that cannot be explained accurately by human-designed models.

    3a) Free will is one of these phenomena.

  29. Ben,

    I agree with your conclusions qua models; I think the implications of and responses to this constitute some of the most interesting topics in modern epistemology.

    Regarding your 10:07 argument, 1, 2, and 3 all seem true.

    3a, however, is resting on an additional implicit claim I'd like you to address explicitly before I accept it: If 3a is true, then from 1 we see that it must be true because we either a) have a deficiency in what we perceive (we are missing pieces of the puzzle), or b) have a deficiency in our ability to order what we perceive (we have the pieces but can't fit them together).

    When you assert 3a, which of these do you assume? If we are missing necessary pieces, what brought you to believe that?

  30. Mel,

    I'm pretty comfortable asserting either a) or b)--I think our observations of human behavior are fairly limited and that our models of human behavior are far from accurate. But I can't prove either claim.

    Instead, I'd propose that we accept s viable a number of different models--hard determinism, determinism plus randomness, determinism plus randomness and free choice. Any of these could be true, and the question of which to use in a particular application is, in my mind, largely pragmatic. I find worldviews that omit free will to be fairly useless for ethical or legal questions, for instance, and I consider these questions important. (I'm not very fond of compatibilist solutions)

    I'm really not trying to convert you to a particular belief set. I'm trying to defend the concept of free will as viable, by showing its implications to be internally consistent, and consistent with observable phenomena.

  31. Hmm, well, when you're missing a needed piece of the puzzle, you know it because the whole thing won't fit together -- conversely, it seems that if the whole thing does fit together, neither a) nor b) can be true.

    As far as I can tell, hard determinism has no explanatory deficiency, nor does it lack proscriptive power -- even when it comes to ethical or legal questions. Oddly, I see explanatory deficiency and proscriptive impotence arising once we leave the determinist perspective.

    I think a lot of ideas are held simply because they're familiar or otherwise comforting, but I have to agree with Socrates on this one; I want the truth, whether I like it or not. I rely on people who disagree with me to set me straight when I'm wrong, so I feel like an intellectual truce is the only condition under which no one benefits from the conclusion of a debate.

  32. Mel,

    Surely there are resolutions besides truce and and outright victory/defeat. Going back to the map/territory divide, I'd always prefer two models that are different and usable to just one. We agree that both a deterministic model and a model that includes free will are consistent with available observation. We don't know which model is accurate, unless you have further evidence to support a deterministic view. Isn't this a different--and better--place than where we started?

    I'll also bite on the question of which model is more useful. The foundational ethical question is "What ought I to do?". In my reading, this question presupposes the ability of an actor (I, the self) to freely choose between different options. I don't see this question making any sense in a deterministic world, even one that allow randomness. If there's a determined rule governing my behavior, one that's outside my influence, why am I even thinking about my behavior?

  33. "I don't see ["What I ought to do?"] making any sense in a deterministic world, even one that allow randomness. If there's a determined rule governing my behavior, one that's outside my influence, why am I even thinking about my behavior?"

    Determined or no, we all experience what we call "free will"; the debate isn't over whether we have that experience or not, it's over what that experience is in a metaphysical sense. Since a determined universe threatens only our conception of free will rather than the actual object of that conception, if we make choices today--and we do--we'll still make choices in a determined tomorrow.

    The difference is that we'll have a better understanding of why we make the ones we do, rather than them just magically appearing out of our "freedom".

  34. For an excellent techincal treatment of the quantum mechanics and free will, see Henry P. Stapp, Mind, Matter and Quantum Mechanics and his later Mindful Universe: Quantum Mechanics and the Participating observer.

    He does a good job of sticking to the theory (i.e. leaving out most of the math). Not sure I understand it quite yet, though...