Monday, January 20, 2014

... except under some very specific circumstances

In my previous post I argued that evolution does not, as a whole, have any direction or goal, but I qualified that with the claim that there are some circumstances when it does, and when it even makes sense to think in terms of better and worse or more primitive and more advanced solutions to an adaptive problem. So what would those be?

What I am thinking of here is a situation like this: A lineage enters a new habitat or acquires a mutation that allows it to exploit a hitherto unavailable resource. At this stage, the organisms in question are not particularly good at making use of the new environment but since there is nobody better around they can nonetheless flourish. Crucially, among the members of this lineage the race is now on to get better at making use of the new environment, and that is what gives this situation a directionality.

As an example from the plant kingdom, we may consider the first green plants that made it onto dry land and moved into what we might term the 'big land plant niche'* hundreds of millions of years ago. The ancestors of our contemporary vascular plants were, as far as we can infer, basically green blobs bound to the moistest conditions. What happened then?

  1. The first step was to evolve some protections against water loss. This came in the form of a waxy cuticle covering the surface of the plant body. However, that also makes gas exchange with the environment more difficult. A conundrum: you need to breathe but breathing loses you water. The solution: stomata, that is pores in the surface that can be opened or closed depending on circumstances.
  2. Now often it is an advantage to grow bigger, and in fact we are talking about the big land plant niche here. But then arises another problem because it is difficult to distribute water, nutrients and assimilates throughout a big body. So the next crucial innovation were vascular bundles, strands of tissue that act as the transport system inside a plant body. They contain two types of tissue: the xylem transports water and nutrients from the base to the apex, the phloem transports sugar and amino acids to wherever they are needed. As an added bonus, lignified xylem can also provide structural support to the plant.
  3. By now the plants are growing into longer and longer branch systems but they are still only anchored and acquire water and nutrients with a few ridiculously small cells called rhizoids. That won't do. How about sticking a few branches into the ground? Sure, they will not be able to photosynthesise but they can anchor the plant safely in place. Even better, they can reach for deeper water sources, allowing it to grow in superficially dry places. Congrats, we have evolved roots!
  4. We also notice that there is a limit to how big we can grow if we only rely on branching stems. If they are too thin, there is not enough structural support. If they are too big, there is too little surface for gas exchange and photosynthesis to feed the massive stems. Perhaps it is time for another round of specialisation. How about we have a fat stem for support but make the terminal branches flat so that they have a lot of surface? We need a different word for that kind of branch though, what do you think of "leaf"?
  5. Now everybody can grow quite big already but that only means that it is important to grow even bigger. After all, we need to over-top our neighbours to get more of that delicious sunlight. Sadly, we have hit another obstacle. There is a limit to how thick we can reasonably grow our stems and roots in one go before they need to reach for the light and water, respectively. Perhaps if we put a continually growing layer of cells into the stem and roots, to make more xylem and phloem as need, we could grow even thicker and thus support even more height. Achievement unlocked: secondary growth. You are now a tree!
  6. Somewhere along the line, we might wonder whether it is really such a good idea to have motile, flagellated sperm cells. That worked well when the ancestors of the land plants were aquatic, but now that they are on dry land it seems a bit silly that one still needs a film of water to connect male and female reproductive organs to have sex. Any ideas?
I will stop here lest I get lost in the wacky minutiae of land plant sexuality.

It is important to understand that I simplified a lot and that the steps described above did not necessarily occur in that order. Indeed different lineages of land plants evolved some of these innovations in varying order, or not at all. Obviously also individual sub-lineages of the land plant lineage went back into the water: in the case of duckweed or sea grass, the race is on to become more algae-like because that is what makes them better at surviving in the new habitat they have now entered.

The point is simply that the first land plants were poorly adapted to the new environment whereas their descendants have had a long time to get better at dealing with it. Contemporary plants can deal with the problems of living on dry land in a way that makes their ancestors look primitive.

But that is just it: their ancestors, not other contemporary land plants. Even those that have not yet evolved secondary growth, for example, or even secondarily lost it, are apparently good enough to thrive. They aren't primitive, otherwise they would not have survived. But still we might say that something like Aglaophyton, without leaves and roots, represents a first primitive solution to the adaptive problem of living on dry land compared with later, more advanced solutions represented by its own descendants. In other words, it does not make sense to talk of primitive and advanced in one time slice but it can sometimes be justified when looking across time.

Similar situations abound, especially at small scales: after a pollinator shift, in plants colonizing a newly arisen mountain or island, in a parasite changing hosts, etc. Thus evolutionary biologists distinguish stabilizing or purifying selection, in which a population is already well adapted to its environment and all deviations are selected against, directional selection, in which a population is over time becoming better adapted, and disruptive selection, in which a population is slowly split in two directions. In the second case the concept of directionality is already included in the name.

Of course, none of this implies that there is any plan or purpose behind evolution. The directionality comes simply from the involuntary movement of a population onto a new peak on the fitness landscape and as far as we can tell not from some cosmic intelligence pushing things in that direction.


*) A keen observer will notice that I am neglecting the bryophytes here. Theirs is a second viable strategy for living on dry land but it is completely different in that they do not attempt to keep their water behind a cuticle when it gets drier, they simply dry out and wait for new water to arrive. That also means that they cannot get very big. Still, a contemporary moss is not any more primitive than an oak tree, it has merely found a second adaptive strategy to the problem of living on dry land. Among all the plants trying the same adaptive strategy, however, we could distinguish better or worse solutions.


  1. Hi Alex,

    This is "Paul" from whyevolutionistrue blog.

    I've been prohibited from posting further on there in this particular discussion, under threat of a ban (issued via private communication). The moderator there doesn't appreciate our back-and-forth dominating the post's comments section. I'd advise you against posting further on it, too, btw, or you too might become subject to such warnings or a "probation" period. (Or perhaps not)

    As such, I'd like to reply to your last post here instead, given that I've no other way to communicate with you ;-)

    If you look at a generally accepted definition of free will, namely (quoting wikipedia here) "free will is the ability of agents to make choices unconstrained by certain factors."

    The way I see it, this definition is uncontroversial, and accepted by supernaturalists, as well as naturalists/rationalists of different persuasions. The problems begin when the definition is "unpacked".

    Here, the incompatibilist and the compatibilist part ways, because they mean different things by the word agent. The compatibilist assigns what I still assert (I'll make a case for why this is so in a short while) is an arbitrary, but actual meaning to the word "agent", while the incompatibilist says the word is meaningless in principle, yet for the sake of his sanity uses words like "action", "choice", etc in a strictly metaphorical sense.

    As for your objection to my having calling the compatibilist "agency" an arbitrary abstraction, consider this:

    The compatibilist seems to only consider those entities, that are intuitively thought of as individuals, as "agents". The point here is that this intuition about individuals is also just a naive conception, born of our own human nature. What constitutes one human individual is a conglomeration of particles and their interactions. The very fact that billions of those particles aren't a part of my body anymore tomorrow and billions of new one will be, should underline the arbitrary nature of it. Furthermore, can you point to the exact boundary of this "individual"?

    Arguments along many other lines can be made for why human intuition is nearly always naive, with respect to this, but also with respect to other things: just consider Newton's theory of gravity. Perfectly fine to describe everyday human dealings with interactions between things with mass, but also clearly naive.

    Now you may be wondering what the relevance of this particular example of gravitation I just gave is... Well I chose it, because I want to circle back to my question about what you would delineate as boundary of the individual. You either find this question meaningful, in which case you feel you can delineate it, or you don't, in which case you must admit that your term "individual" has no coherent definition in a physical sense. Assuming the first position (if you are of the second opinion, my argument is already done here, since then I've shown the arbitrary nature of what you call "agent"), consider that all the physical that consitutes what you've now delineates as this individual is completely and utterly a slave to gravitational forces (as well as the three other forces, but we need only consider this one). This means the individual cannot be decoupled from everything that is not-this-individual, by the very fact of gravitation between the matter constituents of the two different entities (the individual and "everything else").

    In addition, I'll refer you to for some philosophers' point of view on compatibilism, most of which amount to saying what I did in my previous post, namely that they feel compatibilists water down free will so much that nothing is left of what is generally understand by it. It's a bit like philosophical homeopathy in that sense ;-)

  2. This is not the ideal post on this blog to continue this discussion, there are more on-topic ones.

    It should perhaps give you pause that even the incompatibilist needs those words. Perhaps that is because there is some empirical to be described with them after all: a sane human could, for example, learn not to kill any more people, a landslide cannot.

    We surely have non-arbitrary ways of deciding what is an agent and what isn't, and although they are clearly too trigger-happy there was tremendous selection pressure on our intuitions to make them at least a decent heuristic. Consider being in a forest, with a boulder to your left and an angry mother bear to your right. Which of these would have a 'will' relevant to you? Which of these would you consider more likely to, ahem, 'act'? If a boulder comes rolling towards you and an angry bear comes bounding at you, which of these might you be able to convince to reconsider their planned course of, ahem, 'action' if you suddenly had a dozen comrades appear next to you, shouting and waving spears?

    Sorry but no, there is nothing arbitrary about these considerations, there are demonstrable, reproducible, empirical differences.

    And the facts that there is no one atom boundary to humans, that gravity is a simplified model, or that I am not quite the same person as I was a month ago don't change that either. Of course science is a necessary simplification of reality but that is not the same as arbitrary. The difference between a toddler and a grown-up isn't arbitrary just because there is teen age between them. Would you let the toddler drive a car? I hope not.

    I know the criticism of compatibilism because it features so strongly on WEIT, but to me the rejection of that criticism boils down to two considerations: (1) Given that incompatibilists demonstrably also differentiate between these situations, it would be nice to have terms to describe the difference. (2) Although you may disagree, the incompatibilists are, in my eyes, demonstrably wrong thinking that the terms used to describe these differences have supernatural connotations. Although few people read this blog I should perhaps do a post expanding on the latter.

  3. I realize this post really has nothing to do with it, but I wasn't sure if you'd notice it if I posted this on a potentially really old, but more relevant post ;-)

    Also, just let me know if you'd rather end this discussion altogether! I find it interesting enough to continue, but if you don't, I can respect that of course.

    I think you're misinterpreting what I mean by "arbitrary", here. Of course, there is a difference between a landslide and a murderer, it's a plain as day and I don't deny that. The point is that the compatibilist pinpoints or fixes the individual ((usually) human) as the "sphere of things" in which talking about "will" "makes sense", which is an arbitrary choice (note that arbitrary doesn't mean random here), motivated only by some intuition of this human compatibilist doing so, an intuition that his humanness has some implicit significance.

    I like that you brought up toddlers and adults, since I can turn that argument around on you: at what point does a toddler stop being a toddler? When does an adolescent turn into an adult? The meanings of these concepts are also fairly arbitrary, obviously somewhat motivated by developmental milestones, but still, essentially (pardon the pun) arbitrary.

    As for your question as to why the incompatibilist needs "those words", I thought I adequately explained that earlier: to not have to think of everything in terms of the underlying particles and their interactions. The incompatibilist however recognizes that the concepts he abstracts the actual situation into are flawed. As for your boulder, forest, bear story, the incompatibilist would run a way from the bear not being he thinks the bear "wills" to eat him, but because he realizes it poses more of a threat than the boulder after having abstracted from their respective underlying physical natures.

    Also note here that, in the forest analogy, there is also the implicit attribution of "agency" for the entity presented with the dilemma, which is a real problem that further muddies the argument, since all of it is just on rails and the incompatibilist presented with the scenario, being a determinist, realizes not even his own behavior is anything but _on rails_. The classic mistake is then to think this will naturally lead him to fatalism or anarchistic behavior. At this point I want to circle back to my wondering about the origins of this watered-down notion of "free will", which I asked about in my last post on WEIT. Did it originate as a reflexive action motivated by this fear of a chaotic, lawless society emerging from the realization of the illusion of free will? I've heard Dennett speak about "free will" along these kind of lines a couple of times, and he is one of the most "well known" compatibilists today, unless I'm mistaken.

    Just before introducing the forest analogy, you used another word which can be used against you, namely the word intuition itself, another essentially arbitrary word used to refer to the following: something that guides some of your actions, but whose origin you can't really account for. What do you intuitively know about the potentional behavior of another human (or even yourself), and what do you consciously know about it? Can you draw a line? You can't, of course, leading me to note yet again how arbitrary the compatibilist notion of "agency" is. (And let's not go into the quagmire of realizing that human intuition about human intuition itself is also just a naive conception, lest our brains explode.)

    Text continues in next post.

  4. (Continued from previous post)

    I still find your objection to my programs analogy odd, by the way. If I mapped your forest analogy to a "computer world", what then? At what point does the concept of an actual definition of "agency", as opposed to simply a flawed abstraction, stop making sense to you? What if I run some computer simulation, wherein deterministic subprocesses, which accept inputs, interact? Would you still speak of subprocesses of potentially different degrees of "will" within it?

    If you argue that
    - one these processes, an instance of, say program A
    might be
    - provided with input by another program, B
    such that
    - its output would be in some way more "favorable" to program B (the precise way is inconsequential)
    and still
    - insist on using a term like "free will" for the "malleability" of the behavior (here: execution trace) of program A,
    can you then
    - admit you've sufficiently watered down what the word conventionally means to people,
    so as to
    - be (nearly?) completely devoid of the "essence" of its original meaning?

    Commenting quickly on the last paragraph, (1) yes, we do have use of the words. but only as inherently flawed abstractions, not as essential, meaningful terms. (2) I don't think a compatibilist has a supernatural meaning for the terms, per se, as I've tried to make clear. I just think their definitions are based in naive human intuitions, which limits their thinking such that it can't transcend this intuition.

    - Paul

  5. I really don't get where your problem is.

    You keep talking about intuition as if it were the primary or even only way we can think about agency. No, as I keep saying we have empirical approaches to decide. People are not being declared insane based on intuition, just as one obvious example.

    You also appear to be making the argument, wearily familiar from Jerry Coyne, that agency or free will immediately cease to be useful concepts if they come in grades. Again, free agent versus not, toddler versus adult, red versus yellow, shrub versus tree, short story versus novel. All of these have some kind of gradient between them but that does not mean that their distinction is arbitrary, that their distinction is an illusion, that their distinction is not empirically demonstrable, or that knowing which is which is not immensely useful.

    The expectation that we can only accept concepts as real that have an unquestionable, extremely clear separating line between them is weird from a scientific perspective because we scientists are (or at least should be) all constantly aware that our concepts are abstractions but still empirical, non-arbitrary and useful. (It is perhaps particularly weird coming from a scientist who accepts, of all things, the concept of biological species, but that only as an aside.) If you want everything to be clear-cut you will find every concept outside of pure mathematics to be arbitrary and useless.

    I'm getting the feeling that the modus operandi of incompatibilists is to unilaterally define the topic so that it is easily knocked over. They say we should only call something a choice if the choice has been made by an immaterial, supernatural soul. Sorry but no, it is a well accepted use of the term choice when applied to computers or lab rats, so why not to human meat computers? They say we should only accept free will if it is a simple yes/no attribute instead of something that comes in grades. Sorry but no, I am happy to accept it comes in grades, just like adulthood or redness, and that a snail has more of it than a rock.

    And once more, it is no help talking about simple computer programs here because none of them has agency. If we had a hard AI, on the other hand, then why should we not say that it makes choices, that it has agency and a will just like we do? There is nothing special about humans.

    since all of it is just on rails and the incompatibilist presented with the scenario, being a determinist, realizes not even his own behavior is anything but _on rails_.

    I thought you had acknowledged by now that the compatibilist is also a determinist and realizes the same.

  6. I said _nothing_ about the simplicity you suppose of the programs in this latest example, simply that they accepted input and produced output. However, the very fact that *you plainly stated that a "simple program" has _no agency_, even if it accepts input (even if minimal), shows me that you _implicitly_ reject (due to your naive intuition) that (perhaps very) minimal agency is a thing* worth giving that name, further making my point about the arbitrary nature of what you call agency for me. (I bet you didn't even realize you did this just now! And this fact is part of the essence of the problem with your blurring the distinction between comp. & incomp.)

    The programs are just black boxes, accepting input and producing output. Their internal behavior is abstracted from entirely, and can be as complex as it you want it to be. Now we finally seem to get to the heart of the matter.

    "If we had a hard AI, on the other hand, then why should we not say that it makes choices, that it has agency and a will just like we do?"

    Because that is clearly not what is generally understood by "choices", "agency" and "will". So it ultimately boils down to the compatibilist who seems to inhabit a linguistically separate universe than anyone who isn't a compatibilist, one in which the words used are - as I asserted earlier, many times - in some sense naive, and introduced as a supplanting of established meanings.

    Let me again stress that what you seem to be understanding my word "arbitrary" to mean is not I mean by it. I've tried to explain it, but you seem to be misunderstanding it, and I think I give up on trying further on this specific thing, since I've tried to point out my meaning of it, in terms on intuition, to no avail. I hope my earlier pointing out of your utter rejection of agency, as opposed to attribution of very minimal agency, of a simple I/O program will finally drive the point home.

    - P

    PS. (my "being a determinist" was just a qualification of the incompatibilist, not a disqualification of the compatibilist as one. In hindsight, I should have made that clearer)

  7. Oh come on, even a snail has more agency than the best chess computer we have come up with so far.

    Because that is clearly not what is generally understood by "choices", "agency" and "will".

    I disagree. As for the first, it is fairly clear that, say, autopilots or computer opponents in games make choices, let alone animals in behavioural experiments. We simply have no alternative expressions to describe what they do.

    As for the others, who is it now who arbitrarily declares that a computer cannot possibly have agency? You seem to be alternately claiming that I did that (to show that I rely only on my flawed intuition) and then turn around and say that everybody except me does that (to show that the terms have a supernatural definition). Please pick one.

    1. "who is it now who arbitrarily declares that a computer cannot possibly have agency? You seem to be alternately claiming that I did that (to show that I rely only on my flawed intuition) and then turn around and say that everybody except me does that"

      You did do that, do you deny it? I don't see how you could deny your denial of agency of that simple I/O program? Saying "oh come on" won't do to gloss over that, I'm afraid. And here the naivite of compatibilism is ultimately exposed.

      As for your confusion about "who it is now that...", again you seem to be playing a bit of a word game. A naturalistic connotation of the words can still be "not what is generally understood by it". But perhaps I should refrain from this particular line of "attack", since our language seems inadequate to properly address this particular issue. (Or at least, my command of the English language is.)

      I concede that compatibilism is fully naturalistic; the other point from the first paragraph of my reply still stands, however, making it philosophically naive and therefore not _really_ coherent or meaningful.


      (As for your subtle suggestion that chess computers are pinnacles of AI... I reject that too. AI researchers have all but completely stopped any work relating to chess; chess engines today operate basically on brute force computation. But that's an objection detached from the discussion entirely)

  8. I do think though, that we roughly make the same type of mental models in dealing with things, but we basically look at the models in perhaps a slightly different way. The exact way in which they differ is perhaps hard to explicitly denote. Perhaps compatibilism is more constructive to philosophers (more useful in thinking about thinking in a human context), and incompatibilism is more coherent (in realizing the arbitrary nature of the sphere of considerations about will).


  9. Yes, you have caught me! Although I have repeatedly stressed that we have empirical approaches for determining the degree of free agency that a person has, although I repeatedly wrote that computers can have varying degrees of agency, although I repeatedly wrote that animals and hypothetical aliens clearly have agency, my reaction to your program example totally demonstrated that I intuitively assign agency to humans and to nothing else because I think humans are special.

    Well, either that or you only read what you want to read. One of those.

  10. The first one ;-) It's OK. If you don't want to admit you do and did intuitively do that, even if not always *explicitly*, that is fine. It doesn't change the reality that it is true. -P

  11. The way I see it, you either

    (a) retract your earlier denial of agency of the simple I/O program, on the basis that it misrepresents your own view
    (b) keep contending that a program that accepts the minimal amount of input, namely 1 bit, has no agency (to speak of)

    if you pick (a) and still insist that a program accepting just 1 bit information of input is something conventially understood using a term like "free will", I cease all further discussion since that is not what it means to either supernaturalists or non-compatibilist naturalists
    if you pick (b) then clearly what is worth of the name "agency" for your view of compatibilism is informed by some arbitrary intuition, whether you deny it or not

    there is no in-between! -P

    1. the only way to really check the claim about the semantics issue would be to poll every person in the world, supernaturalists and naturalists alike, about whether they would say that 1-bit-input-accepting-program has what they call "free will". as I said, I don't think a whole lot of people share your viewpoint, but this seems an interesting thing to test in a scientific context. -P

  12. Look here, this is not rocket science. You see it as a binary issue, I don't.

    Let's say I argue that 'maturity' is something can be present to differing degrees, and that a newborn doesn't have any but a thirty year old, all else being equal, will generally have a lot of.

    If then you hold up a three week old baby and ask me if it has any maturity, me saying 'no' does not demonstrate that maturity is an illusion, it does not demonstrate that maturity is arbitrary, it does not demonstrate that there is no empirical approach to determining degrees of maturity, it does not disprove that a six year old has more maturity than a baby and a twelve year old has a bit more yet, and it does not demonstrate that I am inconsistent. It merely means that I think that the three week old baby has too little of it worth mentioning.

    Same for agency and free will and a very simple program that has neither desires nor any relevant options to act on them. Does that finally penetrate? I am losing patience here.

    And no, obviously not a lot of people would hold the belief that a 1-bit-input-processing program has free will because that is a stupid belief. The questions you would have to ask a representative subsample of people is whether they believe that something is only a choice if it has been made by a supernatural being or whether a computer, for example, can also make choices; whether they believe that 'I did that out of my own free will' implies that I have an immaterial soul or merely that I was able to do what I wanted; whether if they assume determinism to be true it would still make sense to talk of choices having been made and people doing things out of their own free will; stuff like that. While you are at it, maybe poll physicists and statisticians if 'degrees of freedom' in mechanics and stats implies something supernatural.

    And there you will find that contrary to incompatibilist assumptions none of these words mean anything supernatural, neither per definition nor as commonly used.

  13. You seem fundamentally unable to understand the point about the arbitrary nature of what I say you use the word agency for: I don't argue this point as a criticism of compatibilism, but as a reason the incompatibilist isn't really a closeted compatibilist, despite your insistence. The incompatibilist would not dismiss the "agency" of the 1-bit input program outright.

    In any case, I think you need to take a step back and reevaluate what we are actually disagreeing on. Not on whether incompatibilism or compatibilism makes more sense, but on your assertion that tried to blur the line between them by attributing "stances" incompatibilists don't *really* take to them. - P