A blog about problems in the field of psychology and attempts to fix them.

Monday, May 21, 2012

Beyond the Brain: Difficult Metaphors

Continuing coverage of Beyond the Brain, by Loise Barrett. The first part of the extended blog review is here, there is a succinct published review in PsychCritiques.
Chris asked me a question over at Manchester Psychiatry and I realized three important things: 1) I have written a lot about Beyond the Brain, 7 posts. 2) I still have a few more posts to write. 3) I better do that soon. 

This post is the most "asking for help" of the things I have left to write. It is about metaphor. I love a good metaphor, and most of the metaphors in Barrett's book were good. However, there are two pairs of metaphors that just aren't working for me, and they are somewhat common metaphors in the complexity/dynamics systems literature. In particular, I found the ‘short leash’ vs. ‘long leash’ analogy weird, similarly the ‘loose coupling’ and ‘soft assembly’ analogies have never quite gelled for me. I would appreciate some help figuring out why these labels make sense.

 Leash Length
‘Short leash’ refers to situations in which a central control mechanism has a high degree of control over all the structure in a set of behaviors, as when an owner (the central control mechanism) has tight control over a dog on a short leash (the dog being the body doing the behavior). ‘Long leash’ refers to situations in which the central control mechanism does little work, because the structure of the organism-environment interaction takes care of a lot of things you might otherwise think requires control, as when an owner of a very well controlled dog still gets the desired behavior even with a very long leash. Barrett invokes this metaphor to help readers understand how a properly built body can engage in sophisticated movements without each little detail being controlled by the brain. Good constrained-dynamic-systems stuff, which I agree with.
But while I like the message, the metaphor doesn’t quite work, for me. I find it confusing, I think, because there are two candidate ‘central controllers’ in the dog walking situation, the human’s brain (or the human more generally) and the dog’s brain. Thus I'm not sure what to think when Barrett uses mars-exploring robots as an example of the short leash vs. long leash distinction. Of particular interest is a new breed of distributed-control robots which are being developed, which spread out and explore without continuous human input... you know, cause any signal given by a person on Earth takes several minutes to reach Mars, so you standard video game style controls don't do much good. Instead of building one big robot that can handle anything (you hope), the new approach is to have lots of robots that will each "know" a little something about their local surroundings, and so can overcome challenges as a group, such as finding good paths around obstacles or implementing an idealized random-search pattern for resources.
However, because I don't know whether the human brain or the dog brain is the central control mechanism, I'm not sure what to think about the robots. The well healed dog's behavior is based, almost completely, upon a modified dogs brain. The mars robots could just as easily be controlled by a relatively centralized control system, which would render the 'long leash' metaphor odd. If, on the other hand, there was no centralized control mechanism on mars, and the robots acted in a very autonomous-but-coordinated fashion, accomplishing goals that no central mechanism is regulating, then I'm not sure that there is any leash. Thus an example of a robot that can explore mars correctly without direct human control doesn’t necessarily mean that there isn’t a strong central-control element, it only means there is not a ‘short leash’ connecting it to earth. This metaphor isn't too bad, it just gets muddled, and there should be a better metaphor we could use.

Coupling Strength
‘Loose coupling’ and ‘soft assembly’ cannot be criticized in the same way, because those terms suffer primarily from not drawing a metaphor to a clear and concrete situation. At the least, I can't really tell what situation they are trying to invoke. The gist of these terms can be easily gathered in context, but the book would have been stronger if better metaphors could have been created. So far as I can tell, the idea is to contrast something assembled like a die - with the relationships between the numbers on the side rigidly determined - versus something like a Rubic's Cube - where many patterns can be quickly made by rearranging the parts. Of course, that example is not very good. The skin of an octopus might be a good example of a soft assembly system, it can become spiky or smooth, change color, and variously serve countless functions by becoming a series of different tools, while still being one thing. 
Why loose or soft vs. hard? I can't say. I intuitively know what it means for two things to be tightly coupled, but when I use that to think about two things being loosely coupled, I just end up with the conclusion that they are connected, but will fall apart easily. Is the joint between two train cars a tight or a loose coupling? Well, it usually seems tight, but a big enough jostle and they fall apart, and the swivel of a lever and they come apart. If the lever is half-way, then the first car still pulls the second car, but things at the connection point are loose. Similarly, when I think of a softly assembled structure, I think of the block towers my children build and knock down... because they are not assembled more strongly. Or maybe I think of a structure made out of toothpicks, with marshmallows as the joints. I don't think either of these ideas are anywhere near what the author is trying to invoke, but it is where the metaphor seems most naturally to lead. 
I know these terms comes from computer science and dynamic systems work, and I know that sometimes you have to use what others give you, but the labels are so unintuitive that something better must be possible. Any help?


  1. Soft assembly isn't about how well the assembled device is held together. It's about how difficult it is to disassemble the device and reuse the pieces. My hand is 'hard assembled' in that I can't make it into a leg, but it can become temporarily coupled to other resources to form a grasping device, or a punching device, and I can stop grasping and start punching. Grasping devices can be perfectly stable while operating, but that stability must be actively maintained; remove the control and the assembly will fall apart, while my hand requires no active control to remain hand shaped.

  2. Yes, yes... and thank you for the better example... but "soft assembly" is a terrible term for that! It is not about being "soft", and it is only loosely about being "assembled" in a certain way.

    After all, my grasping-arm and my punching-arm are assembled in almost exactly the same way (especially if I am grasping something small, or grasping something used for punching ;- ). Further, people make arms into legs all the time, and evolution has certainly turned hands into a variety of things (e.g., wings and fins), if not legs per se. Developmentally speaking, tons of resources have been used to maintain your hand's hand-shape.

    At any rate, to talk about the tension maintained in our arm muscles as part of my arm "assembly" is just odd. If one were building a robotic person with hydraulics for muscles, one would talk about assembling the robot first, then, as a second step, adjusting the pressure of the hydraulics. If the robot had a central control mechanism that adjusted the pressure in the hydraulics to perform different tasks, most people would not think of the control mechanism as adjusting the way in which the robot was assembled.

    So I get the idea, and I like the idea, but can't we get a better metaphor for explaining it? I think we are talking about two different ways to maintain functional flexibility, two extremes along a dimension. One way is to be a multi-tool, like a Leatherman or a Swiss Army Knife. The other way is to have your substrate itself reform to create different tools as needed. Ideal examples would be the octopus skin, or comic book examples such as Spawn or Venom. If that is what we are talking about, then a human arm which changes tool-type primarily through the alteration of muscle tension is somewhere in the middle.

  3. Soft assembly is pretty accurate, when you remember you're assembling task specific devices, not hands or legs. These are functional assemblages of task resources, and that assembly is soft in that it's reversible over short timescales.

  4. Alright, so a few days after posting this, I got to struggling with the part of our joint chapter in which you start invoking the 'soft assembly' terminology. I think the terms 'flexible' and 'calibrate' get at everything you are trying to say in a much more clear way. I just don't like the implication that my torso is 'reassembled' when it transitions from a walking device to a running device.

    Couldn't we just say that my muscle tension is recalibrate? And that this is the functional flexibility created by a soft-tissue + neurons system?

  5. Your torso isn't reassembled, but the systems it participates in are. Soft assembly is about systems, not elements. Flexible is too vague and non-specific, while calibration has a specific meaning that isn't the same as soft-assembly. Calibration is part of the assembly process; you have to tune the device. But it's not identical and it's important to keep those separate.

  6. But what does it mean to say that "the systems your torso participates in are reassembled"?

    I don't mean to be obstinate, I just can't make the words work for me. Can you give me any other context in which the word 'assemble' is used in this manner, so I can try to make the connection?

  7. Lego is a great example of soft assembly. You have identifiable elements that do nothing much on their own but can be temporarily assembled into anything within the range of their properties. That thing is stable until the pieces are required for something else, at which point you disassemble the first thing and use the pieces to build the second thing.

  8. Is "assembly", in soft assembly, being used as a noun or a verb? That is, does "assembly" refer to the set of pieces or to the act of putting them together?

    I'm not sure I like it better either way, it just occurred to me that the word had both possibilities.

  9. Softness refers to the assembly, and means that it is reversible.

  10. I am still unsure if "assembly" in that context is a noun (i.e., the pieces) or a verb (i.e., the act of putting the pieces together), but in any case, why not just say "reversible assembly". I don't know of any other context in which the word "soft" is used as a synonym for "reversible".

    This whole expression is odd. What my body does when it changes from a slow-moving machine to a fast-moving machine (say, from walking to running) is not at all like what happens when I change a Lego castle into a Lego chair. Now, there is a sense in which my body becomes part of a larger system, and that process could almost (but not quite) be captured by the word "assembled". If that is what we are saying, then my body is analogous to a single Lego piece. Is that what we are trying to get at?!?

    I will be meeting in a few days with my old grad school adviser who does a lot of dynamic systems stuff (Jeff Schank). Maybe he can explain it to me.

  11. Try this,from here:

    "Soft assembly features
    These are those assembly features where the
    two mating form features connect together
    without any physical connector between them.
    For example, in plane-plane assembly feature,
    the two plates connect each other without the
    use of any physical connector."

    Softness means the coupling is informational, which is exactly what we're talking about. Dynamics makes a big distinction between causes that are physical (eg me pushing a button) vs informational (me telling you to go push the button). Kugler had some good stuff on this in the 80s or 90s.

    I really think you're overly worried about this. Softness is indeed a technical term from dynamics and it has a fairly specific meaning according to this. That meaning is precisely what we want.

  12. I think that thanks for the valuabe information and insights you have so provided here.
    best cognitive enhancers