Tuesday 19 April 2011

Chemero (2009) - A Brief Pause to See Where I'm At

There are still a couple of chapters left for me to go over in Chemero, but the last two were the crux of the book for me, and I want to try and summarise where my analysis has left me before I go on holiday for Easter.

First, I'd like to separate out a couple of themes. First, there's the overall 'radical embodied cognitive science' programme, and second is the specific form of the 'shored up ecological psychology' that Chemero advocates as a basis for this programme. I am entirely on board with the basic idea of RECS; specifically, I think that cognition, whatever that is, is non-representational and that we can make great progress by making our science non-representational. I think Chemero has written a clear exposition of what such a radical science might look like, and why we might want to bother, and I think this makes RECS a must read for the field. 

Second, I think that Chemero is right to go to ecological psychology for a theoretical basis. The more I look, the more I have come to believe that James Gibson is about the only psychologist to have actually proposed a genuine scientific theory in psychology. A true theory provides you with tools to empirically attack novel problems in your domain, and provides you with a clear basis to interpret the results of your tests and to begin to tell a coherent story. Psychology has been chasing phenomena for most of it's scientific life, with no clear framework emerging to tell us a story about why things are the way they are. Ecological psychology is a genuine theory, and it's about a critical feature of our psychological lives: how we maintain contact with our world and move through it successfully. Regardless of what topic you're specifically interested in, you need to understand how we come to have knowledge about our environments.

Finally, though, I have problems with the proposals Chemero makes about the two pillars of an ecological psychology, affordances and information. It's these problems I want to try and sum up here, to focus the conversation a little.

Chemero wants affordances to be relations; not properties of the world, but relations between features of the world and abilities of an organism. He motivates this with several specific points (e.g. the problem of two minds) but as I described last post, none of these work to motivate the account he wants.

But this is just the starting point, and Chemero has commented that the differences between the net results of the relational and dispositional accounts aren't that great. What he really wants to do in the book is to define affordances dynamically, as time varying features of a system which spans world and organism. What affordances are currently available would then depend on the recent activity of the organism, as well as it's developmental history and the available resources in the immediate surroundings. 

I think this is fine; the general point is well taken. But the capacity to do all this is already present in the dispositional, laws account - the time varying, relational aspect is the act of perception itself. This is part of my main concern with Chemero's analysis; it repeatedly sounds like he's confusing the world with information about the world. Information, our access to information, our ability to use information to guide action; these things are already time-extended and dynamic, so it's not clear to me that affordances have to do the work here.

I also actively prefer the dispositional formulation because dispositional properties, grounded via anchoring properties, capture everything about Gibson's description of affordances, including being the kind of thing which can interact with energy arrays to create information. These properties can then be projected into light, or sound, and can be measured by a perceptual system calibrated to it's own capabilities (perhaps body scale). This measurement by a calibrated device is all that's required to ensure that perception is direct, and of affordances.

The Turvey-Shaw-Mace approach grounds information in ecological law. Affordances (dispositional properties of the world) interact with energy arrays over time and produce structures in those arrays which are invariant over the transformations which come with a change in perspective, etc. Those structures are the lawful consequence of the way the properties interact with the arrays; any organism which detects this structure is, in effect, detecting the property, and the ecological hypothesis is effectively that we detect these structures and thus perceive the world. This is underwritten by the symmetry principle; the laws account therefore provides a mechanism by which the structure contains information.

Chemero feels that this account is too restrictive, and proposes that organisms can happily use sufficiently regular features of the arrays, which are the result of constraints in the situation. But this approach loses the symmetry principle and Chemero does not replace it with anything. This is a problem.

Chemero has a good point lurking here. As Geoff Bingham says, there is no 'peeking behind the curtain' in perception. All an organism has is the information which is presented to them, and we have to come to know what this information means via perceptual learning. If I present an organism with a structure in light which I have generated via a lawful process, or by a merely constrained process, and these structures are the same, then the organism has no basis on which to tell the difference and could readily end up using the non-law based variable. This is effectively what Chemero would like to allow into ecological psychology - the possibility (or rather, the likelihood) that a busy and needy organism will happily latch onto anything that's useful.

At one level, I agree with this. There is no peeking behind the curtain, and if a merely constrained situation was capable of producing a sufficiently stable information variable, then an organism might indeed learn to use it. However, while I agree this is possible, I am not at all convinced that this is all that likely. It's hard to make genuinely invariant features of an optic array. We have so much scope to move and reveal any problems that stability must be underpinned by something very robust. Can a mere constraint produce such an unbreakable structure in a flowing perceptual array? I don't think so, and I have yet to hear an example - the two in the book chapter (tau and entropy) are law based. A mere constraint is simply unlikely to be good enough to produce anything which will survive typical exploratory behaviour during learning.

Essentially, Chemero's critique of law based information and dispositional affordances is missing too many of the pieces which already do the work he thinks is needed. Dispositions have anchoring properties which can interact with energy arrays to create spatial-temporal structures in those arrays. Organisms will learn to differentiate those structures from the flowing arrays only when they are sufficiently stable over space and time to allow the process to run; they will be stable enough only when they are generated by a robust, lawful process. This lawful process allows the invariant to be informative about the world via the symmetry principle; the invariant is directly  informative about action relevant properties when measured by a suitably calibrated perceptual system.

Ecological psychology is the right theoretical framework for a radical embodied cognitive science. It's strength is that it is a true scientific theory, and the dispositional/laws formulation provides a clear research programme and methodology. This means that we owe it a rigorous empirical trial before we give it up because we can't think how to cope with one or two results, and frankly we have a lot of hard work ahead of us before we've done due diligence.

A Side Note
I don't actually object to people raising these kinds of questions; I think it's healthy to pay attention, so this isn't directed at anyone in particular. But more than anything I want to see psychology make something that looks like progress in my lifetime, and I'm tired of the endless goddamn fiddling we do with the basics of our field, because it's part of the reason we never get anywhere. I've come to the conclusion recently that Gibson and Turvey went to a lot of trouble to provide us with a powerful research programme, and it's time the field stopped fucking around and knuckled down to a little hard work and normal science based in this, without trying to stamp our own little tweaks on it. Admittedly I've come to this conclusion by arguing with Ken Aizawa and reading Chemero carefully, so it's not like there isn't a good place for this work. But there's a reason I stuck with science over philosophy, and it's because, at the end of the day, I really just want to get on with it. I have the tools I need, and having banged them around pretty hard recently I feel pretty good about them, so Game. On.


  1. Re your "A Side Note":
    I think Tony and I are in complete agreement with you. This was one of the first things we talked about when I visited his class a couple of weeks ago. When paradigms are functioning properly, most people don't think about them much - most people just do research.

    I'm not sure why psychological science can't find a balanced relationship with philosophy of psychology, as physics research or biology research have found with philosophy of physics or biology. Either the philosophy is totally ignored and shunned (compare top university psych departments vs. bio or physics) or the philosophical issues completely dominate and little research happens.

    One thing I liked about Tony's book is that the examples came from real research programs, reinforcing the idea that the point is for this stuff to be useful. Even his discussion about the "reality" of affordances can be summarized as: If you can do good research with them, then we might as well just call them real and get on with it.

  2. I agree with all that, basically, and I'm sure you guys are on board with the idea. It wasn't a swing at anyone, promise :)

    I think psychology has problems with philosophy of science because we don't actually have a central theory. Sabrina and I started this blog based on discussions we were having about our frustrations about this issue, and the more I think about it the clearer this fact seems to be.

    Then the fact that Gibson gets so little play in psychology makes me cranky, because it really is one hell of a theory. It could 100% wrong and still be the best theory psychology has had, possibly ever. People nick the ideas, talk about them wrong, then dismiss Gibson because their version makes no sense. It's like the psychologist's meta-fallacy :)

    Even though I don't buy Tony's shored up eco-psych, at least I know where I stand now having engaged with his book and I trust that stance. That's a win.

    Clearly, it's time for a holiday :) I'm off as of tomorrow for a week, so if I don't reply, that's why.

  3. Yes, I am on board.

    In fact, one of the main reasons I wrote the book is to let philosophers of science and (ugh) mind know that there is high-quality Normal Science going on that rejects all of their assumptions. The second main reason, and the reason for the "shoring up" (a phrase from the publicist that I'm not sure I actually used in the book) is that I wanted to do away with all the reasons said philosophers of science and mind would reject this Normal Science as being conceptually confused in one way or the other.

    So, I think Andrew's arguments with Ken Aizawa should be different from his arguments with me because what I'm really trying to do is save Andrew from having arguments with Ken! As Andrew hinted a while back in an exchange with Ken, the criticisms Ken has made of Turvey et al and Gibson over on his blog don't apply in the same way to the slightly altered version of ecological psych that I (and Rob Withagen) have been pushing.

    I realize that it's making all my eco psych friends mad at me, but I really am trying to help.

  4. I don't think Ken's arguments apply anyway, but that's another story :)

  5. Tony, before I forget: as much as I've been banging on about the affordances and information stuff, I do think you've achieved your goal as you described here - laying out that you can do good cognitive science without representations and here's what it might look like, etc.

    Speaking of doing good science: Geoff and Mark Mon-Williams have a new paper on the affordance structure shaping the spatial features of prehension: it's another Bingham tour de force on the methodology of studying affordances:

    Mon-Williams M, & Bingham GP (2011) Discovering affordances that determine the spatial structure of reach-to-grasp movements. Experimental Brain Research, 211(1), 145-160 Link

    I'll blog this in some detail when I have time, but it's a damn good read.

  6. Mon-Williams & Bingham (2011): TLDR

  7. Then you lose. Turns out doing this properly is hard.

  8. good thing i have you around to summarize the contents of those 13 figures for me!

  9. Have you ever seen Geoff give a talk? The number of figures that man can fit on a slide and still have the whole think make sense is epic.