Chemero (2009) Chapter 7: Affordances, etc (Pt 2)

Last time I went over affordances-as-dispositions, and Chemero's first swing at affordances-as-relations. Affordances can't be dispositions, claims Chemero, because

1. Dispositions manifest when the conditions are met; this is compulsory. But I am not currently trying to effect all the affordances in my vicinity, so they can't be dispositions. Relations are functions, and thus support malfunctions.
2. Dispositions require complements - for perception-action, the complement of an affordance is an effectivity. But what exactly is this? Body scale (e.g. leg length)? Actually, it's more likely in terms of ability (per some unpublished experiments Chemero has run); people's judgements of stair climbability are a relation between the riser height and the person's ability to step that high.
3. If affordances are properties that are directly perceived, then when two people perceive the same affordance their minds will overlap: the problem of two minds. Relations solve this problem by making the overall relation which the directly perceived affordance is part of unique to each observer.

This would be all well and good, except that 

1. Affordances and effectivities are complex dispositions, and the conditions for being realised can be a long list. In addition, I can only be one kind of effecting device at a time, so when seated I am literally not capable of complementing the climbing affordances of my stairs at that moment in time.
2. Noting that 'body scale' is an imperfect proxy for an effectivity, and then claiming that this means nothing is an effectivity makes no sense. In addition, 'abilities' are equally approximate. The issue (being careful what you claim is the actual complement of the affordance) is valid but applies equally to dispositions or relations.
3. The solution to the problem of two minds that Heft outlined and Chemero thinks supports his case lies in making the act of perception relational, not the thing perceived. The affordance does not, itself, need to be a relation.

So far, nothing has convinced me that affordances need to be relational. But to round the story out, I want to finish the chapter and address the final tweak Chemero adds: Affordances 2.0.

There's actually not a lot more to Affordances 2.0. Because it assumes that affordances are relations between abilities to act and features (not properties) of the environment, there is room for both affordances and abilities to evolve and change over time;

Over developmental time, an animal’s sensorimotor abilities select its niche the animal will become selectively sensitive to information relevant to the things it is able to do. Also over developmental time, the niche will strongly influence the development of the animal’s ability to perceive and act. Over the shorter time scales of behavior, the animal’s sensorimotor abilities manifest themselves in embodied action that causes changes in the layout of available affordances, and these affordances will change the way abilities are exercised in action. The key point here is that affordances and abilities are not just defined in terms of one another as in the dispositional and relational views discussed above, but causally interact in real time and are causally dependent on one another.

Chemero, 2009, p. 150-151

This makes the whole thing time-extended and dynamic. 

Some of the problems Chemero's identified are genuine problems to which relations are the answer (e.g. the problem of two minds; the dynamics within developmental timescales of our fit with the environment). The error I believe he's made, though, is to ignore the fact that perception-action already has a relational component that does this work - the act of perception itself. Need the potential for malfunction? How about the ready made potential for perception to fail (I look in the wrong place, attend to the wrong thing, have yet to learn to discriminate the relevant information, etc). Worried about keeping the contents of your mind private? Easy; just take advantage of the fact that your unique viewpoint on the affordance means you are in a unique relation to it. Want to allow room for affordances and your abilities to causally interact over time? Move around and gain a time-varying stream of information about the world and use this information to coordinate a control your next actions (by forming the supported task specific device). 

I like that Chemero is trying to firm up the connections between ecological psychology and developmental systems biology; I think this is exactly the right place to go looking for a home. But dispositions don't prevent this, and the Turvey-Shaw-Mace focus on ecological physics was mostly, I think, a result of the laws paper being about how to project specifying information into energy arrays - a side effect of the topic at hand, nothing more. Getting dynamical and trying to get process thinking into cognitive science is also an excellent idea; but the natural home for that is in information, not what the information is about. All the tools Chemero wants already exist, and the motivation for an account which I think eventually falls flat anyway simply isn't there.

Dynamic Touch

One final note; Chemero ends by tackling Gibson's central question about affordances; not whether they exist, but whether there is information for them. To complement his discussion of visual perception from Chapter 6 (τ and entropy) he reviews some of the work on dynamic touch. This is a good section on the one hand (as he notes, dynamic touch doesn't get a lot of attention outside the ecological literature and it's a productive programme), but there are a couple of issues I wanted to note in passing.

Dynamic touch is the name given to the process that allows you to have non-visual information about objects. Pick up a book, and close your eyes; now heft the book, or move it around. You will have a very clear (and generally pretty accurate) perception of it's size, etc. The basis for this perception is the inertia tensor. All objects with mass have inertia - they resist a change in their position. This resistance can vary in three spatial dimensions, and the inertia tensor is a matrix describing this three dimensional resistance. People's haptic perception of, say, an object's size, varies as a function of this tensor. If you experimentally alter an object's inertia tensor by altering the mass distribution, people's perception follows the tensor and not the size. The tensor, in other words, is the dynamic entity that is being specified and thus perceived.

Two brief comments - first, the tensor is not information. It is a 'world' variable, because it is dynamic (it's units include mass). Information is kinematic (motion based; time, position and the temporal derivatives of position only). The tensor is, as I say, quite clearly the world variable we perceive; but the information specifying this world variable is yet to be identified. So, strictly speaking, dynamic touch is a poor example of an answer to Gibson's central question.

Second, Chemero discusses the proposed role of the inertia tensor in the size-weight illusion. The suggestion is that people are not perceiving size, or weight, but moveability, and there is evidence to support this. I just want to flag up that recent work by my colleagues Geoff Bingham and Qin Zhu suggests the size-weight illusion is actually specifically about throwability, and that when they tested the inertia tensor explanation for these data the tensor account did not work. However, this last bit is new and as yet unpublished data, so I just want to flag up that while there is a good literature on the illusion in the dynamic touch literature, the story isn't over yet.

Conclusions

This ends the psychological part of the book. I'll get to the last chapters soon, but not immediately due to work constraints. But I wanted to get to this stage, to have all the pieces available to discuss.

Broadly speaking, I don't buy Chemero's 'shored up' ecological psychology. I don't buy the motivation, I don't buy the solutions (situation semantics, relations), and I think that the law based dispositions account stands up just fine to the critique. I will concede that the issue of using non-specifying variables has me interested; to preview my thinking, I agree that an organism cannot tell the difference between a lawfully generated variable and a sufficiently robust constraint based variable. I am not yet convinced, though, that the latter can stand up to the kind of scrutiny that it will be exposed to over the course of perceptual development and learning - but I need to invest a little time in this, and to finally get round to reading that literature. I think that the laws account will still hold up well, and I hope it does because the constraint account currently lacks the critical mechanism of the symmetry principle to allow specification to work; but there's work still to be done on the latter. I think.

Comments, please! I'm enjoying the discussion and there are still question to which I'd like answers.