[FRIAM] Why depth/thickness matters

[glen ☣]

On 02/14/2017 09:51 AM, Eric Charles wrote:
> Thanks for the reorientation! If you want to discuss complexity, I think an interesting question regarding perception-action systems is how much of the complexity has to be inside the organism, and how much of it can be encapsulated in the larger organism-environment system. The more the complexit is spread across the system, the more the organism can get by with much less "mental" complexity that it might originally seem. That tension is at the heart of Gibson vs. traditional theories, though, of course, Gibson described the tension in different terms.

Yes, and that's exactly what the Hoffman article is about, too, with their exploration of simpler or more complex environments.  Your criticism of their (rather common) concept that seeing more takes more energy also exists in the "fly ball" and locomotive examples.  And the well-kept or poorly-kept radio metaphor simply raises the spectre of "adaptation" and the target of selection pressures.

In other words, the boundary between the organism, the environment, and the organizational relationship between them is nowhere near as crisp as we assume.  It's that assumption that is the target of Hoffman's (anti-realism) project.

And that brings me back to my original point about loopiness.  We not only have the problem of distributing the logic beetween organism and environment.  We also have the problem of how to grade/categorize the spectrum _between_ the two.  E.g. to what extent is, say, a pair of eyeglasses a part of the organism?  E.g. to what extent is the eye's cornea part of the environment?

Computations over the organism strike me as one layer.  Computations over an objectively extant landscape are another layer, perhaps of similar complexity than those over the organism.  Computations over both are another layer.  Computations over a collection of organisms, with a purely co-constructed "environment", is another.  Computations over all 4 (each organism, extant environment, organism-extant-env couplings, multiple organisms in extant environment) is yet another layer.  Loops within loops.

> However, that doesn't necessarily speak to our ability to jettison "representation" and replace it with dynamic-systems accounts more generally.  
> [...]
> So, to recap: The questions for the list are 1) Where will we look for the complexity in question? In the organism, in the environment, or in the system that includes both? 2) Once we have a decent account of that complexity, is anything added by inserting representation-talk in the middle of it?

It's not clear to me why you focused on a juxtaposition of representation vs. dynamical systems.  It sounds a lot like Marcus' argument in the loopiness thread.  You seem to be arguing that we can "flatten" the system to a dynamical systems account, with some exogenous accuracy and precision or error.  (By "exogenous", I mean typical sources like however we measure it or purely mechanical noise caused by a kind of "simple" uncertainty ... things like how well a nut fits a bolt, etc.)

By arguing that some types of loops within loops are only amenable to lossy compression, I'm asserting that _some_ of the loss is due to non-isomorphic mappings across boundaries.  The interfaces between actors are somehow smaller than what's on either side of the interfaces.  (Hence my comments about the holographic principle.)  In that sense, the question isn't merely about _where_ the complexity is (organism, environment, both), but also to what extent that complexity would be invariant if it were a) moved or b) modeled by something on the other side of a (smaller, lossy) interface.

This raises questions like: to what extent do organisms model their environment or vice versa?  Or to what extent are co-constructed scientific theories validated?  How to falsify them?  Etc.

-- 
☣ glen