Chapman, Agre, and PCT

[From Bill Powers (931214.0920 MST)]

RE: Chapman, Agre, and PCT

The following is from Chapman and Agre, Abstract Reasoning as
Emergent from Concrete Activity. MIT research report; Proceedings
of 1988 workshop, Timberline, Oregon: Reasoning about Actions and
Plans. Here are some pertinent exerpts.

"Routine activity is _situated_; it makes extensive use of the
immediate surrounds and their accessibility for observation and
interaction.. Rather than using internal datastructures to model
the world, the world is immediately accessed through perception.
Much of the theory [Agre's] is concerned with the emergence of
_routines_,which are dynamics of interaction: patterns of
activity that occur not as a result of their representation in
the head of the agent, but due to a Simon's-ant relationship to a
complex environment...." [whatever that is]

"We think that _indexical-functional_ representations are used
... These are fully indexical representations which can be
evaluated relative to the current situation very cheaply by
virtue of their grounding in sensorimotor primitives. They can be
thought of as definite noun phrases: an indexical-functional
representation picks out, for example, _the cereal box_ or _the
tea strainer_ from the visual field. The indexicality of such
representations is mostly a matter of egocentricity. _The cereal
box_ implicitly means the cereal box in front of _me_....

"For indexical-functional representations, things that look the
same _are_ the same. All you can do is see if there is a bowl
there or not. There is no way to represent permanent objects in
Piaget's sense."


The language here is so ambiguous and abstract that it's hard to
make out what the authors are really talking about. When they
speak of "immediate" access to the world through perception, do
they mean "unmediated" or "present-time?" Are they saying that
perception simply picks up what is there to be perceived? They
contrast immediate perception to "using internal datastructures
to model the world," which seems to imply that immediate
perception is not dependent on internal operations --
unfortunately, they use the word "datastructures," which omits
mention of the process required to create a datastructure, and
implies some sort of memory process rather than ongoing
interpretation of input data. In the terms of Martin Taylor's
post this morning, they don't distinguish between perceptual
functions and perceptual signals.

I don't know what they mean by "sensorimotor primitives." I could
perhaps understand either sensory primitives (intensity signals)
or motor primitives (muscle contractions), but using a term that
sort of smushes them together ends up saying nothing about

I also don't know what "indexical" means, and thus what
"indexical-functional" representation means, either. Is
indexicality that sort of "aboutness" that Searle and others use
as an explanation of intentionality, as in "indicating"? Does it
mean that a representation indicates something in the external
world? When you add "functional", does that imply that the
representation is some sort of operation on something? Or that it
is used to accomplish some purpose? Or that it relates the values
of one or more arguments to the value of a function? When you try
to attach meanings to these terms, they dissolve into mush.

The "egocentricity" of which they speak is present in a way of
which the authors are apparently not aware. They see a
representation of "the cereal box" as egocentric because it
assumes that it means "the cereal box in front of _me_". But the
real egocentricity is in the label "the cereal box," not in its
location relative to the observer. The assumption is that anybody
can see that the object in the environment is a cereal box, a
naive and culture-centered assumption that ignores the role of
perceptual functions in defining what is seen to be "in front of
me." In all the modeling efforts by Agre and Chapman, this same
naive assumption is evident: the world is transformed into
definite-noun phrases, as if the only perceptual problem is what
to call the items that are found in the world. The role of
perceptual functions in _creating_ those items is never

Note, in the opening line above, the assertion that indexical-
functional representations "are fully indexical representations
which can be evaluated relative to the current situation very
cheaply...". The cheapness is a minor consideration in comparison
with the glossed-over assumption that a representation can be
evaluated relative to the current situation. That evaluation
requires knowlege of both the situation and the representation of
it, implying a mode of access to the world that bypasses

That sentence deserves further study. It's constructed to build
up momentum enough to carry the reader past this fundamental
question about representation to the issue of the cost of
representation, cutting off any discussion of the major issue and
forcing attention onto the minor one. This technique is very
common in discussions of this sort.

One indication that the authors aren't even aware of the real
perceptual problem (and its proposed solution by several people)
is this: "_Constructivism_ is the position, held by many
developmental psychologists, that human development occurs
roughly in stages, and that each stage is built upon the previous
one." With this understanding of the term, how could they
possibly understand a constructivist position on perception, the
idea that perceptual representations are constructed by the
brain? I have actually never heard the term used as the authors
define it; one wonders if they didn't just guess at a meaning, a
typical bad habit of verbally adept people. They obviously didn't
get that meaning from von Glasersfeld, or me.

The greatest difficulty in evaluating writing like that of
Chapman and Agre is that they use terms in such sloppy ways,
defining them primarily by italicizing them, that one can read
almost anything into them. If you're convinced that they have
shared the insights of PCT, you can always pick one of the
possible meanings that is congruent with PCT, because they say
nothing to rule it out. On the other hand, they say nothing to
rule it IN, either, so we must fall back on trying to see what
they mean by analyzing context. If one had to evaluate the
authors' ideas strictly on the basis of the clarity and precision
of the writing, one would have to conclude that they don't know
what they're talking about. If that's the case, how can they
expect anyone else to understand them?

Bill P.