A modeling question

From Kent McClelland (930630)

Reply to Tom Bourbon (930623.1306)

Thanks, Tom, for your thought-provoking answer to my earlier post (Kent
McClelland [930622]) on the possibility of "Collective Controlled Variables."
I've taken a while to get back to you on this, but I've needed some time to
consider the points you made.

In response to my argument that some "alignment" of reference levels between
independent control systems is necessary for cooperative actions to occur,
you first point out that cooperation often requires individuals to adopt
different reference levels, not the same ones:

What matters is that the participants adopt reference
signals (that they adopt goals which we model as reference signals -- Hans?)
that result in each acting in a way that produces a match between personal
goals and present perceptions. The same constraints apply to my two
hemispheres when "they cooperate" to produce one-person performance of what
can also be a two-person cooperative task: they need not adopt similar
reference signals; all that is necessary is that each adopt reference
perceptions that result in the intended perceptions. . . .
Cooperation can, probably always
does, entail some necessarily different reference perceptions in the

In sociological terms, we're talking about "division of labor," and your
point is well taken. I'm a little unclear, though, about what you mean by
"adopting goals" and how that's different from my "practical alignment" of
reference levels, if the goals are necessarily shared in order for
cooperation to take place. You go on to suggest that even when people
perceive themselves to be sharing the same goals, they probably aren't, at
least not in any detail:

. . . that kind of agreement says nothing about the
specific contents of the participants' heads. The socially affected and
approved options all reside in individual heads, as products of individual
interpretations or understandings. Necessary and important, to be sure; the
same in each person, probably not.

A good many sociologists would heartily agree with you on this point,
especially those who call themselves "ethnomethodologists." This band of
maverick sociologists have made their careers by studying the methods people
use to convince themselves that their perceptions of reality match everybody
else's perceptions, in other words, that "social reality" is really real.
The "ethnomethods" described by ethnomethodologists are techniques by which
people succeed in glossing over ambiguities and putting off any discussion of
apparent conflicts. In various ways, such folk techniques make it possible
for "normal" social interaction to take place on the mistaken assumption that
everybody understands each other and sees the world in the same way. The
ethnomethodologists themselves apparently believe that the semblance of
social reality achieved by these techniques is merely a shared illusion and
that, were it not for the use of ethnomethods, social life would be even more
chaotic than it already is.

As I've listened at various times to discussions of epistemology on the net,
I've thought that the most ardent PCTers might enjoy reading a little
ethnomethodology. And I recall now that Rick Marken described to me how he
once studied under one of the founding fathers of ethnomethodology (Schegloff,
was it, or Sacks?), which might explain a lot about Rick! These founding
fathers (especially Harold Garfinkel) are mostly unreadable, but a recent
book on ethnomethodology I'd recommend is MUNDANE REASON by Melvin Pollner
(Cambridge University Press, 1987).

Anyhow, my sociological conclusion about what you're saying is that you are
undoubtedly right in emphasizing that each individual's perceptions are
different, but I still think that at some abstract level people must have
their control systems at least crudely aligned in order for cooperation to
take place and that any such alignment has got to be an important
sociological issue.

In the last part of your post, you dismissed my suggestion that independent
control systems when aligned can become "super powerful", if there are enough
of them. You described in tantalizingly brief terms some cooperation
experiments that you and Michelle Duggins will be presenting at the CSG
conference. In your words. . .

In none of the cases I have modeled did a quartet take on super powers. . .

I'm curious about what exactly this means, because of some spreadsheet
simulations I've been playing with. I find that when I couple two 50-gain
control systems to the same "external variable" and I give them identical
reference levels, their joint performance (whether measured as their
perceptual variables or their summed output) is indistinguishable from the
performance of a single 100-gain system given the same reference level and
the same pattern of disturbances. The correlation between either the input
or the summed output of the two coupled systems and the input or the output
of the single "double-power" system is 1.000. Actually, any combination of
two or three systems seems to work the same way, as long as the gains add up
to 100 (or whatever the gain of the single comparison system is set to be),
and as long as all the slowing factors are the same.

I take from these simulations (which would no doubt be easier to do in Simcon
if I weren't working on a Mac), that "perfect" alignment of reference levels
has the practical effect of adding up the gains of the participating systems.
I find it easy (maybe too easy) to make the jump from this highly abstract
simulation to a human social world where alignment is always imperfect but
nevertheless there seems to be strength in numbers. Does this make any sense
to you?

Till later,


PS Tom, I hope you'll soon be inspired to share your thoughts on

PPS It's raining again in Iowa tonight. Pretty soon the whole state will
have washed down the Mississippi, and all of us Iowans will be moving to the
Louisiana delta country whether we like it or not!

Kent McClelland Office: 515-269-3134
Assoc. Prof. of Sociology Home: 515-236-7002
Grinnell College Bitnet: mcclel@grin1
Grinnell, IA 50112-0810 Internet: mcclel@ac.grin.edu