[From Kent McClelland (2013.4.22.13:25 CDT)]
Subscribers to CSGnet may perhaps be interested in a draft article that I've recently written on models of conflict escalation. The paper is to be presented at a sociological convention this summer, and it's aimed at an audience of sociologists, but it uses PCT models and conflict is a topic of some interest in a lot of disciplines.
Best to all,
Kent
Cycles of Conflict 03-26-13.pdf (618 KB)
[From Bill Powers (2013.04.26.1155 MDT)]
Kent McClelland (2013.4.22.13:25
CDT) –Subscribers to CSGnet may perhaps be interested in a draft article that
I’ve recently written on models of conflict escalation. The paper is to
be presented at a sociological convention this summer, and it’s aimed at
an audience of sociologists, but it uses PCT models and conflict is a
topic of some interest in a lot of disciplines.
Finally getting around to reading this **MAGNIFICENT
PAPER.**This is the kind of relevant, substantive, grown-up paper I
have always wanted to write, but have never come close to achieving. You
treat Collins’ paper with complete collegial support and respect and
offer your own ideas with modesty and the least possible confrontation.
The spirit of scientific honesty sings in every sentence. By the time I
came to the discussion section I felt uplifted and inspired. This is how
I used to idealize science as a profession – a joy to be part of. And
here, it’s real!
I have nothing to offer by way of substance that wouldn’t be an
anticlimax. You might want to consider the relationship between loop gain
and leakage, in which steady-state loop gain is expressed as the
multiplier of the integration term divided by the fraction of output
leaked away on each iteration (the point where increases in output per
iteration with a constant error just equal the losses per iteration). But
the paper will not be diminished by leaving that for another
time.
By my reckoning, PCT reaches the age of 60 some time in 2013. A good
anniversary year in which your paper will serve as a milestone.
Thanks.
Bill P.
Cycles of Conflict 03-26-13.pdf (618 KB)
[From Kent McClelland (2013.04.26.1435 CDT)]
Bill Powers (2013.04.26.1155 MDT)
Good to hear from you, Bill, and thank you for your very kind words on my paper. It took me three or four drafts to get to the relatively restrained tone that you talk about in your post, and I hope that the paper is finally toned down enough that some
sociological journal will decide to print it.
I’d be really interested in hearing some more of your ideas about how the relationship between loop gain and leakage could be included into the substance of the paper. As I think about these issues, the loop-gain parameter seems more and more important
to me. I’d like to be better informed about whether loop gain could be included as a dynamically changing parameter in models like these.
My best,
Kent
Cycles of Conflict 03-26-13.pdf (618 KB)
On Apr 26, 2013, at 1:45 PM, Bill Powers wrote:
[From Bill Powers (2013.04.26.1155 MDT)]
Kent McClelland (2013.4.22.13:25 CDT) –
Subscribers to CSGnet may perhaps be interested in a draft article that I’ve recently written on models of conflict escalation. The paper is to be presented at a sociological convention this summer, and it’s aimed at an audience of sociologists, but it uses
PCT models and conflict is a topic of some interest in a lot of disciplines.Finally getting around to reading this MAGNIFICENT PAPER. This is the kind of relevant, substantive, grown-up paper I have always wanted to write, but have never come close to achieving. You treat Collins’ paper with complete collegial support and respect and offer your own ideas with modesty and the least
possible confrontation. The spirit of scientific honesty sings in every sentence. By the time I came to the discussion section I felt uplifted and inspired. This is how I used to idealize science as a profession – a joy to be part of. And here, it’s real!I have nothing to offer by way of substance that wouldn’t be an anticlimax. You might want to consider the relationship between loop gain and leakage, in which steady-state loop gain is expressed as the multiplier of the integration term divided by the fraction
of output leaked away on each iteration (the point where increases in output per iteration with a constant error just equal the losses per iteration). But the paper will not be diminished by leaving that for another time.By my reckoning, PCT reaches the age of 60 some time in 2013. A good anniversary year in which your paper will serve as a milestone.
Thanks.
Bill P.
[From Rick Marken (2013.4.26.1820)]
Kent McClelland (2013.4.22.13:25 CDT)
KM: Subscribers to CSGnet may perhaps be interested in a draft article that I've recently written on models of conflict escalation. The paper is to be presented at a sociological convention this summer, and it's aimed at an audience of sociologists, but it uses PCT models and conflict is a topic of some interest in a lot of disciplines.
RM: I asked Kent to send the paper to CSGNet to see if it could
generate some discussion.It looked like it wasn't until Bill chimed in
with his nice post. But I told Kent that I had a couple comments on
the paper so here they are. These comments are not criticisms because
they are not really relevant to the main point of the paper (the
temporal course of a conflict) but I bring them up because they might
be relevant to future studies of conflict.
First, on p. 12 you say: "Conflict is expected to occur whenever two
control systems operating in
the same environment have incompatible goals,..." After goals I would
just add either "for the same perceptual variable" or, possibly
better, "for the same aspect of the environment". I think this is
really the essence of conflict: conflict exists when two (or more)
control systems with incompatible (different) goals for the same
aspect of the environment. This point is kind of nit picky in the
context of your paper but I think it might be something interesting to
look into in future studies of conflict. For example, one thing I have
wanted to study about conflict was how similar the aspects of the
environment are controlled by different control systems have to be for
there to be conflict between the systems. This question comes up
because conflict can occur when systems control very similar aspects
of the environment which are not actually the same. In all the
simulations described in your paper you assume that systems are
controlling the exact same aspect of the environment relative to
different references. But suppose they are controlling "similar"
perceptions but not the same. The similarity of perceptions that are
linear functions of environmental variables can be measured in terms
of their orthogonality of the vectors that define the perceptions. For
example, if p1 = x + y and p2 = x-y, where x and y are two
environmental variables, the p1 and p2 are completely orthogonal; p1
and p2 can be controlled by two different control systems w/o any
conflict. If p1 = x+ y and p2 = x+ y then p1 and p2 are completely
non-orthogonal -- they are the same perception and they cannot be
controlled by two different systems without conflict. Intermediate
levels of orthogonality (dissimilarity) of perceptions can be defined
by using different coefficients (other than 1 and -1) for x and y.
Second, on p. 14 you say "Whenever the organism�s perceptions are
reasonably accurate, successful control of perceptual variables
results in a corresponding stabilization of environmental variables".
I think this can make sense from an observer's perspective. If the
person is supposed to control x-y and he is actually controlling a
perception that is p=x-y then when he successfully controls p he will
be successfully controlling x-y. But from the controller's
perspective, all they are controlling is p, whatever p is a function
of in the environment. So in this case, successful control means p is
kept = to r, which stabilizes whatever it is in the environment that p
corresponds to. The success of control -- the ability to keep p = r,
depends on the design of the control system, which means the
appropriate output function that "takes into account" the type of
perception being controlled and the characteristics of the
environmental connection between efferent neural signal and afferent
perceptual signal. Considering charateristics of the output function
is related to Bill's suggestion that you " might want to consider the
relationship between loop gain and leakage, in which steady-state loop
gain is expressed as the multiplier of the integration term divided by
the fraction of output leaked away on each iteration...". But I agree
with Bill that "the paper will not be diminished by leaving that for
another time".
Best regards
Rick
--
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com
[FromMike Acree (2013.04.26.2050 PDT)]
Bill Powers (2013.04.26.1155 MDT)]–
I’m pleased to see Bill’s benediction on Kent’s work. That makes it much easier to voice a criticism that I had decided to withhold; but I see now that I can frame it more constructively, as a
couple of suggestions.
My concern arose from judging the work by an impossible standard. The paper speaks about conflict as though it were a concept relevant to sociology; the Discussion
speaks, for example, about political controversies and family feuds as applications. Yet what is being modeled is only a physical tug-of-war, in a single dimension. The restriction to a single level means that there are no strategies except brute force:
to pull harder, or to pile more people on your end of the rope. And I think that induces some distortion in the operations of polarization or solidarity.
The very simplest, and still feasible, accommodation to a more complex reality would be to allow the tug-of-war to take place in a two-dimensional rather than
a one-dimensional space. That way members of opposing teams might discover that they shared, say, an upward component in their pulls, rather than merely left-right, and could actually end up a cosine closer to each of their goals. Solidarity, in such a model,
might be represented by the degree to which the members on one side were aligned in the direction they were pulling. That might do a kind of justice to the common experience in groups where all of us want to pull, say, to the left, but a lot of energy is
wasted when some people are pulling northwest, as it were, and others are pulling southwest.
Any model adequate to actual conflict in humans (or even other animals) would not only have to be multilevel, to account for phenomena like the value of conflict
itself (at what point does the tugging cease to be worth it?)—I gather Martin’s Layered Protocol Theory, though I haven’t studied it, may incorporate such multilevel phenomena—it would also have to deal with meanings—strategies of negotiation, deception, or
whatever—and it is beyond my ability to imagine how that could be given a mechanical representation. I feel some uneasiness personally to see a one-level unidimensional model put forth as the PCT account of conflict; I fear that it sets PCT up to be dismissed
out of hand as ridiculously simplistic. For that reason, although you would naturally want to focus more on what you had accomplished than what you hadn’t, I would be more comfortable (for what that may be worth!) with an additional sentence or two acknowledging
these limitations.
Mike
[From Rick Marken (2013.04.27.1812)]
Mike Acree (2013.04.26.2050 PDT) --
First, I got a letter published in the LA Times that I'm sure you
would all like to read. It's at:
Now to Mike's post:
MA: The paper [by Kent McClelland] speaks about conflict as though
it were a concept relevant to sociology; the Discussion speaks, for
example, about political controversies and family feuds as applications.
Yet what is being modeled is only a physical tug-of-war, in a single
dimension.
RM: I think Kent's model of conflict is just as relevant to real world
conflicts, like those you mention, as our simple tracking
demonstrations of control are to real world controlling. Kent's
conflict model illustrates the basic principles underlying all
conflicts; two or more people trying to keep the same perceptual
variable in different reference states. In a political controversy,
for example, one possible perceptual variable that is being controlled
for might be "the appropriate tax rate on the wealthy". Liberals want
that variable at "high", conservatives want it at "low" . The argument
involves trying to get the opponent to agree with you; the liberal
wants the conservative to say "high" and the conservative want the
liberal to say "low". The means used to "tug" the perception in the
desired direction is verbal. So the situation is exactly the same as a
tug of war except that, in the political discussion, the "flag" in the
tug of war is what the opponents say. The rope that connects the
opponents to this "flag" is the verbiage they use to try to get the
"flag" over the line to their side.
MA: The restriction to a single level means that there are no
strategies except brute force: to pull harder, or to pile more people
on your end of the rope. And I think that induces some
distortion in the operations of polarization or solidarity.
RM: Yes, that is a simplification. But Kent is dealing only with
conflict at the level at which it is "expressed" (to use Tim Carey's
terminology), ignoring the fact that the higher level systems that are
setting the conflicting goal references could resolve the conflict any
time. I think I refer to this possibility in my discussion of the
"Cost of Conflict" demo at
In this demo, a conflict exists if you have the goal of controlling
the cursor in both the x and y dimensions. The situation is set up so
that it is physically impossible to achieve these goals
simultaneously. But if you (a higher level system in you) are willing
to give up controlling the cursor in one dimension you can control it
in the other just fine. That's really the basis of the Method of
Levels; reorganizing conflicts away by changing the way higher order
systems set (or don't set) lower level goals.
MA: The very simplest, and still feasible, accommodation to a more complex
reality would be to allow the tug-of-war to take place in a two-dimensional
rather than a one-dimensional space.
RM: I would like to see what you have in mind. Maybe a functional
diagram? I think the only thing that is one dimensional about Kent's
model is that both systems are controlling a scalar perceptual
variable, p. I think all conflicts are one-dimensional in this sense;
all conflicts involve two (or more) systems controlling the same or
similar perceptions. Even the political conflict is one-dimensional in
this sense; both parties are perceiving what the other says is "the
appropriate tax rate on the wealthy". This perception is a very
complex function of lower level perceptions and, ultimately, of
sensory input. So in that sense the controlled perception of both
parties is derived from multidimensional input. But a controlled
perception, regardless of the level in the hierarchy, is, per PCT, a
uni-dimensional (scalar) variable.
MA: Any model adequate to actual conflict in humans (or even other animals)
would not only have to be multilevel
RM: And, indeed, PCT is such a model.
MA: it would also have to deal with meanings�strategies of negotiation,
deception, or whatever
RM: Absolutely, and PCT does deal with these. Meanings are dealt with
throughout the hierarchy; strategies are, I think, programs; so they
are perceived and controlled at the program level; negotiation and
deception sound like principles, controlled at the principle level by
varying references to the lower level program level.
MA: �and it is beyond my ability to imagine how that could
be given a mechanical representation.
RM: I can imagine it but I have no idea how to implement it. The main
problem of mechanization, from a PCT perspective, comes on the
perceptual rather than on the output side of control. So the problem
of mechanizing a model of conflict that can negotiate and deceive is
building a perceptual system that can perceive negotiation and
deception as well as the strategies that are being carried out to
achieve the desired levels of these perceptions.
MA: I feel some uneasiness personally to see a one-level unidimensional model
put forth as the PCT account of conflict; I fear that it sets PCT up to be
dismissed out of hand as ridiculously simplistic.
RM: I think this is a good point. But I think Kent's paper is OK as it
is because it is explicitly developing a model to account for the
"expressed conflict" that is dealt with (so poorly) by the Collins
model.
Best
Rick
--
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com
from Kent McClelland (2013.04.29.1130 CDT)]
Mike Acree (2013.04.26.2050 PDT)
Rick Marken (2013.04.27.1812)
Rick Marken (2013.4.26.1820)
Thank you to Mike and Rick for your comments on my paper models of conflict. You are making excellent points.
MA: The paper speaks about conflict as though it were a concept relevant to sociology; the Discussion speaks, for example, about political controversies and family feuds as applications. Yet what is being modeled is only a physical tug-of-war, in a single dimension. The restriction to a single level means that there are no strategies except brute force: to pull harder, or to pile more people on your end of the rope. And I think that induces some distortion in the operations of polarization or solidarity.
The very simplest, and still feasible, accommodation to a more complex reality would be to allow the tug-of-war to take place in a two-dimensional rather than a one-dimensional space. . . .
Any model adequate to actual conflict in humans (or even other animals) would not only have to be multilevel, to account for phenomena like the value of conflict itself (at what point does the tugging cease to be worth it?) . . .
RM: In all the simulations described in your paper you assume that systems are
controlling the exact same aspect of the environment relative to
different references. But suppose they are controlling "similar"
perceptions but not the same. The similarity of perceptions that are
linear functions of environmental variables can be measured in terms
of their orthogonality of the vectors that define the perceptions. For
example, if p1 = x + y and p2 = x-y, where x and y are two
environmental variables, the p1 and p2 are completely orthogonal; p1
and p2 can be controlled by two different control systems w/o any
conflict. If p1 = x+ y and p2 = x+ y then p1 and p2 are completely
non-orthogonal -- they are the same perception and they cannot be
controlled by two different systems without conflict. Intermediate
levels of orthogonality (dissimilarity) of perceptions can be defined
by using different coefficients (other than 1 and -1) for x and y.
KM: As Mike comments, by restricting myself to one-dimensional models, I've portrayed conflict in a drastically simplistic way, and it would require multi-dimension models to provide a more sociologically satisfactory account of the dynamics of conflict. Some of Rick's comments, too, pertain to the dimensionality of conflict, when he talks about the correlation between the aspects of the environment being perceived by two control systems in conflict.
As it happens, I played around years ago with models of two-dimensional conflicts, in which the perceptions controlled by two agents are linear functions in an X-Y plane. Each agent is modeled as having two outputs, one working in the X direction, and the other in the Y.
Just as Rick suggests, I tried to see what happens when you vary the correlation between the two lines that represent the references for the two simulated agents. At the time, I wasn't sure how to relate this modeling work to empirical situations, but your comments have suggested to me some ways these models could be applied. Here's what I found:
Situation A: When the two preference lines are parallel to each other (do not intersect), you get conflict dynamics similar to the one-dimensional situation, with unlimited escalation of output in opposite directions (perpendicular to the preference lines).
B: When the two preference lines are orthogonal to each other, and the initial position of the environmental variable is at a point different from the intersection of the two lines, the combined outputs of the two agents quickly bring the environmental variable to the point of intersection of the lines of preference, and then the agents hold it at that compromise point (approximately) against any random disturbances in the X-Y plane.
C: When the two preference lines meet at an acute angle and thus are somewhat correlated, the outputs of the two agents bring the environmental variable quickly to some point along a center line that splits the difference between the two preference lines, but not immediately to the point of intersection of the preference lines. Then, more slowly, their combined outputs pull the environmental variable along that compromise line till it gradually reaches the point of intersection of the two preference lines. When the environmental variable has finally reached this compromise point, the two systems hold it approximately there against disturbances.
In situation A, the preferences of the two agents are truly incompatible, continuing conflict is inevitable, and the environmental outcome is a set of compromise positions that are fixed with regard to the direction orthogonal to the two preference lines but unstable and free to vary in directions parallel to the preference lines.
In situation B, in which the preferences of the two agents are orthogonal to each other, compromise is easily reached, and the two agents stabilize the environmental variable around their compromise point.
In situation C, in which the preferences are similar, but not exactly the same, we see some initial conflict that diminishes as the environmental variable moves along the line between the two preference lines toward the compromise position at the intersection of the preferences.
A lot of real-life interactions, as Mike suggests, must be some variant of situation C, where people start with similar but not identical preferences, and the initial conflicts diminish as people find solutions that are OK with everyone involved� a degree of solidarity, if not complete agreement. One person may want more of Y when X is low and less of Y when X is high than the other does, but there's some X-Y combination that both can live with. It's only when their preference lines don't intersect at all (within the region over which they both can exert control) that conflict becomes inevitable.
You made some other good points, as well, but this seems like enough from me for the moment.
My best,
Kent
On Apr 26, 2013, at 8:19 PM, Richard Marken wrote:
[From Rick Marken (2013.4.26.1820)]
Kent McClelland (2013.4.22.13:25 CDT)
KM: Subscribers to CSGnet may perhaps be interested in a draft article that I've recently written on models of conflict escalation. The paper is to be presented at a sociological convention this summer, and it's aimed at an audience of sociologists, but it uses PCT models and conflict is a topic of some interest in a lot of disciplines.
RM: I asked Kent to send the paper to CSGNet to see if it could
generate some discussion.It looked like it wasn't until Bill chimed in
with his nice post. But I told Kent that I had a couple comments on
the paper so here they are. These comments are not criticisms because
they are not really relevant to the main point of the paper (the
temporal course of a conflict) but I bring them up because they might
be relevant to future studies of conflict.First, on p. 12 you say: "Conflict is expected to occur whenever two
control systems operating in
the same environment have incompatible goals,..." After goals I would
just add either "for the same perceptual variable" or, possibly
better, "for the same aspect of the environment". I think this is
really the essence of conflict: conflict exists when two (or more)
control systems with incompatible (different) goals for the same
aspect of the environment. This point is kind of nit picky in the
context of your paper but I think it might be something interesting to
look into in future studies of conflict. For example, one thing I have
wanted to study about conflict was how similar the aspects of the
environment are controlled by different control systems have to be for
there to be conflict between the systems. This question comes up
because conflict can occur when systems control very similar aspects
of the environment which are not actually the same. In all the
simulations described in your paper you assume that systems are
controlling the exact same aspect of the environment relative to
different references. But suppose they are controlling "similar"
perceptions but not the same. The similarity of perceptions that are
linear functions of environmental variables can be measured in terms
of their orthogonality of the vectors that define the perceptions. For
example, if p1 = x + y and p2 = x-y, where x and y are two
environmental variables, the p1 and p2 are completely orthogonal; p1
and p2 can be controlled by two different control systems w/o any
conflict. If p1 = x+ y and p2 = x+ y then p1 and p2 are completely
non-orthogonal -- they are the same perception and they cannot be
controlled by two different systems without conflict. Intermediate
levels of orthogonality (dissimilarity) of perceptions can be defined
by using different coefficients (other than 1 and -1) for x and y.Second, on p. 14 you say "Whenever the organism�s perceptions are
reasonably accurate, successful control of perceptual variables
results in a corresponding stabilization of environmental variables".
I think this can make sense from an observer's perspective. If the
person is supposed to control x-y and he is actually controlling a
perception that is p=x-y then when he successfully controls p he will
be successfully controlling x-y. But from the controller's
perspective, all they are controlling is p, whatever p is a function
of in the environment. So in this case, successful control means p is
kept = to r, which stabilizes whatever it is in the environment that p
corresponds to. The success of control -- the ability to keep p = r,
depends on the design of the control system, which means the
appropriate output function that "takes into account" the type of
perception being controlled and the characteristics of the
environmental connection between efferent neural signal and afferent
perceptual signal. Considering charateristics of the output function
is related to Bill's suggestion that you " might want to consider the
relationship between loop gain and leakage, in which steady-state loop
gain is expressed as the multiplier of the integration term divided by
the fraction of output leaked away on each iteration...". But I agree
with Bill that "the paper will not be diminished by leaving that for
another time".Best regards
Rick
--
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com
from Mike Acree (2013.04.29.1450 PDT)]
Kent McClelland (2013.04.29.1130 CDT)--
Thanks, Kent. I might have known you had already explored two-dimensional models. It's true that the second dimension doesn't add anything fundamental to the concept of conflict per se, but some complication like that still seems to me important in understanding derivative concepts like solidarity and polarization. Such "two-dimensional" conflicts within conflicts have seemed ubiquitous in my experience; as it happens, there is one going on now that is attracting international attention, as indicated in Glenn Greenwald's magnificent column in the Guardian: http://www.guardian.co.uk/commentisfree/2013/apr/27/bradley-manning-sf-gay-pride. The Board of San Francisco Pride announced the selection of Bradley Manning as Grand Marshal (in absentia) in the June parade, then quickly reversed itself. Apparently corporate sponsors with military ties were displeased, so there is now a major struggle underway between hawks and doves in the LGBT community. This could of course be analyzed as its own separate conflict; but it's the existence of other dimensions of difference which makes it possible to look for allies in the overlaps or correlations.
Mike
[From Rick Marken (2013.04.30.0825)]
Mike Acree (2013.04.29.1450 PDT) to Kent McClelland (2013.04.29.1130 CDT)–
MA: Thanks, Kent. I might have known you had already explored two-dimensional models. It’s true that the second dimension doesn’t add anything fundamental to the concept of conflict per se, but some complication like that still seems to me important in understanding derivative concepts like solidarity and polarization. Such “two-dimensional” conflicts within conflicts have seemed ubiquitous in my experience; as it happens, there is one going on now that is attracting international attention, as indicated in Glenn Greenwald’s magnificent column in the Guardian: http://www.guardian.co.uk/commentisfree/2013/apr/27/bradley-manning-sf-gay-pride. The Board of San Francisco Pride announced the selection of Bradley Manning as Grand Marshal (in absentia) in the June parade, then quickly reversed itself. Apparently corporate sponsors with military ties were displeased, so there is now a major struggle underway between hawks and doves in the LGBT community. This could of course be analyzed as its own separate conflict; but it’s the existence of other dimensions of difference which makes it possible to look for allies in the overlaps or correlations.
RM: I guess I don’t really understand what you guyd mean by “dimensions” of a conflict. Maybe one of you could explain it to me?
I, for example, don’t see Kent’s X,Y conflict as “two dimensional”. The perception that s in conflict is a function of the two variables (dimensions of variation), X and Y. But if there is a conflict the systems must be controlling the same or a similar function of X and Y; so there is really only one dimension of “contention” on any conflict (I think). That is p1 (the perception controlled by system 1) must be a1X+b1Y and p2 must be a2X+b2Y and there is conflict when the coefficients a1, a2, b1 ,b2 define highly correlated perceptions, p1 and p2.
The multidimensionality here is about the perceptions, not the conflict. The conflict you describe above seems just like the conflicts modeled in Kent’s paper; the variable in conflict is “Manning as Grand Marshal” which is a binary variable that can be in two states, true or false. The conflict is apparently an interpersonal one; one group want’s this variable to be true, the other wants it to be false. The latter group seems to have “won” the conflict since the current state of the variable is false. The people who want this variable to be true must be experiencing considerable error but the group on the other side of the conflict can apparently has higher gain. this conflict was not resolved; it is just won by the stronger side. The losers will either rise up again at a later time or they will reorganize and figure out some some new way to control for their higher level goals.
Best
Rick
–
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com
[FromMike Acree (2013.04.30.0842 PDT)]
Rick Marken (2013.04.30.0825)–
RM: I guess I don’t really understand what you guyd mean by “dimensions” of a conflict. Maybe one of you could explain it to me?
MA: I left the primary conflict implicit in this account: Everyone involved in Gay Pride was pulling for the gay community, and against the oppression of it by others. But as we were all pulling to the left, let us say, on that
issue, some of us at the same time were pulling up and others were pulling down, on the status of Bradley Manning. Either conflict, as I said earlier, can be modeled separately; the two-dimensional approach seemed to me useful in representing more complex
phenomena like solidarity and polarization, and in opening a space for actions like negotiation.
Mike
[From Rick Marken (2013.04.30.2200)]
Mike Acree (2013.04.30.0842 PDT)–
Rick Marken (2013.04.30.0825)–
RM: I guess I don’t really understand what you guyd mean by “dimensions” of a conflict. Maybe one of you could explain it to me?
MA: I left the primary conflict implicit in this account: Everyone involved in Gay Pride was pulling for the gay community, and against the oppression of it by others.
RM: So “dimensions” are different conflicts. There’s the conflict over Manning being Grand Marshal, the conflict over the oppression of the gay community, etc. I don’t see how one conflict is “primary” relative to the others. But PCT certainly can handle multiple conflicts between multiple control systems.
MA: But as we were all pulling to the left, let us say, on that
issue, some of us at the same time were pulling up and others were pulling down, on the status of Bradley Manning. Either conflict, as I said earlier, can be modeled separately; the two-dimensional approach seemed to me useful in representing more complex
phenomena like solidarity and polarization, and in opening a space for actions like negotiation.
RM: So the two dimensional approach is more than about there being two separate conflict; it has something to do with the relationship between conflicts. I guess I see the situatoin you describe more in terms of the good ol’ PCT hierarchy. You are dealing with lots of people controlling for all kinds of different perceptions but I imagine most in the Gay Pride parade were acting to push against a perception of oppression from non-Gays, many of whom probably do control for doing things that could be perceived as oppression of Gays. Making Manning Grand Marshall was one way of pushing against this oppression; but there was push back against Manning from the military. Some in the Gay Pride group appeared to fold on this conflict. Some may have folded because they had goals regarding the military that created an internal conflict that made it impossible for them to push for Manning; others may have folded because they didn’t have the skill (gain) to push back; others may have folded because they had goals about avoiding interpersonal conflicts that they felt would hurt their cause. And some probably didn’t care (not controlling for Manning being or not being Grand Marshal).
I think it would be possible to create a model of this situation as an interaction between hierarchical control systems, some of which are controlling for the same perceptions (such as oppression of Gays) at different levels (none versus some), some of which are also controlling for perceptions that put them in internal conflict about fighting the military, etc. All you have to do is do it and given your math skills I think that would be a great project for you, Mike.
Best
Rick
[FromMike Acree (2013.04.30.2219 PDT)]
Rick Marken (2013.04.30.2200)–
MA: But as we were all pulling to the left, let us say, on that issue, some of us at the same time were pulling up and others were pulling down, on the status of Bradley Manning. Either conflict, as I said earlier,
can be modeled separately; the two-dimensional approach seemed to me useful in representing more complex phenomena like solidarity and polarization, and in opening a space for actions like negotiation.
RM: So the two dimensional approach is more than about there being two separate conflict; it has something to do with the relationship between conflicts. I guess I see the situatoin you describe more in terms of the good ol’ PCT hierarchy.
. . .
RM: I think it would be possible to create a model of this situation as an interaction between hierarchical control systems, some of which are controlling for the same perceptions (such as oppression of Gays) at different
levels (none versus some), some of which are also controlling for perceptions that put them in internal conflict about fighting the military, etc. All you have to do is do it and given your math skills I think that would be a great project for you, Mike.
MA: No question you can do a lot more with hierarchical PCT; I suggested as much in my original post, and no doubt Kent was aware of that long ago. I was merely trying to suggest a more adequate representation
within Kent’s constraint of nonhierarchical modeling.
Mike
[From Erling Jorgensen (2013.05.07.17:50 EDT)]
Kent McClelland (2013.4.22.13:25 CDT)]
Hi Kent,
I'm finally getting around to responding to your paper on conflict
escalation. I find myself with some measure of 'internal conflict'
over the various things I want to get done. I suppose I use a
time-sharing strategy, to avoid the kind of runaway feedback process
that you elegantly model & describe.
I agree with Bill P. & others, that it is a masterful paper. You seem
to have operationalized several of the constructs that Collins proposes
for explaining conflict. I like how you show (on p.23) that the
positive feedback process of his theory can be arrived at by two or
more _negative_ feedback controllers working at cross purposes, &
escalating their efforts with each iteration even though it does not
necessarily make a difference for the "contested variable" they are
trying to control.
I also like the contrast you display between Figure 5B & Figure 5C,
showing that the same objective outcome can emerge from either
conflictual or coorperative means. When it is cooperative, the output
resources of each of the parties are drastically reduced. Those
figures do imply, however, that if the reference standards for the
parties are not perfectly aligned, the variable will be contested &
the outputs will escalate & diverge.
I wonder about the kind of method Martin Taylor has discussed from
time to time, with putting thresholds or 'a zone of perceived no-error'
(my term, not his) around an outcome that is considered "good enough."
For instance, with the situation simulated by Figure 5B, maybe a value
of '20' for the contested variable is close enough to the preferred
values of '25' & '15' for Agents 2 & 4, respectively, & some of the
intensity of the conflict in terms of divergence of outputs could be
reduced. This would obviously entail adding some kind of hierarchical
appraisal of the current state of the contested variable, or its
prospects for change over time (would that be some kind of derivative
input function?). With my group therapy clients, I sometimes
illustrate the kind of reappraisal I'm talking about, using the words
of a song from the sixties: "If you can't be with the one you love,
love the one you're with!"
I especially like how the clean & elegant modeling you have done,
which you with proper modesty call "radically simplified" (p. 15),
can nonetheless allow the reader to wonder about these kinds of
supplements & modifications. Whenever a paper draws a reader in,
to want to improve upon the model, I think that is a powerful sign
of the key argument getting through to your intended audience.
I do have a question about how you are interpreting a portion of your
simulation. On page 28 or your draft, you discuss proposed changes
to the situation depicted in Figure 5B. It is part of your discussion
about how to understand "solidarity," whether as coalescing around
a common reference standard or as an increase in "the loop gain that
actors contribute to the joint action."
You make the claim that coming together around "a more extreme
reference value" is an instance of increased solidarity which
"improves effectiveness in conflict" (p.28). But you seem to contrast
that with coming together around a less extreme value. You say:
"And if all four agents had adopted the reference value of 10, the
lowest value in their group, their solidarity as a group would have
increased, but with a decrease in their collective impact on the
contested variable." (p.28)
I don't understand how that situation would lead to "a decrease in
their collective impact" on the controlled variable. With all four
agents now pulling it from a value of close to '20' to one close to
'10', isn't that having exactly the impact that all of them want?
If anything, by increasing their solidarity in terms of a shared
reference standard, their impact & effectiveness in terms of the
variable actually arriving at that value seems to have increased.
I don't think the corresponding conclusions that you draw, (on that
same page & later in the paper,) necessarily hold up, regarding "more
extreme" versus "less extreme" positions.
I do agree that Collins' view of solidarity as always increasing the
intensity of conflict does not match what your modeling displays. When
reference standards align -- which seems the best way to operationalize
solidarity -- the whole conflictual enterprise can even be made
cooperative, with a corresponding decrease in respective outputs (as
shown by your Figure 5C).
However, it does seem that when there is a difference of reference
standards, then the agents pulling in a common direction will be
pooling the effects of their loop gain. In that case, then a
conceptual definition of solidarity along the lines of 'shared
sacrifice', as demonstrated by loop gains, may come into play. As
you note, the solidarity variable is "the least clearly defined of
the [Collins] model's core variables" (p.23).
All in all, Kent, this is a very well writen & thought provoking paper,
one that makes a significant contribution to the literature, by laying
out a straightforward mathematical model so we can see the effects
of different configurations of conflictual variables. Thanks for
your work.
All the best,
Erling
[David Goldstein (2013.04.27.11:10)]
[Mike Acree (2013.04.26.2050 PDT)]
Bill Powers (2013.04.26.1155 MDT)]
Even though you are
correct in what you say, the one-dimension can be at a high level, for example, religion. Higher level perceptions are created from combining many lower level perceptions. Thus, this one dimension can be very complicated.
In addition, as a clinical psychologist, conflicts within a couple, can be simple–a woman likes the house temperature higher than the man she is living with; another example, a husband is more frugal than his wife and they conflict about money and staying within a budget.
In the political world, some people want more regulation of gun purchases for some
reasons and other people do not for different reasons. When it comes to
voting for law,
I thought the article was very well written and based on the literature. It shows how what appears to be a positive feedback situation can be understood in terms of negative feedback. It was dealing with interpersonal conflict, not intrapersonal conflict. However, the method of resolving conflicts is the same–putting awareness on the conflict with the hope that there is some way of coming up with a solution that is acceptable.
From: “Acree, Michael” AcreeM@OCIM.UCSF.EDU
To: CSGNET@LISTSERV.ILLINOIS.EDU
Sent: Friday, April 26, 2013 11:50 PM
Subject: Re: Paper on models of conflict escalation
[From Mike Acree (2013.04.26.2050 PDT)]
Bill Powers (2013.04.26.1155 MDT)]–
Iâm pleased to see Billâs benediction on Kentâs work. That makes it much easier to voice a criticism that I had decided to withhold; but I see now that I can frame it more constructively, as a
couple of suggestions.
My concern arose from judging the work by an impossible standard. The paper speaks about conflict as though it were a concept relevant to sociology; the Discussion
speaks, for example, about political controversies and family feuds as applications. Yet what is being modeled is only a physical tug-of-war, in a single dimension. The restriction to a single level means that there are no strategies except brute force:
to pull harder, or to pile more people on your end of the rope. And I think that induces some distortion in the operations of polarization or solidarity.
The very simplest, and still feasible, accommodation to a more complex reality would be to allow the tug-of-war to take place in a two-dimensional rather than
a one-dimensional space. That way members of opposing teams might discover that they shared, say, an upward component in their pulls, rather than merely left-right, and could actually end up a cosine closer to each of their goals. Solidarity, in such a model,
might be represented by the degree to which the members on one side were aligned in the direction they were pulling. That might do a kind of justice to the common experience in groups where all of us want to pull, say, to the left, but a lot of energy is
wasted when some people are pulling northwest, as it were, and others are pulling southwest.
Any model adequate to actual conflict in humans (or even other animals) would not only have to be multilevel, to account for phenomena like the value of conflict
itself (at what point does the tugging cease to be worth it?)I gather Martinâs Layered Protocol Theory, thoough I havenât studied it, may incorporate such multilevel phenomenait would also have to deal with meaningsstrategies of negnegotiation, deception, or
whateverand it is beyond my ability to imagine how that could be given a mechanical representation. I feel some uneasiness personally to see a one-level unidimensional model put forth as the PCT account of conflict; I fear that it sets PCT up to be dismissed
out of hand as ridiculously simplistic. For that reason, although you would naturally want to focus more on what you had accomplished than what you hadnât, I would be more comfortable (for what that may be worth!) with an additional sentence or two acknowledging
these limitations.
Mike
[From Rupert Young 2013.05.08 19.00 BST]
None of the figures are visible in the document for me. Though I see from other's responses that they discuss the figures. Is there a way get them to appear?
Regards,
Rupert
On 22/04/2013 19:33, McClelland, Kent wrote:
[From Kent McClelland (2013.4.22.13:25 CDT)]
Subscribers to CSGnet may perhaps be interested in a draft article that I've recently written on models of conflict escalation. The paper is to be presented at a sociological convention this summer, and it's aimed at an audience of sociologists, but it uses PCT models and conflict is a topic of some interest in a lot of disciplines.
Best to all,
Kent
Did you look at the end of the document. That's where they appear.
Fred Nickols
From: Control Systems Group Network (CSGnet)
[mailto:CSGNET@LISTSERV.ILLINOIS.EDU] On Behalf Of Rupert Young
Sent: Wednesday, May 08, 2013 1:54 PM
To: CSGNET@LISTSERV.ILLINOIS.EDU
Subject: Re: Paper on models of conflict escalation[From Rupert Young 2013.05.08 19.00 BST]
None of the figures are visible in the document for me. Though I see from
other's responses that they discuss the figures. Is there a way get them
to
appear?
Regards,
Rupert> [From Kent McClelland (2013.4.22.13:25 CDT)]
>
> Subscribers to CSGnet may perhaps be interested in a draft article that
I've
recently written on models of conflict escalation. The paper is to be
presented at a sociological convention this summer, and it's aimed at an
audience of sociologists, but it uses PCT models and conflict is a topic
of some
-----Original Message-----
On 22/04/2013 19:33, McClelland, Kent wrote:
interest in a lot of disciplines.
>
> Best to all,
>
> Kent
>
Ah, I'd printed out everything up to the references to save paper, so didn't spot them. Thanks.
Regards,
Rupert
On 08/05/2013 19:37, Fred Nickols wrote:
Did you look at the end of the document. That's where they appear.
Fred Nickols
-----Original Message-----
From: Control Systems Group Network (CSGnet)
[mailto:CSGNET@LISTSERV.ILLINOIS.EDU] On Behalf Of Rupert Young
Sent: Wednesday, May 08, 2013 1:54 PM
To: CSGNET@LISTSERV.ILLINOIS.EDU
Subject: Re: Paper on models of conflict escalation[From Rupert Young 2013.05.08 19.00 BST]
None of the figures are visible in the document for me. Though I see from
other's responses that they discuss the figures. Is there a way get themto
appear?
Regards,
RupertOn 22/04/2013 19:33, McClelland, Kent wrote:
[From Kent McClelland (2013.4.22.13:25 CDT)]
Subscribers to CSGnet may perhaps be interested in a draft article that
I've
recently written on models of conflict escalation. The paper is to be
presented at a sociological convention this summer, and it's aimed at an
audience of sociologists, but it uses PCT models and conflict is a topicof some
interest in a lot of disciplines.
Best to all,
Kent
[From Kent McClelland (2013.05.08.1555 CDT)]
Erling Jorgensen (2013.05.07.17:50 EDT)
Thank you Erling, Bill, Michael, Rick, and David for saying nice things about the paper. I'm glad you liked it. I only hope your favorable impression is eventually shared by some group of editors and reviews so that I can get it into print.
On my first try (admittedly with a highly selective journal in sociology) I struck out. But the reviewers did have some useful things to say, which I think have helped me to improve this draft. I've now sent it to a less prestigious journal, and we'll see what happens.
EJ: I do have a question about how you are interpreting a portion of your
simulation. On page 28 or your draft, you discuss proposed changes
to the situation depicted in Figure 5B. It is part of your discussion
about how to understand "solidarity," whether as coalescing around
a common reference standard or as an increase in "the loop gain that
actors contribute to the joint action."You make the claim that coming together around "a more extreme
reference value" is an instance of increased solidarity which
"improves effectiveness in conflict" (p.28). But you seem to contrast
that with coming together around a less extreme value. You say:"And if all four agents had adopted the reference value of 10, the
lowest value in their group, their solidarity as a group would have
increased, but with a decrease in their collective impact on the
contested variable." (p.28)
KM: By "effectiveness in conflict" here, I meant ability to move the position of the contested variable to a value far from zero, the presumed neutral point along the continuum of the conflict, despite any disturbances or the efforts of agents in the opposing coalition. More effective coalitions of agents, as I see it, can combine their efforts to move the contested variable farther in the direction their side's preferences. That's what I meant by having more of an impact.
In Figure 5B, the combined efforts of all four agents have moved the contested variable to a value of (almost) 25, the average of their preferences. If the average value of their preferences were only 10, instead of 25�whether or not they agreed precisely in their reference values� their combined efforts (in the absence of any disturbances) would move the variable to a value of approximately 10, not 25, so it seems to me that they would then be making less of a collective impact.
EJ: I don't understand how that situation would lead to "a decrease in
their collective impact" on the controlled variable. With all four
agents now pulling it from a value of close to '20' to one close to
'10', isn't that having exactly the impact that all of them want?
If anything, by increasing their solidarity in terms of a shared
reference standard, their impact & effectiveness in terms of the
variable actually arriving at that value seems to have increased.
I don't think the corresponding conclusions that you draw, (on that
same page & later in the paper,) necessarily hold up, regarding "more
extreme" versus "less extreme" positions.
KM: For a real-life example, suppose that there is a heated controversy between two sides, where both sides stake out extreme positions, but then the combatants on one side get together and decide that their position has been much too extreme, and they agree to take a more "moderate" position as a group. If the other side continues to hold to its extreme views, the the compromise outcome in the conflict between the two sides (the resultant of everyone's efforts) will shift in the direction of the side that hasn't moved toward the middle (on average).
Some observers might argue that something like this has happened with the two political parties in the US over the last 30 years, in which the Democrats have decided that some of their previous liberal positions were too liberal and Republicans have shifted in an ever more conservative direction, so that the entire center of gravity of the political discourse has shifted to the right. In this case, neither side has shown a great deal of solidarity within its ranks, but because average reference values have shifted, the Republican views have had more of an impact on the outcome of the conflict.
The amount of solidarity (defined as agreement within a side) might make a difference, of course. If the distribution of reference values held by one party overlaps with the distribution of of references of the other (like "blue-dog" Democrats and liberal Republicans�a breed that seems now to be extinct), the efforts of some members of the party may end up aiding the other side. But Erling's criticism hasn't changed my opinion that the extremity of references (on average) among members of one side of a conflict has more impact on the outcome of the conflict than their degree of consensus.
In most social movements, the leaders hold more extreme views than the rank-and-file. The willingness of leaders to stake out extreme positions helps them to get attention and thus to move into leadership positions. When such leaders call for more solidarity, what they really want is for other members to join them in taking these more militant positions. If they can get the membership mobilized around their own extreme positions, the organization will become more effective as a fighting unit, but if leaders become discouraged and begin to moderate their own extreme positions, the organization will lose some of its combat effectiveness. (Note that I'm not necessarily arguing that combat effectiveness is a good thing. Depends on what they're fighting for. And a strategy of less combativeness may actually be more effective in the long run for accomplishing some of an organization's goals. That's another story entirely.)
EJ: I do agree that Collins' view of solidarity as always increasing the
intensity of conflict does not match what your modeling displays. When
reference standards align -- which seems the best way to operationalize
solidarity -- the whole conflictual enterprise can even be made
cooperative, with a corresponding decrease in respective outputs (as
shown by your Figure 5C).
KM: Agreed. (But then it's no longer a conflict!) Thanks again for your comments.
To Bill, Michael, Rick, and Martin: I'm still trying to pull my thoughts together on the question of conflict in multiple dimensions. A posting will follow eventually.
Kent
[From Erling Jorgensen (2013.05.10.16:45 EDT)]
Kent McClelland (2013.05.08.1555 CDT)
Thanks for your clarifications, Kent. Allow me to continue my own
convergence toward understanding your perspectives.
By "effectiveness in conflict" here, I meant ability to move the
position of the contested variable to a value far from zero, the
presumed neutral point along the continuum of the conflict
Two thoughts here. I had the impression that the Y-axis on your
figures was essentially in arbitrary units. So I am not sure that
"zero" represents the neutrality that you seek.
More centrally, when we are talking about reference standards, a
value of "zero" is an expressed preference for a given perception.
That is to say, "keep it at zero," rather than "I don't care what
it is."
I believe the way to express "I don't care," i.e. genuine neutrality,
would be to bring the gain down to zero.
More effective coalitions of agents, as I see it, can combine their
efforts to move the contested variable farther in the direction [of]
their side's preferences. That's what I meant by having more of an
impact.
Yes, & I believe the leverage for this to happen (although, see the
further discussion below) is the combined gain that such coalitions
bring to bear on the process. Output pulling on a variable towards
a different reference point is just that, & is treated by controllers
with their own preferred reference as a disturbance acting on the
variable of interest.
I believe your simulations of conflict compute the average of the
reference values _as weighted by_ the gain, as shown in the two
'stale-mate' scenarios of Figure 4 & Figure 5B. So the relative
difference in gain between competing sides affects where the
variable lands, according to what we sometimes call a "virtual
reference signal" compounded out of the gain-weighted references of
all the parties.
Having said all that, I do believe you are onto something when you
talk about "extreme positions" in a conflict:
For a real-life example, suppose that there is a heated controversy
between two sides, where both sides stake out extreme positions,
but then the combatants on one side get together and decide that
their position has been much too extreme, and they agree to take
a more "moderate" position as a group. If the other side continues
to hold to its extreme views, the the compromise outcome in the
conflict between the two sides (the resultant of everyone's efforts)
will shift in the direction of the side that hasn't moved toward
the middle (on average).
I have heard that essential analysis, & I think I agree, as applied
to some of the political polarization in the US that you cite next
as an example:
Some observers might argue that something like this has happened
with the two political parties in the US over the last 30 years,
in which the Democrats have decided that some of their previous
liberal positions were too liberal and Republicans have shifted in
an ever more conservative direction, so that the entire center of
gravity of the political discourse has shifted to the right. In
this case, neither side has shown a great deal of solidarity within
its ranks, but because average reference values have shifted, the
Republican views have had more of an impact on the outcome of the
conflict.
I do think, however, that if the more moderate position had sufficent
gain -- consider the talk way-back-when about "the silent majority"
-- then the center of gravity does not necessarily shift just because
there are more extreme positions in play.
I'm thinking of the example of a New England town meeting, which in
New Hampshire is still held in high regard! There may be significant
spirited promotion & spirited dissent over whether the Town should
spend X amount of dollars, with Y amount of increase in the tax rate,
on a new fire engine for the fire hall. And yet it is frequently
the case that a more moderate position carries the day. We even have
a colloquial term for it -- "cooler heads prevailed."
I don't think that simply racheting out one's reference to its most
extreme level is any guarantee of getting one's way. And, indeed,
don't the stale-mate solution sets of your simulations demonstrate
that?
In any event, I'm not fully convinced yet, but that's okay.
Fortunately, we as more complex control systems have hierarchical
variables & preferences that can temper things -- such as valuing
& maintaining a respectful relationship & dialogue.
With much respect,
Erling
[From Rick Marken (2013.05.14.0835)]
Erling Jorgensen (2013.05.10.16:45 EDT)–
EJ: I believe the way to express “I don’t care,” i.e. genuine neutrality,
would be to bring the gain down to zero.
RM: Yes, exactly. A reference of 0 means “I do care, and I want none of it”. Sort of the way I feel about free market economics (and Republicans in general). Whereas a gain of 0 (with reference of 0) means “I want none of it but I’m not going to do anything about it”.
Best
Rick
–
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com