Elections as collective control

RM: In another thread Martin Taylor said:

MT: …Elections are prime examples of stochastic collective control. Changes of “public sentiment” obtained by frequent enough polling could be seen as continuous collective control in N-dimensions, the N dimensions being the properties being polled, such as satisfaction with the government’s performance on fishing rights.

RM: And I asked:

RM: What is the collectively controlled variable here? I don’t see any variable in this example that corresponds to one where the actions of a collective of controllers keep the variable in a reference state that is the same as would be observed if that variable were being controlled by a single controller.

RM: Apparently I asked this in the wrong thread so I am hoping to get an answer in this one.

Best, Rick

In a general election there are N parties. Each voter has a reference for party X to be perceived as the governing party. That’s a variable controlled by the action of voting. Many voters cast a vote for party X, the entire collective controlling their perception of the same environmental variable. The collective controlling for seeing party X to be in government is just one example of collective controllers in the election, while other groups collectively control for perceiving party Y, Z, W, etc. to be in government. The collective controllers controlling for government to include their preferred party (usually) are in conflict, the N-1-dimensional virtual controlled perceptual variable being the mix of parties included in government.

Just as in Kent’s original conflicted control demo, the pulls by the N Giant Virtual Controllers result in a party mix that is not the reference mix for any of the individual GVCs. It is only changed by another election (unless members of one party collective move to join a different one — in the Anglo-Canadian Parliamentary, this is called “crossing the floor” — and cause the governing coalition to lose its majority).

Elections are a form of stochastic collective control, whereas membership in a party collective is pseudo-continuous, as I described in the preceding thread Varieties of neural activity - #5 by MartinT.

Based on the fragments quoted, choice of person for the office or choice of outcome for the issue being voted on is the collectively controlled variable, and the result of the election (or the result of the current poll) is de facto the collectively controlled value of that variable.

In PCT analysis, it’s always important to note that “to control” is not equivalent either to “to bring a perception to its reference value” or “to control successfully”.

“To control” is wider than that. It includes the effects of any actions that influence the environmental variable corresponding to the perceptual variable, if they are the result, however ineffectual or counterproductive, of output from the controller of that perception.

Voters may have any reason to cast a vote for Party X. They may be voting for the party that best agrees with their opinion on a specific issue, say the right to fish all the salmon out of a river, or they may vote for X because although they might not specifically control with high gain any particular policy of any party, they might perceive the general tenor of the party’s policies as showing that X is generally more aligned with the voter’s reference values than are the specific policies of the other parties. Either way, it takes the combined effects of a collective for a GVC to exist that controls succesfully for party X to be in the government.

If ai understand you, Martin, you are saying that to control is simply to have your actions affect the environmental variable that corresponds to the perceptual variable (a.k.a. the controlled variable).

Fred Nickols

Yes, Fred.

The point here is that reorganization is what improves the Quality of Control (QoC) over time. Either the QoC for that perception improves to a level where the control has some useful effect on the dynamics of the intrinsic variables, or reorganization will eliminate that particular perceptual function as a waster of energy, and therefore a bad influence on the system of intrinsic variables, which can make good use of any energy available. Either way, controllers with low QoC that do nothing useful but do have side-effects that disturb other controllers are unlikely to survive long.

In your domain of interest, I would think of such poor controllers as analogous to the person who does no useful work but spends his/her time haranguing the effective workers about what they should be doing to make their lives better rather than what they are currently doing in their normal work.

MT: In a general election there are N parties. Each voter has a reference for party X to be perceived as the governing party. That’s a variable controlled by the action of voting. Many voters cast a vote for party X, the entire collective controlling their perception of the same environmental variable.

RM: Right off the bat this situation is not the same as that in Kent’s “collective control” demo. In Kent’s demo, each party to the conflict can affect on and, thus, control the controlled variable on their own; in an election no single party to the conflict can affect on the controlled variable and, thus, could not control it on their own.

MT: Just as in Kent’s original conflicted control demo, the pulls by the N Giant Virtual Controllers result in a party mix that is not the reference mix for any of the individual GVCs.

RM: But the situation is not very much like Kent’s demo. One difference is noted above: unlike in Kent’s demo, none of the parties to the conflict could control the state of the controlled variable on their own. Another difference is that none of the parties can perceive the current state of the controlled variable; so the actions of the parties cannot be based on the state of that variable (relative to their reference for it), as they are in Kent’s demo.

RM: These differences (and others that you see when you approach the study of elections without trying to fit them into the Procrustean bed of “collective control”) require a very different approach to applying the PCT model than that used in Kent’s demo. Indeed, the model used in Kent’s demo really only applies to a situation such as a tug of war between N parties, each attached by rope to the same flag, the position of which is the controlled variable. It also requires that all parties be of approximately the same strength (gain) with the same strength limits. In that case you end up with the flag in a virtual reference state that is the average of the reference states for the position of the flag for all parties.

RM: An election is similar to Kent’s conflict inasmuch as the all parties to the conflict could end up not getting the results they want. This could be true in parliamentary elections if the voters (the parties to the conflict) are controlling for a particular distribution of parties represented in the parliament. But even in parliamentary systems, a member of only one of the parties eventually becomes the leader. So, assuming that voters are also controlling for which party leads the country, it is possible that some voters will get what they want and others will not. In this case the controlled variable doesn’t end up in a “virtual reference state” as it does in Kent’s dem; instead the controlled variable ends up in the reference state of the voters who wanted the winning party to lead.

RM: The point of all this is to show the problem involved in trying to impose an explanation (“collective control”) on a phenomenon (elections) rather then going from phenomenon to explanation (PCT). When I do the latter with elections what I see, first, is not just conflict but also cooperation. The conflict comes from the fact that different members of a group have different references for who should lead them. One way to deal with this conflict is by force; the one with the biggest army and/or mouth claims leadership. The other way is by some form of democratic process, such as an election. This requires cooperation: the parties to the conflict have to agree, among other things, to vote at a particular time in a particular manner and, of course, they have to agree to accept the results of the vote tally. We saw in the US election that these tacit agreements cannot be taken for granted! Cooperation requires trust ,so it’s a risky business, but the benefits usually outweigh the costs.

RM: Cooperation is obviously an important aspect of social behavior. It’s a kind of behavior so it must involve control since, as we know from PCT, behavior is control. Cooperation comes in at least two forms: intended and unintended (See Chapter 7 in The Study of Living Control Systems). Elections involve intentional cooperation, which is very risky; the flocking of birds and the honey making behavior of bees involves unintentional cooperation; much less risky.

RM: The point of all this is simply that there are more things in the behavior of collections of living control systems than are dreamed of in the “collective control” demo. Which is why I didn’t use that term in the Social Control Chapter of my book (I did mention Kent’s demo as one example of “conflictive control”). But I do hope (which is probably like troubling deaf heaven with my bootless cries) that some readers will take a look at Chapter 7 in my book and see if it gives you any ideas about how to look at social behavior thorough a broader PCT lens.

Best, Rick

Of course it isn’t. There are lots more kinds of collective control, some of which look like Kent’s demonstration, some of which don’t, as I said right at the beginning of this transferred thread when you asked me what I meant by “collective control” and you told me that I didn’t meen whar I said I meant.

Think about it.

Martin

RM: Right off the bat this situation is not the same as that in Kent’s “collective control” demo.

MT: Of course it isn’t. There are lots more kinds of collective control, some of which look like Kent’s demonstration, some of which don’t, as I said right at the beginning of this transferred thread when you asked me what I meant by “collective control” and you told me that I didn’t mean what I said I meant.

MT: Think about it.

RM: And I think (actually, I know) that this definition of collective control does not include elections or most other examples of social control. The problem is that your “GVC” exists only when all participants "…have some influence on a variable that is the same as would be observed if that variable were perceived and controlled by some single controller with a virtual reference level [Emphasis mine]. As I said in my previous post, you get the GVC phenomenon only when (as in Kent’s demo) each participant would be able to control the controlled variable on their own. And obviously, that can’t be done a voter in an election.

RM: Think about it. Or, better yet, write a control model of an election and see what happens. Have each voter controlling for a win by either candidate 1 or candidate 2. See if you get the model to end up controlling for a win by candidate 1.5.

Best, Rick

“It ain’t what you don’t know that gets you into trouble. It’s what you know for sure that just ain’t so” (Mark Twain).

That is correct. The PCT understanding of collective control has advanced quite a bit since that demo, and is much more comprehensive. Chapters 9 and 10 in The Handbook (LCS IV) are a good place to start, though not current due to the publication delay.

RM: The point of all this is simply that there are more things in the behavior of collections of living control systems than are dreamed of in the “collective control” demo.

BN: That is correct. The PCT understanding of collective control has advanced quite a bit since that demo, and is much more comprehensive.

RM: Has it advanced beyond the understanding of collective control as defined by Martin in the Neural Activity thread:

MT: I understand a taxonomy of types of collective control, which all have in common just one property, that the actions of two or more distinct controllers all have some influence on a variable that is the same as would be observed if that variable were perceived and controlled by some single controller with a virtual reference level.

RM: It’s that “one property they all have in common” that I find troublesome.

BN: Chapters 9 and 10 in The Handbook (LCS IV) are a good place to start, though not current due to the publication delay.

RM: All I see in those chapters is theory based on arm chair speculation and anecdote. The only way to demonstrate (to me, anyway) that collective control is a useful theoretical addition to PCT is to show how collective control models can account for the real world behavior of groups of living control systems. I do this using plain vanilla PCT in the Social Control chapter of The Study of Living Control Systems. Until I see that that can be done with collective control theories I’ll continue to consider them useless extensions to PCT.

Best, Rick

Thank you for summarizing your position so well.

Conjectures cannot be modeled until they have been conjectured. That is their primary usefulness. (They are also useful in other ways for the advancement of PCT and of IAPCT.)

‘Armchair speculation’ and anecdotal observations have a legitimate place in the advancement of any science and should not be discouraged or stifled.

BN: ‘Armchair speculation’ and anecdotal observations have a legitimate place in the advancement of any science and should not be discouraged or stifled.

RM: Their legitimate place is in net discussions, working papers or even peer reviewed articles (if the peers rate them as acceptable), not in a Handbook that purports to contain authoritative papers on PCT.

Best, Rick

No error, no actions, but it is still being controlled.
Low gain, little or no action, but it is still being controlled.
But:
No actions affecting a (purported) controlled variable, no evidence of control.