The Nature of Collective Control

Collective Control is a term introduced by Kent McClelland. I first heard of the concept at the 1993 CSG group meeting in Durango, where McClelland demonstrated that two controllers attempting to influence a single environmental variable that corresponds to a perception they both control produce effects on that variable that would look to an external observer as though a single controller was acting with a reference value that was a weighted average of the two reference values, weighted by their individual gains and a gain that was the sum of their gains. This single controller was a virtual controller, but without perceiving the two individual controllers it would be impossible for any experimenter to distinguish it from a real controller.

McClelland’s 1993 demonstration was of a single variable that changes only in its location along some curve in a complicated environment. Collective control is instructive in this simple environment, in that it illustrates the creation of a virtual controller in a very unusual situation. The concept, however, becomes more interesting and powerful when we recognize that it is unlikely that any two living control systems actually control “the same” perceptions within the complex environment in which they live, unless the environmental variable’s motion is constrained to change only along a well-defined curve. Even when the variable is as constrained as the location of a patio door sliding along its track, it is unlikely that two conflicting control systems would push or pull the environmentally constrained object from precisely opposed directions.

Suppose this variable were the X-location of a marker on a board. The marker also has a Y-location. Suppose another two players want to control its Y-location with different reference values. Between them they would create a second “Virtual Controller” (VC). The two virtual controllers would not interfere with each other. The X-direction VC would correct against disturbing force applied in the X-direction, while the Y-VC independently controlled against disturbing force in the Y-direction. But what if the disturbing agency were a fluctuating wind generating a force that pushes the marker in varying directions? Then it would, most of the time, disturb the controlled perception in both controllers, and both would act to oppose the magnitude of the force in the direction it controls. The result would be that the X and Y VCs would together create a two-dimensional Virtual Controller that would control against the disturbance created by the wind with a two-dimensional reference value that is the location defined by the reference values of the individual VCs.

There is much more to be said about Collective control, but this is probably enough to show how complex it may become if you follow the details in social situations with anywhere from tens to millions or even billions of people, or trillions of bacteria. Any reasonable analysis of a situation in which multiple controllers interact in a common environment will have to be simplified, and that simplification is likely to follow the Powers lead, in his simplification of billions of neurons each equipped with thousands of synapses, into “bundles” simplified into “wires” that carry “neural currents”.