For the 35 or so years that I have been involved with PCT objections to the word “control” have recurred. Emotional discomfort with controlling others and being controlled is a bump in the road for many people.
The Control Systems Group started as a group within the cybernetics world, and in the first years the term of art was simply control theory, CT. Kent introduced the more explicit phrase perceptual control theory, PCT.
PCT is the theory of agency. The control systems that PCT describes are agents. The control systems that engineering control theory describes have no agency, the operator of the ‘plant’ is the agent.
I don’t propose a name change. There would be a lot of push-back because “PCT”, “IAPCT”, etc. are collectively controlled variables with mutual influences between them and other CVs both private and collective. I don’t think we’re about to become the IAAT.
Though we can say “PCT is where it’s AT”.
And talk about agency can be very helpful in communicating what PCT is about. PCT is the science of agency; there is no other.
I think this is a great phrase! It encodes that PCT is about agency and has a meme-like resonance.
For what it’s worth, spreading the ideas of Powers and his disciples in the current digital environment is going to require these bite-size chunks of PCT to be disseminated. Perhaps some sort of IAPCT outreach should include a program to meme-ify Perceptual Control Theory.
I think a far worse “bump” is the word “perceptual”, because it leads people who have made it past the “control” word bump – people who “accept” PCT – to imagine that what distinguishes PCT from engineering control theory is that it is about how the behavior of living control systems is based on “fallible” perceptions. But this is not what distinguishes PCT from engineering control theory. Indeed, PCT is the same as engineering control theory.
What distinguishes Powers’ version of control theory from engineering control theory is its point of view. Powers’ theory is called “perceptual” control theory to emphasize the fact that the theory is applied from the point of view of the system being modeled. From this point of view it can be seen that living systems control their perceptual inputs, not (as has appeared to be the case for centuries) their motor outputs.
Apparently, you have not yet taken a ride in a Waymo! I have and I don’t see any difference in agency (whatever that is) between a Waymo driver and human driver (the car being the plant in both cases) except, perhaps, that the Waymo is a bit more skillful;-)
I think PCT is the science of the CONTROLLING done by living systems. Control is a well understood and perfectly objective phenomenon. Understanding control is absolutely essential to understanding PCT. I know that some people are put off by the word “control” but there’s really nothing you can do about it. If agency is just another word for control and people are less afraid of it then I’m fine with it. But they are still going to have a tough time when the cat is inevitably let out of the bag and the phenomenon called “agency” is revealed to point to the phenomenon of control.
I think the best approach is Tim Carey’s. Give a full throated explanation of what an amazing and wonderful thing control is. But like all wonderful things it can be abused. PCT explains the good, the bad and the essential of the controlling done by living systems.
As the science of agency, PCT also has a lot to offer in terms of the traditionally binary free will vs determinism debate by offering a model where agency exists through control - not by choosing from an infinite number of options or by denying interdependency but by actively controlling our perceptions through the agency of behavior.
What is “agency”? How do we know that it exists? How do we observe it?
In PCT, we know what control is. We know it exists because it is easily observed. It is seen when a variable is maintained in a fixed or variable reference state, protected from disturbance. Control can be seen in this demo. The distance between cursor and target is the variable that is kept in a fixed reference state (~0 distance). That is, the distance between cursor and target is controlled; it is a controlled variable.
Is there a comparable way of demonstrating the existence of agency?
You say that agency exists by “actively controlling our perceptions through the agency of behavior”. This suggests that agency is what, in PCT, we call the “output” component of a control loop. So PCT would say that control exists through what you call agency rather than agency exists through control.
I think I’m going to stick with calling Powers’ theory Perceptual Control Theory rather than Agency Theory. PCT communicates everything that is important about Powers theory: People are control systems and what they control is their perceptual input. The rest is detail;-)
That is very true, I think. From PCT I would say that we (all living organisms) are free to control - or at least try to control - what ever perceptions for which we happen to have a control unit (input-comparator-output). Whether the control will realize successfully depends on the relationship between output and the feedback function and the disturbance. If we cannot manage to control a perception we are free to reorganize and acquire better methods of output. As animals and especially human beings we are also free to use our consciousness to direct the reorganization to right places in the whole control hierarchy for to make process quicker and more effective. (Perhaps we are also free to think and choose consciously different alternative methods of output.)
So our freedom is real but always limited and relative to the situation and our history.
This is, I think, an important topic to discuss and develop.
I think agency is distinct from control and is what distinguishes human beings from animals. Humans and animals are both control systems where control is easily observable. However, animals lack agency. Agency is more akin to the system concepts level of control (and reorganization), which could also be called consciousness.
In PCT, agency is not directly observable in the way that control is, but it can be inferred from patterns of control - especially across higher levels of the control hierarchy, specifically at the highest level of system concepts.
Control is observable because we can see what variable a system is keeping stable and how behavior changes to keep that variable constant.
Behavior is the control of perception and that’s testable.
Agency in PCT is an emergent property of a reorganized, hierarchical system that sets its own references at high levels, consciously resolves internal conflicts, adapts over time, and initiates shifts in goal structures (like changing life directions).
So you don’t see agency the way you see a hand move. You infer it when a person reflects and changes a long-held belief, someone acts against short-term urges for higher-level values, conflicting goals are resolved internally, or reorganization occurs to improve functioning.
These patterns point to self-directed reconfiguration, not simply the random changes often associated with the organization system - and that’s where agency appears as a phenomenon in PCT.
Agency is the functional outcome of a system that controls and manages its own references at multiple levels and can reorganize itself in response to internal error. In that sense, agency is an emergent, inferable, dynamic phenomenon - not an illusion, but not a “thing” you can point at directly either.
I think this is precisely the function of the system concepts level of control and is what sets human beings apart from animals. I theorize that animals lack a system concepts level of control - that is, we could say they lack consciousness. They are aware, complex control systems but they lack the conscious control of human beings. Animals’ highest level of control is the principles level of control, which I also theorize could be called ego - the reference which seeks the minimization of total system error, wanting perceptions to be as desired.
The system concepts level of control gives human beings the capacity just as you said to “think and choose consciously different alternative methods of output”. It also provides the capacity to consciously manage and change references, reference levels, hierarchy of references, etc - it’s ultimately the capacity for conscious reorganization, not simply reorganization based on random system changes.
I couldn’t agree more that this is “an important topic to discuss and develop”.
It’s true, I’ve never taken a ride in a Waymo vehicle.
The description tells me that it’s using GPS and Google maps or the equivalent, as a human driver would, and it’s navigating lanes and turns, merging in and out of traffic, obeying traffic signals, avoiding pedestrians and other vehicles, etc. as a human driver would. Sounds like PCT would describe a Waymo vehicle as a control system which (except for the passenger destination) sets and varies its own reference values autonomously much as human driver (a living control system) would. That’s not how engineering control theory describes a control system.
If it’s using Friston’s FEP I wouldn’t ride in it. Any cab ride is probabilistic enough already, thank you.
Yes, the engineer designs a system to produce outputs that the engineer (or the customer) wants, products or behavioral actions or a combination.
The engineer designs systems to CONTROL variables that the engineer wants controlled. In the case of the Waymo, the engineer is designing a system that controls the variables that a human driver has to control. Therefore, I think that we have much to learn from Waymo engineers about the perceptual variables that drivers control in order to get around town successfully. The control systems built by Waymo engineers are surely not implemented in the same way those systems are implemented in humans. But since the Waymos are driving around town quite successfully, they are surely controlling perceptual variables that are equivalent to the ones controlled by humans.