[From Bruce Nevin (2017.12.23.11:00 ET)]
Bruce Nevin (2017.12.21.08:46 ET) [Subject was: Watching your p’s and q.i’s (was Re: Kenneth J. W. Craik on levels of perception and control)] –
A little more about points of view.
We don’t know in detail how evolution, embryology, and reorganization have implemented control loops and the control hierarchy in our bodies, although Henry is making progress with simpler mammals. From the wetware point of view, our control diagrams are vastly simplified schematics. The synapsing of axons and dendrites is rather more complex, not to mention the neurochemical soup (cf. The War of the Soups and the Sparks).
Our conventional diagrams are fully adequate for modeling control as an observed phenomenon. For that, the physiological and psychophysical details are not needed. They do not even register in the perceptual input functions for controlling the crafting of PCT models with software. From the modeling point of view they are irrelevant, distracting, and dwelling on them may even become annoying–a disturbance.
Conversely, for a person interested in how the actual neural connections may function to achieve what is schematically represented in control diagrams, the waving away of these matters may well be a disturbance.
Computer simulations have been a central achievement and demonstration of PCT. The core methodology of PCT is understood to be: to identify controlled variables, quantify them, measure disturbances and outputs and infer reference values in commensurate terms, and then to verify all this by constructing a working simulation whose numerical inputs and outputs accord closely with the measured values. This has the crystalline appeal of logic and mathematics.
How does an interest in the messiness of neurophysiology and neurochemistry find a place in this canonical methodology of PCT? What is required for them to be in the same discussion without quarrel?
Another matter now. Quantification is central to our core methodology. But some of us work in fields where it is not obvious how to represent an individual’s controlled perceptions (and disturbances to them) by numerical quantities. This is because, physiologically, the quantification is done by the lowest-level input functions, the receptors in various sensory modalities, but the perceptions being investigated are at higher levels. To model quantitatively something like the control of word choice in a sentence, for example, would seem to require modeling numerous parallel lines of control all the way down to Intensity inputs; not to mention modeling memory and imagination far more comprehensively than has ever been done in any PCT simulation so far. Is it possible (would it be legitimate!) to finesse the problem by providing inputs to higher levels without actually modeling the lower levels that generate them? What then would be the basis or principle for quantification?
Add to this that in many fields–certainly for language–the controlled perceptions and the reference values for them are subject to collective control. We can adopt the simplifying assumption that e.g. parties to a dialog speak the same dialect of the same language, but we cannot ignore that engaging in dialog, understanding the other, and establishing agreements (or not) are matters of collective control. Kent has modeled collective control with great success, but limited to easily quantified variables or to abstract quantities without ‘real-world’ referents. How greatly will adding the above factors of memory and imagination complexify research into collective control?
A number of fields face these challenges as they come to PCT. I suppose collectively we could be said to represent another point of view, albeit rather inchoate, and one not as likely to result in misunderstanding and quarrel as the above. I think we tend to be rather more meek and unpretentious about our PCT chops. There’s no good response when we’re told how essential quantification and modeling are to doing ‘real PCT research’, with the implicit (or explicit) scolding at what dilletante sluggards we are. Perhaps it is only that I lack imagination and initiative, but these challenges have seemed daunting to me, and that is why I have asked Rick and Bruce (Abbott) to apply their expertise and experience to articulating the methodology of PCT more comprehensively.
Another point of view takes quantification in the direction of analyzing the systemic properties of control systems as mathematical objects or in information-theoretic terms. This may turn out to be important for understanding reorganization, the evolution of control systems, their limitations, and the perhaps as yet unrealized capacities of control systems, e.g. when they are technologically supported or augmented. Martin has been pioneering this aspect.
Are there other points of view or avenues of engagement in PCT that I have omitted?
If any of these points of view is beyond the pale of legitimate PCT, now would be an optimum time to make that case. Failing that, it behooves each of us to consider that true statements may be put in different words, and to distinguish genuine contradictions from differences of interest, emphasis, and quantifiability.