Saturday, 4 February 2017

Ashby's Experimental Method in "Design for a Brain"

Ashby's Design for a Brain is a remarkable book containing a lot of detail about how Ashby saw cyberentics as a science. Some of the most interesting passages concern his defence of his own methodology as he sought to create a "mechanical brain", building on his earlier work with homeostats. In contrast to much second-order cybernetics, Ashby remains a down-to-earth and practical scientist. But he is looking at the world in a different way: he is exploring constraints and relations rather than causation. The principal relation which concerns him is the relation between the experimenter and the experimental situation. In this, of course, he is very close to second-order cybernetics - but with a penetration of analytical thought which has been unfortunately overlooked by many of his cybernetic colleagues.
"It will be appreciated that every real 'machine' embodies no less than an infinite number of variables, most of which must of necessity be ignored. Thus if we were studying the swing of a pendulum in relation to its length we would be interested in its angular deviation at various times, but we would often ignore the chemical composition of the bob, the reflecting power of its surface, the electric conductivity of the suspending string, the specific gravity of the bob, its shape, the age of the alloy, its degree of bacterial contamination, and so on. The list of what might be ignored could be extended indefinitely. Faced with this infinite number of variables, the experimenter must, and of course does, select a definite number for examination - in other words, he defines the system." (Design for a Brain, pp15-16).

Ashby goes on to describe specific examples of empirical practice:

"In chemical dynamics the variables  are often the concentrations of substances. Selected concentrations are brought together, and from a definite moment are allowed to interact while the temperature is held constant. The experimenter records the changes which the concentrations undergo with time.
In experimental psychology, the variables might be "the number of mistakes made by a rat on a trial in a maze" and "the amount of cerebral cortex which has been removed surgically" [ugh!]. The second variable is permanently under the experimenter's control. The experimenter starts the experiment and observes how the first variable changes with time while the second variable is held constant, or caused to change in some prescribed manner.
While a single primary operation may seem to yield little information, the power of the method lies in the fact that the experimenter can repeat it with variations, and can relate the different responses to the different variations. Thus, after one primary operation the next may be varied in any of three ways the system may be changed by the inclusion of new variables or by the omission of old; the initial state may be changed or the prescribed courses may be changed. By applying these variations systematically, in different patterns and groupings, the different responses may be interrelated to yield relations
By further orderly variations, these relations may be further interrelated to yield secondary, or hyper-relations; and so on. In this way the "machine" may be made to yield more and more complex information about its inner organisation." (pp17-18)
What Ashby was arguing was that the internal relations of a system - its internal constraints - are revealed by applying constraints to their investigation.  ("Relation" and "constraint" Ashby saw as synonymous terms)

As to the detail of what it means to apply constraint, Ashby argues that it is about creating regularity. This is not to suggest that regularities are necessarily real, or external to the observer, but that they arise in the relations between the experimenter and their subject.

"If, on testing, a system is found not to be regular, the experimenter is faced with the common problem of what to do with a system that will not give reproducible results. Somehow he must get regularity. The practical details vary from case to case, but in principle the necessity is always the same: he must try a new system. This means that new variables must be added to the previous set, or, more rarely, some irrelevant variables omitted."
I find "he must try a new system" a very powerful statement. So often in educational theory, psychology - even cybernetics - regularity is assumed at various levels of the system. It is treated as a foundation upon which all other variables are tested. Maturana's autopoietic theory, for example, makes great statements about the need to adapt to regularities in perception: as if regularities are real, the mechnaism of perception is regular too, and what needs to happen is that the two come together in some way. Ashby doesn't say this. If a scientist discovers a regularity, it exists in the relationships of the experimental situation. If they fail to find regularities, they seek (create) another experimental situation. The conclusion one might draw from this is that there are no real regularities beyond a relationship.

Ashby wrestles with the idea of objectivity always conscious that this is one of the fundamental criteria for a commonsense view of the world. He wants to make the distinction between a "natural system" and an "absolute system" and reflects the challenges of making association between the mechanisms of physics and those of biology. He asks:

"both science and common sense insist that if a system is to be studied with profit its variables must have some naturalness of association. But what is natural?"
His definition of naturalness of association is related to his definition of an "absolute system" where the state of a system in entirely dependent on its historical (previous) states.  With a operating definition of an absolute system, he sets the criteria for defining a "natural association". These criteria insist on some alignment to common sense, and some sense of "objectivity". In effect this places the scientist's own reflexivity in the frame: "only experience can show whether it [an idea of a system] is faulty or sound". It is not to deny objectivity, but it is to resituate as a relation, or a constraint.

No comments: