September 19, 2017

Learning

Appendix 7 of The Stable Society: its structure and control: Towards a Social Cybernetics, Wadebridge Ecological Centre, UK, 1978

In line with empiricist theory, the tendency has been to explain behaviour in terms of the influence of the environment, i.e. as responses dictated by external stimuli, and there has been a corresponding neglect of the influence of the inherited set of instructions or innate releasing mechanisms (I.R.M.), as Lorenz calls them, that must determine the generalities of a behaviour pattern.

An organism is not born with a blank mind—the tabula rasa of the empiricists—but with a general set of instructions and a corresponding general model reflecting its evolutionary experience, which will slowly be differentiated so as to adapt them to the requirements of its specific environment.

This fact is apparent from a number of experiments on different types of animals. Fantz 151 experimented on a sample of 1,000 chicks. He found that they pecked 100 times more often at sphere-shaped objects than at pyramidal-shaped ones. The chicks were hatched in darkness and tested on their first exposure to light, from which it must follow that this behaviour can only be explained on the basis of innate tendencies. Tinbergen 23 found that newly-hatched herring gulls preferred pecking at objects which resembled the bill of the parent from which they were fed. Marked preferences for certain specific objects were also established among baby chimps. Fantz conducted similar experiments with children. In one of these, 49 children, aged from 4 days to 6 months, were presented with:

“. . . three flat objects, the size and shape of a head. On one we painted a stylized face in black on a pink background, on the second we rearranged the features in a scrambled pattern, and on the third we painted a solid patch of black at one end, with an area equal to that covered by the features . . . The results were about the same for all age levels: the infants looked most often at the real face, looked slightly less often at the scrambled face, and largely ignored the control pattern.”151

From such experiments, Fantz concludes:

“Lowly chicks as well a lofty primates perceive and respond to form without experience if given the opportunity at the appropriate age of development. Innate knowledge of the environment is demonstrated by the preference of newly-hatched chicks for forms likely to be edible and by the interest of young infants in kinds of form that will later aid in object recognition, social responsiveness and spatial orientation. This primitive knowledge provides a foundation for the vast accumulation of knowledge through experience.”

The modifications brought about to this general model are normally referred to as ‘learning’. Its role is not to alter innate behavioural tendencies so much as to enable them to be satisfied with ever-greater precision.

Thus an embryo will develop not merely because its environmental conditions are correct, but also because it contains a complex set of general instructions, that, as a result of interaction with its environment, will be slowly differentiated.

One of the consequences of adopting this model of behaviour is that the notion of learning by trial and error is no longer tenable. Thus, if we put a rat into a maze, we know that it can be taught to find its way out. However, before hitting upon the correct route, it will have to make a series of unsuccessful trials. Now, would it be possible to put order into this series, or must each trial be considered merely a random one? The trial-and-error theory appears to favour the latter hypothesis. However, it can be shown that random behaviour is not a scientific concept. Where there is order, there must be instructions, and a corresponding organisation of information, i.e. a model.

Looked at slightly differently, if the rat is supposed to find its way through the maze, what will determine its first trial? If a hundred possible moves are open to it, why should it start off by making one rather than any other? If the move it makes is haphazard, why should not all the others be haphazard, too? Is a rat behaving in a haphazard way when it chooses to eat a piece of cheese rather than an iron nail, or when it feeds its family rather than that of some other rat? If not, which actions are to be considered haphazard and which are not?

The answer is that the notion of non-directive haphazard trial-and-error learning is irreconcilable with our basic knowledge of behaviour. Instead, one must regard each action as based on what, in the light of a systemic model, constitutes the most probable hypothesis. This hypothesis will have a higher probability of being correct, i.e. of leading to adaptive behaviour, the greater is the organisation of the model on which it is based. Thus it will be higher in the case of the action undertaken by a man than in that undertaken by a rat, which in turn will be higher than that undertaken by an earthworm. Nevertheless, all the actions of the man, the rat and the earthworm, as Craik 13 was the first to show, must be regarded as based on hypotheses as to the nature of their environment, or rather, of their relationship with their environment.

Supposing that the first move made by the rat, rather than lead it towards an opening, on the contrary led it to a further cul-de-sac, the model responsible for this error would have to be modified. Thus, when the rat made its next move, the environmental situation would be interpreted in the light of a new model—one that had taken into account the failure of its predecessor to interpret correctly the short-term environmental situation in which the rat had found itself, and one furnishing an ever-more precise representation of the system for the purpose of finding its way out of the maze. In other words, each action could be regarded as a correction of an error, and if we regarded these actions as forming a series, the latter would be ‘damped’, in the sense that the errors were being progressively reduced, i.e. the system would be tending towards ever-greater stability.

References

13. Kenneth Craik, The Nature of Explanation Cambridge University Press, Cambridge, 1952.

23. N. Tinbergen, The Study of Instinct The Clarendon Press, Oxford, 1951.

151. Robert L. Fantz, ‘The Origin of Form Perception’ in Scientific American May, 1962.

  • Twitter
  • Facebook
  • Digg
  • Reddit
  • StumbleUpon
  • Diaspora
  • Identi.ca
  • email
  • Add to favorites
Back to top