Premium
A NOTE ON A MACHINE THAT ‘LEARNS’ RULES
Author(s) -
SHOTTER JOHN
Publication year - 1968
Publication title -
british journal of psychology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.536
H-Index - 92
eISSN - 2044-8295
pISSN - 0007-1269
DOI - 10.1111/j.2044-8295.1968.tb01130.x
Subject(s) - function (biology) , set (abstract data type) , task (project management) , psychology , perception , carry (investment) , organism , reinforcement learning , cognitive psychology , cognitive science , reinforcement , restructuring , artificial intelligence , social psychology , computer science , neuroscience , paleontology , management , finance , evolutionary biology , economics , biology , programming language
A machine (C) is described which will learn to mimic creatively the behaviour of another machine (M) in this sense: after a certain number of ‘observations’ of M, C will begin to produce sequences of symbols not only like those that M produced but also like those that M might have produced but did not in the circumstances do so. If we say that M's behaviour is as if it were regulated by rules then we could say that C has learnt them. For human beings, postulation of an ability of this type seems to be necessary to account for the learning of language and for other social activities where behaviour is regulated by conventions. ‘Reinforcement theory’ is unable to account for this type of learning. C carries out its task without any ‘feedback’ or ‘reward’ from M; indeed, the only function for ‘motivational’ factors is to direct C's ‘attention’ from observing M to observing another machine or to cut off C's input entirely. If C were an organism it would need a set of primitive perceptual categories and an innately active nervous system to carry out certain restructuring operations on the knowledge that it acquires.