Artificial Intelligence and the Empiricist Picture of Thought

Abstract

The empiricist picture of human thought and language tends to guide people’s ideas about the possible roles of computers in society. Our understanding of the world is thought to be drawn from the objects we perceive and learn about, such as colours, shapes, “empirical regularities”, and so on. Our actions are subject to “rational reflection” and are shaped by our factual beliefs and by our desires and decisions. This is a simplistic view. However, it is one that numerous people adopt. Some modify it by presenting a dualist argument: this picture, it is thought, may explain part of human experience, but it fails to deal with deeper, more intangible aspects. The author argues, however, that the empiricist picture contains a deeper error: the priority given to theory over practice. This must in fact be reversed so that theoretical understanding is seen to be conceivable against a background of practical understanding. Where artificial intelligence is concerned, the empiricist picture seems to pose small problems. This is because thinking is assumed to consist merely in the processing of information, which is regarded as strings of symbols. Once it is seen that theoretical understanding is only possible in a context of practical life, however, it is clear that human thought cannot be understood as the processing of information. The difference between human and artificial intelligence becomes evident when we consider what it means to make an error. An error made by a person may be made intelligible by regarding it in the context of his perception of things, his life and culture. An error made by a computer is not intelligible: the computer has no life into which its judgements may fit or fail to fit.