Artificial experts: social knowledge and intelligent machines

Abstract

A machine is what we make of it. It can mimic us if we can mimic it, or help it out, or overlook its mistakes. In Artificial Experts sociologist Harry Collins explains what computers can't do, but also studies the ordinary and extraordinary things that they can do. He argues that although machines are limited because we cannot reproduce in symbols what every community knows, we give them abilities because of the way we embed them in our society. He unfolds a compelling account of the difference between human action and machine intelligence, the core of which is a witty and learned exploration of knowledge itself, of what communities know and the ways in which they know it. In the course of his investigations, Collins derives enlightening metaphors for the relation between artificial intelligence and prosthetic technologies such as artificial hearts. He provides an intriguing explanation of why pocket calculators work and shares his own experience in constructing an expert system designed to teach people to grow specialized semiconductor crystals. He describes a novel development of the Turing protocol for the definition Of intelligence, a new classification of human skill, and an original way of understanding our relationship to machines. From an AI point of view, the acquisition of knowledge and the selection of applications are critical to the success of expert systems. Collins offers an original approach to both problems for AI researchers and practitioners, providing a sociological perspective on the kinds of knowledge expert systems can and cannot capture, and on the domains that are and are not likely to be successful. Harry M. Collins is a widely published sociologist. He isProfessor of Sociology, Head of the School of Social Sciences, and Director of the Science Studies Centre at the University of Bath.