This survey article examines the nature of expert systems, describes conventional approaches to expert systems construction, discusses contexts in which expert systems are and are not appropriate, provides private and public sector examples of expert systems, and inventories various types of expert systems development tools. In addition, separate sections examine the special topics of inductive expert systems and of neural networks. It is concluded that expert system software could become an important new class of tools for social scientists. Keywords expert systems, neural networks, artificial intelligence, modeling, social science. Expert systems and artificial intelligence are often dated to 1938, when Alan Turing, a British mathematician, set forth a model of machine imitation of human intelligence. Later, around 1950, the Turing test was proposed as a method for calculating the success of such models: that 70% of human interrogators communicating blind with a computer and with another human for five minutes would not be able to tell the difference (Denning, 1986). By 1955 Herbert Simon and Allen Newell at Carnegie-Mellon had programmed a computer to produce logic proofs (Simon, 1969;1960). In 1975 one of the first commercials applications appeared, Intellect, a natural language product invented by Dartmouth University professor Larry Harris and marketed through his firm, Alcorp, later a major expert systems vendor for IBM mainframes.Since then artificial intelligence researchers have further explored problem solving, pattern recognition, cognition, the use of natural language by computers , as well as new procedures for information classification and analysis (Bolduc, 1989). Of these, by far the most practical from an applications viewpoint has been the genre of software called expert systems. The United States, Europe, and Japan have all undertaken major research and development programs in expert systems, which is changing the way management information systems (Mis) is taught (Yen & Tang, 1988 p. i65/.