The artificial intelligence (AI) discipline of machine learning offers the best opportunity for alleviating the critical problem of acquiring the knowledge base necessary for expert systems. This paper examines the characteristics of such tasks and identifies a number of weaknesses with several dominant AI approaches. Genetic algorithms (GAs) are a probabilistic search technique based on the adaptive efficiency of natural organisms and offer an alternative which addresses the weaknesses in conventional methods. This paper describes the implementation of ADAM, a GA driven classifier, and compares the quality of the rules it generates to those of alternative induction techniques on a simulated decision problem.