Why GAs are Hard to Use

Genetic Algorithms (GAs) are increasingly used for such purposes as deriving programs 1] and producing designs for robots 2]. According to the building-block hypothesis and schema analysis of Holland 3] the GA is an eecient search method. However, empirical work has shown that in some cases the method is outperformed by simpler processes such as random-permutation hill climbing 4] and 5]. The present paper reexamines Holland's framework (as formulated by Goldberg 6]) and nds that such in-practice failures are eeectively predicted by the schema analysis. The high eeciency of the GA method is commonly attributed to itsìmplicit parallelism'. However, this eeciency is hard to realise because there is a deep contradiction between the building-block hypothesis and the schema theorem.