It is generally recognized that building systems that discover new knowledge is a challenging enterprise which needs interdisciplinary approaches. Over the last two decades, this sub-area has involved various groups including Researchers from Arti"cial Intelligence/Machine Learning, Cognitive Psychology, History & Philosophy of Science & Sociology. The papers included in this special issue re#ect this interdisciplinary perspective and include a paper by Cognitive Psychologists, Baker & Dunbar who have studied how Biological Scientists design typical experiments, as well as a number of papers which describe Machine Discovery Systems.
The papers in this issue on Machine Discovery Systems can be divided into two groups. The "rst reports systems which have recently been developed (the papers by Kocabas & Langley and ValdeH s-PeH rez, Pericliev & Pereira) and the remaining two papers which review existing systems (the papers by Colton, Bundy & Walsh and that by Langley). Kocabas & Langley's paper describes the ASTRA system which creates pathways (sets of reactions) to explain how some elements might be created from lighter ones in the centre of stars. The task addressed by ValdeH s-PeH rez, Pericliev and Pereira's paper is to "nd minimal discriminations among (several or many) classes. In this paper, they apply their approach to kinship relationships, pro"ling children with brain lesions, determining which metallic catalyst should be used with particular (chemical) reactions, and detecting patterns of proteins within cells.
The paper by Colton et al. reviews "ve systems: AM, GT, the system by Bagai et al. & HR which have discovered Mathematical concepts, as well as the Gra$ti system which creates conjectures using given concepts; many of these systems have lead to publications in the Mathematics literature. The authors review each of the systems under the following headings: plausibility, novelty, surprisingness, applicability, comprehensibility, and utility of the concepts proposed (ValdeH s-PeH rez, 1999). Similarly, Langley reviews machine discovery systems which have been applied to Scienti"c/Technical tasks