This article presents a particular approach to the subfield of artific
ial intelligence referred to as ''aid to scientific discovery.'' The p
rogress made in formal induction techniques suggests that the acquisit
ion and discovery of knowledge may be considered as a refinement cycle
in which the computer can play a central role. Some of the perspectiv
es and the main difficulties are illustrated with a study done on the
discovery of the causes of scurvy, and some links are made to the conc
ept of scientific dogmas.