I will present an algorithm for learning nonmonotonic
logic programs, ie. logic programs that contain negative
literals in the body of the rules. The method is parametric
on a classical learning algorithm (one that learns definite
logic programs) whose generated rules are to be understood
as default rules. This means that these rules must be
tolerant to the negative examples by allowing for the
possibility of exceptions. The same classical algorithm
is then used, recursively, to learn these exceptions.