An AI learning technique that learns (logic) programs given a set of (input, output) examples representing correct behaviour, and also a set of examples showing incorrect behaviour.
Also a book:
The complete text has been made publicly available online by the authors Nada Lavrac and Saso Dzeroski: