Abduction and Induction in Artificial Intelligence

Abduction and induction are two forms of reasoning with incomplete information that are appropriate for many tasks in Artificial Intelligence (AI). To a first approximation, abduction reasons from effects to possible causes, while induction learns general rules to account for particular observations. As our logical and philosophical understanding of these two forms of reasoning continues to grow, and our computational techniques are steadily improved, evidence accumulates that there are potentially significant benefits to be gained from integrating abduction and induction, in a cooperative way, within an incremental cycle of knowledge development. Different conceptual models for integrating abduction and induction are emerging which seem particularly promising for application to scientific modelling such as in the area of Systems Biology. Recent work in Machine Learning and Scientific Discovery is now demonstrating the viability of such an integrated approach by providing the first tentative applications of frameworks and systems designed to exploit the combination of abductive and inductive reasoning. This special issue reports on such recent developments focusing on expressive methodologies for logic-based abduction and induction. The volume grew out of the AIAI'07 workshop on Abduction and Induction in Artificial Intelligence and Bioinformatics, which took place on September 15th, 2007 in Aix-en-Provence, France. This workshop followed the series of workshops on abduction and induction that took place in the 1990s, culminating in the edited volume Abduction and Induction: Essays on their Relation and Integration (Kluwer, 2000); and their recent revival in the two workshops at AIAI'05 and ECAI'06. The AIAI'07 workshop was co-located with the First Franco-Japanese Symposium on Knowledge Discovery in Systems Biology in order to further encourage the link between theoretical advances and applications of scientific modelling and development. From the workshop proceedings seven papers were selected for further expansion and review. Three of the papers focus on the application side of things: one in the area of modelling signalling networks in cell biology (by Nam Tran and Chitta Baral); the second in elucidating requirements specifications in the area of software engineering (Dalal Alrajeh, Oliver Ray, Alessandra Russo and Sebastian Uchitel); and the third in distributed agent cooperation in the area of multi-agent systems (Gauvain Bourgne, Amal El Fallah Seghrouchni and Nicolas Maudet). Three further papers focus more on theoretical issues. One re-examines different forms of induction and their possible unification in a common framework (Koji Iwanuma, Katsumi Inoue and Hidetomo Nabeshima), while the second paper examines equivalence issues in abduction and induction (Chiaki Sakama and Katsumi Inoue), and the third by Oliver Ray presents an extension of the Hybrid Abductive and Inductive Learning framework to nonmonotonic inductive logic programming, showing also how it can be implemented and applied. The final paper by Henning Christiansen is concerned with the computational and practical aspects of hypothesis-based reasoning. We would like to thank all the participants of the AIAI workshops for their interest in this research topic. The AIAI'07 workshop was generously sponsored by the Laboratory for the Analysis and Architecture of Systems at the French National Centre for Scientific Research and the Japanese National Institute of Informatics, as part of the Franco-Japanese collaboration on Knowledge-based Discovery in Systems Biology. We would also like to thank all authors for submitting their papers and working with us on this special issue. A special thanks goes to Oliver Ray who helped us revive the Abduction and Induction workshops in this new series of AIAI workshops and took care of most of the organizational issues of the AIAI'07 workshop.