Human deduction usually does not follow the rules of classical logic. Reasons may be incomplete knowledge, incorrect assumptions or inconsistent norms. Artificial intelligence research aspired from the beginning to implement rationality or mechanisms for rationality in AI-systems. In this context, rationality cannot be limited to cognitive tasks, but must include complex behavior and the interaction with other subjects and the physical environment.

The RatioLog project aims at establishing a common model for deduction and behavior. To this end logical deduction and modeling continuous systems are to be combined, based on preceding work on non-monotonous calculi and hybrid automata. Deduction in classical logic is to be extended with several non-monotonous aspects, for example abduction or refutable argumentation. The extensions will not only be made on a theoretical level, but will also be implemented into the automated theorem prover E-KR-Hyper. LogAnswer, an open-domain question answering system that uses E-KRHyper and Wikipedia to answer natural-language questions (in German), will be expanded to a rational question answering system, providing an excellent testing field for evaluating the rational deduction.

RatioLog is a project by the working group of Prof. Frieder Stolzenburg at the Hochschule Harz and the Working Group Artificial Intelligence of Prof. Ulrich Furbach at the University Koblenz-Landau. The project is supported by the Deutsche Forschungsgemeinschaft (DFG)