Abschlussarbeit

Beschreibung:

Revision is the kind of belief change in which a new belief is incorporated into the initial beliefs such that the new information is prioritized and inconsistencies are solved whenever possible [1]. In the following, we presuppose that belief revision is a fundamental operation of human thought. There are multiple ways of doing belief revision, and we want to analyze what kind of revision predicts performs best to predict human reasoning [2].

The dataset for this thesis comes from an experiment on human propositional reasoning. In the experiment, first, a sequence of propositional sentences is shown to the participants. Second, the participants decide what is entailed (of a set of presented options) by all the propositional sentences. This thesis uses a formal model of the experiment. Each presentation of a proposition in the first phase of the experiment is modelled by a revision. The second phase of the experiment, the decision of which proposition is entailed, is modelled by logical entailment of the belief state after the last revision. For the thesis, the formal model shall be implemented. An existing code base can be used and extended for implementing this approach. Finally, the implementation shall be used to evaluate which kind of revision operation predicts best the answers in the second part of the experiment. For more details on this topic, please contact Kai Sauerwald.

• [1] Peter Gärdenfors, Hans Rott: Belief revision. Handbook of logic in artificial intelligence and logic programming (Vol. 4): epistemic and temporal reasoning, Oxford University Press, 1995, p. 35-132
• [2] Marco Ragni, Gabriele Kern-Isberner, Christoph Beierle, Kai Sauerwald:
  Cognitive Logics - Features, Formalisms, and Challenges. ECAI 2020, p. 2931-2932