Prof Jürg Kohlas

Professor emeritus of Theoretical Computer Science

Research Interests

Algebraic Theory of Information

Information is related to questions, it must be focused on the questions of interest, and it comes often in different pieces, or from different sources and must aggregated. This can be captured in two-sorted algebras, linking elements representing pieces of information and elements representing distinct questions. In this research project, these algebras are studied based on different, but related axiomatic bases. These algebras allow for generic distributed local computation schemes. Different computational architectures are developed and implemented. Further representations of information based either on universal logic systems or on general relational systems are studied. Finally, uncertain information is modeled by random variables taking values in information algebras.

Probabilistic Argumentation

Often information depends on uncertain assumptions. Hypotheses may then be deduced or proved assuming certain of these assumptions. By considering the likelihood or probability of the assumptions involved, the reliability with which the hypotheses can be deduced or proved can be computed. This defines degrees of support for the hypotheses. These degrees form belief functions in the sense of Dempster-Shafer theory of evidence. These probabilistic argumentation systems link logic and probability for inference in a very natural way going back to the founder of stochastics, J. Bernoulli. Computational architectures for probabilistic argumentation systems have been implemented in the framework of propositional logic. Further, in the abstract setting of information algebras they are induced by random variables with value in information algebras. In this direction the project is pursued under the first heading.