The ADRIA team contributes to the development of new approaches to knowledge representation, reasoning and decision in Artificial intelligence.
Formal frameworks are proposed and studied by ADRIA for reasoning, decision, argumentation and learning.
The addressed problems rely on basic formal tools like: possibility and imprecise probability theories for the treatment of epistemic uncertainty; logical or graphical models for the representation of preferences (VCSP, CP-Nets, GAI-Nets, …); non-monotonic, weighted, or fuzzy logics intended to bypass the inadequacy of classical logical inference at capturing specific features of human reasoning.
Current work deals with: The links between (generalized) possibilistic logic, modal logics, multivalued logics and logic programming. The revision and the fusion of uncertain and partially conflicting pieces of information coming from multiple sources. Commonsense reasoning based on analogical proportions.
We focus on axiomatic foundations of decision rules in the qualitative setting, as well as logical and computational tools for multicriteria and/or collective decision and planning problems. Their highly combinatorial nature requires the search for tractable subclasses.
We develop formal models of argumentation and their applications to reasoning, explaining decisions, or modeling dialogues (for negotiation purposes especially).
ADRIA is interested in various aspects of artificial learning (analogy, preference learning, formal concept analysis), and their adaptation to uncertainty frameworks but also to its connection to human learning (à la Piaget). This methodological work is generally carried out in connection to applications, like scheduling, diagnosis, risk analysis, optimization and product configuration.
(see ADRIA website for more information)
This section doesn't contain any article.