ILIKS SCIENTIFIC PROGRAMME
The ILIKS scientific programme for modelling interacting knowledge systems is organized in complementary topics. The basics of interaction, i.e., the notion of cognitive agent, the social relationships between agents, and the notions of action and plan, are addressed in the first three topics. Interaction modelling proper, together with the important special case of communicative interaction constitutes the fourth topic. The last point deals with theories and tools for modelling the contents of interaction, a crucial issue for the efficiency of interacting systems.
- Agents and mental attitudes.
Modelling cognitive agents, artificial or not, is addressed by comparing and establishing synergies between cognitive, social, philosophical and logic approaches. Work in this topic aims at understanding the nature of agents' mind and mental attitudes and the rules that control their coherence and dynamics. Specifically, this addresses: the ontology of mind, mental states, and agency; logics of belief and knowledge and the problems of belief revision; models of agents intentions, goals, plans and commitments. This last item is studied jointly with the notions of action and plan under topic #3 below.
- Social relationships and organizations.
Beyond the individual agent and its private mental states, we address the social dimension and the notion of collective agent, here again combining several approaches in several disciplines. Ontology of social reality is based on a number of social relationships creating links and dependencies between agents, such as trust, responsibility, obligation, delegation, or cooperativity, as well as on the very notions of collective agent and organization. This last point includes the study of
we-attitudes (e.g., mutual beliefs, collective intentions), collective action and collective decision, and the characterization of various kinds of social groups, institutions, organizations and legal entities, together with the specification of roles to be played within such groups. The modelling of social norms and other conventions regulating relations between single agents, collective agents and organizations, for instance in terms of deontic logics, also belong to this topic. All these notions lay the foundations for the characterization of more complex ones such as that of security within a society of interactive agents. We are particularly interested in techniques for the representation and verification of complex systems involving these notions. Typical applications are diagnosis and verification of multi-agent systems.
- Actions and Plans.
As agency is based on the capacity of agents to perform actions, representing and reasoning on actions is necessary. Actions, goals and intentions to act are tightly related, therefore this topic is developed in synergy with the first two. Work in this topic starts with an ontological study of actions, a special kind of eventualities with a special kind of participation relation between agents and actions. The ontology of action provides in particular an analysis of epistemic actions, such as speech acts and other communicative acts, and characterizes the various composition modes of single actions into collective actions, complex actions and action sequences or plans. Existing and new logics of actions are developed, and their semantic aspects ontologically analyzed. The reasoning dimension of plans is studied as well, with work on goal dependencies and goal conflicts, on planning in multi-agent environment with epistemic actions, and planning with model-checking techniques under partial observability and graded beliefs. This topic finally addresses the formal and algorithmic study of the off-line versus on-line trade-off, i.e., interleaving planning and execution, for knowledge-based plans.
- Interaction and Communication
Formal models of interaction are developed on the basis of the above fundamental work. The interplay between logics of actions and logics of mental attitudes yields a logical characterization of speech acts. Studies of the structure and coherence of interaction are also essential for this purpose; such theories have been developed separately in a variety of disciplines. With the help of a formal ontology of interaction, an essential step is to compare them; then, their results can be integrated where possible. Theories of the rhetorical structure of texts, conversations and multi-modal documents are being reconsidered and enriched. Dialogue genres, among which argumentation, negotiation and persuasion, which are modelled with dialogue logics in formal dialectics, another discipline studying the structure of communicative interaction, are also ontologically characterized. Similarly, argumentation theory and other formal models of communication dealing with either linguistic interaction or implicit communication, are ontologically analysed, compared, and integrated with other approaches.
- Lexicon, ontologies, semantic interoperability and information extraction.
Communicative interaction is effective only when mutual understanding of the contents of the information exchanged is assured. The
Semantic Web has imposed the use of computational lexicons and
ontologies (more than often limited to taxonomies), but system interoperability remains problematic. The last topic in the ILIKS research program thus first addresses the relationship between lexical semantics and formal ontology, for well-founded computational lexicons as well as for semantic negotiation and coordination. In parallel to formal ontology methods, techniques of ontology learning from texts and terminology extraction are used to automatically build consensus ontologies. Interoperability is then improved using ontology-based techniques of query reformulation, document annotation and categorization, and document collection browsing. Finally, social and economic models of the dynamics of ontologies and contexts during interaction are developed, thus contributing also to the study of the models of interaction of the previous topic.