Exchanging information among cognitive agents in collaborative environments

Cooperative design of complex artifacts always involves several specialists who necessarily have different technical cultures. Multi-agent systems mixing artificial and human agents are a key technology for supporting collaborative work. The paper first includes a brief description of such a supporting architecture, emphasizing the role of personal assistant agents. Personal assistant agents engage in dialogs using natural language preferably with vocal interaction. Implementing such features calls for complex programming when agents do not share the same ontology and humans may not speak the same language. In a second part of the paper we show how the ontologies and knowledge bases can be used to process messages containing natural language data, language information and patterns. We indicate how this approach can be implemented leading to simpler programming. The last part of the paper discusses the limits of the proposed approach, gives an example of interaction and indicates how it could also be used to exchange concepts, extending an agent ontology. The paper however does not claim that the resulting dialogs are better than in other approaches, but intends to show that developing complex human-agent interaction does not necessarily require complex programming.