Communication and trust in the DALI logic programming agent-oriented language.
Stefania Costantini, Arianna Tocchio, Alessia Verticchio · Intelligenza Artificiale · 2005
Interaction is an important aspect of Multi-agent systems: agents exchange messages, assertions, queries. This, depending on the context and on the application, can be either in order to improve their knowledge, or to reach their goals, or to organize useful cooperation and coordination strategies. In open systems the agents, though possibly based upon different technologies, must speak a common language so as to be able to interact. However, beyond standard forms of communication, the agents should be capable of filtering and understanding message contents. A well-understood topic is that of interpreting the content by means of ontologies, that allow different terminologies to be coped with. In a logic language, the use of ontologies can be usefully integrated with forms of commonsense and case-based reasoning, that improve the “understanding” capabilities of an agent. A more subtle point is that an agent should also be able to enforce constraints on communication. This implies being able to accept or refuse or rate a message, based on various conditions like for instance the degree of trust in the sender. This also implies to be able to follow a communication protocol in “conversations”. Since the degree of trust, the protocol, the ontology, and other factors, can vary with the context, or can be learned from previous experience, in a logic language agent should and might be able to perform meta-reasoning on communication, so as to interact flexibly with the “external world” . This paper presents a novel communication architecture for the DALI agent-oriented logic programming language and proposes an example aimed at showing that the communication architecture is general enough as to model sophisticated concepts such as the level of trust. Trust is a type of social knowledge and encodes evaluations about which agents can be taken as reliable sources of information or services. We focus on a practical issues: how the level of Trust influences communication and choices of the agents. Finally, we consider approaches that use a similar architecture and we outline the meaningful differences between the approaches.