A Semantics for the Interaction of Agents with Coordination Artifacts∗

In this paper we consider the multi-agent systems (MAS) scenario where agents do not only cooperate by direct, communicative speech acts, but may also interact with coordination artifacts — embodied entities dynamically provided by the MAS infrastructure to realise specific coordination rules. We study a semantics for this kind of interaction, based on the idea that agents exploit the coordination artifact according to some operating instructions, which describe the interaction protocol the agent has to follow as well as the mentalistic semantics of each single interaction. By tackling some of the most relevant issues raised in the context of ACL semantics, our framework allows intelligent, BDI-like agents to carry on interaction protocols in a rational way. 1 Mediated Interaction One of the key principles of the research on agentbased systems is the intentional stance [Dennett, 1987]: as far as the complexity of MAS design is concerned, agent behaviour is better understood, predicted, and analysed in terms of mental properties — such as believes, desires, and intentions (fears, hopes, and so on) — instead of just relying on the agent design. This approach has then promoted a connection between the agent abstraction and human behaviour (and between MASs and social groups of people) which drove a number of typical design choices in the agent field. Most notably, these include the idea of basing agent communications on the speech act theory [Searle, 1969]: each communication act is directed to another agent, and is an utterance whose meaning is associated to a human-like kind of speech, also called performative, such as an inform, a request or a propose. This model also made into the agent mainstream: in the FIPA standard, agents interact with ∗This work has been partially supported by MIUR, project COFIN 2003 (ex 40%) “Fiducia e diritto nella società dell’informazione”, and by MIPAF, project SIPEAA “Strumenti Integrati per la Pianificazione Eco-compatibile dell’Azienda Agricola”. each other through direct communication via speech acts, exploiting the so-called FIPA Agent Communication Language (ACL) [FIPA, 2000]. However, direct communication is not the only viable approach: new models are emerging in different research areas of MASs, e.g. in environmentbased coordination such as stigmergy and, more generally, mediated interaction frameworks and infrastructures based on forms of coordination/cooperation without direct communication [Parunak et al., 2003; Fenster et al., 1995; Ciancarini et al., 2000; Bonabeau et al., 1999; Odell et al., 2003]. Even if we stick to human behaviour, we observe that direct speaking is not the only way of cooperating. Rather, as emphasised in the Activity Theory [Nardi, 1996] — a social psychological theory about dynamics in human work activity —, humans often use mediating artifacts such as blackboards, semaphores, mailboxes, form sheets and maps, to make the cooperation task easier, more flexible and effective. So, it is clear that, especially as far as complexity is concerned, other forms of interactions might be interesting that do not involve two directly communicating agents, but rather an agent acting on an abstraction that provides a suitable mediation for agent interactions, an abstraction that might not be fruitfully understood in terms of the intentional stance. 2 Coordination Artifacts Along this idea, in [Ricci et al., 2003] the framework of coordination artifacts has been introduced, based on the very idea of adapting to the MAS field the concept of mediating artifact from Activity Theory. Coordination artifacts are embodied entities provided by the MAS infrastructure playing a similar role to mediating artifacts: their aim is to mediate the interaction between a group of agents, supporting a specifically designed coordination task. An example of MAS infrastructure supporting this notion is TuCSoN [Omicini and Zambonelli, 1999; Ricci et al., 2003]. Coordination artifacts are entities for which the agent abstraction is not an appropriate model. This is because their behaviour is not characterised as “achieving a goal in autonomy”, and their characteristics do not include proactiveness or rationality. Rather, a coordination artifact realises the laws characterising a given coordination task without the freedom of autonomy: such laws can be inspected by the agent and can be adapted according to some emerging need. Besides the notion of mediating artifact from Activity Theory, which provides the basic motivation of our scenarios for MASs, coordination artifacts are inspired also by the notion of coordination medium [Ciancarini, 1996] developed in the distributed systems area — in particular, in the field of coordination models and languages [Omicini et al., 2001]. Coordination media are design-time abstractions conceived to govern and rule the interaction of processes into a concurrent system, typically dealing with synchronisation and communication issues — in one word, with coordination ones. From coordination media, thus, we inherit parts of an engineering methodology approach, based on a substantial tradition of ontological, theoretical and development studies, grounded e.g. on the notion of interaction, on formal tools such as process algebras and behavioural equivalence, and on development techniques such as coordination infrastructures [Viroli and Omicini, 2003]. Indeed, this guarantees the ability of designing and implementing coordination infrastructures — such as e.g. TuCSoN and Mars [Cabri et al., 2000] — that can suitably tackle the complexity of the coordination issue in MAS. The notions of mediating artifact and coordination medium then comes together into coordination artifacts for MAS, which are run-time abstractions used to mediate the interactions of a group of collaborating agents.