Carolina Howard Felicíssimo

Providing contextual norm information in open multi-agent systems

Agents can freely migrate among open MASs in order to obtain resources or services not found locally. In this scenario, agent actions should be guided for avoiding unexpected behaviour. However, open MASs are extremely dynamic and, thus, a solution for guiding agent actions is non-trivial. This work details DynaCROM, our solution for continuously supporting agents in open MASs with updated norm information. The main asset of DynaCROM is that it decreases the complexity of norm representation by using contexts. DynaCROM proposes (i) a top-down modelling of contextual norms, (ii) an ontology to explicitly represent norm semantics and (iii) a rule inference engine to customize different compositions of contextual norms. Thus, DynaCROM offers a solution for both developers to enhance their open MASs with norm information and agents to be continuously supported with precise norm information.

The Autonomic Semantic Desktop: helping users cope with information system complexity

In this paper we present the architecture for the autonomic semantic desktop, an self managing application designed to hide system complexity and alleviate users from burdening and time consuming management tasks. The proposed architecture is founded in the basic principles of autonomic systems, i.e. ,the set of self* attributes. We introduce a semantic layer to the architecture, to allow machines to process and exchange information without requiring human mediation. The underlying conceptual model used in this layer is ontology, currently the lingua franca in application on the semantic Web

Dynamic contextual regulations in open multi-agent systems

Following software engineering approaches for the Semantic Web (SW) and also Hendler’s vision [3], I believe that the SW will not be a unique large complex Web, but it will be composed, mainly, of several small contextualized domain applications. These domain applications will be, in my opinion, Multi-Agent Systems (MAS) [5]. MAS have emerged as a promising approach for the development of information systems, which are constituted of cooperative goal-oriented problem-solving entities (named agents) [6]. Agent-based computing is rapidly emerging as a powerful technology for the development of distributed and complex information systems.

An Approach to Operationalize Regulative Norms in Multiagent Systems

Multiagent systems have emerged as a promising approach to develop information systems that clearly require several goal-oriented problem-solving entities [Jennings et al., 1998]. Following this direction, it is believed that upcoming information systems will be implemented as open multiagent systems, in which agents (their entities) can freely migrate among those systems in order to obtain resources or services not found locally. A multiagent system(s) (hereinafter referenced as MAS1) is/are an example of an open system in which the actions of heterogeneous, self-interested agents may deviate from the expected behavior in a context. Openness has led to dynamic software systems that have no centralized control and that are composed of autonomous entities [Hewitt, 1991]. Key characteristics of such systems are heterogeneity, conflicting individual goals and limited trust [Artikis et al., 2002]. As stated in [Esteva et al., 2004], “openness without control may lead to chaotic behavior”. In order to be a viable solution for dynamic software systems, MAS must be enhanced with norms for defining which actions are permitted, obliged and prohibited to be performed by agents so that the system does not reach an undesirable state. A permitted norm defines that an action is allowed to be performed; an obliged norm defines that an action must be performed; and, a prohibited norm defines that an action must not be performed. Permissions and prohibitions are used to describe positive/negative authorizations, whereas obligations are used to describe responsibilities [Kagal and Finin, 2007]. These three types of norms represent the three fundamental deontic statuses of an action [Alberti et al., 2006] from deontic logic [Wright, 1951] and they are logically connected as presented by the following statements: If an action is permitted, then, it is not prohibited; If an action is obligatory, then, it is permitted and it is not prohibited; If an action is prohibited, then, it is not obligatory and it is not permitted;