In cooperative environments is common that agents delegate tasks to each other to achieve their goals since an agent may not have the capabilities or resources to achieve its objectives alone. However, to select good partners, the agent needs to deal with information about the abilities, experience, and goals of their partners. In this situation, the lack or inaccuracy of information may affect the agent's judgment about a given partner; and hence, increases the risk to rely on an untrustworthy agent. Therefore, in this work, we present a trust model that combines different pieces of information, such as social image, reputation, and references to produce more precise information about the characteristics and abilities of agents. An important aspect of our trust model is that it can be easily configured to deal with different evaluation criteria. For instance, as presented in our experiments, the agents are able to select their partners by availability instead of the expertise level. Besides, the model allows the agents to decide when their own opinions about a partner are more relevant than the opinions received from third parties, and vice-versa. Such flexibility can be explored in dynamic scenarios, where the environment and the behavior of the agents might change constantly.
Em um sistema multi-agente (SMA), é muito comum que os agentes deleguem tarefas uns aos outros. Contudo, devido à subjetividade das informações utilizadas pelos agentes durante o processo de tomada de decisão, um agente pode acabar delegando uma tarefa a um parceiro não confiável. Neste trabalho, apresentamos uma abordagem de cálculo de confiança baseada em argumentação quantitativa com votação (QuAD-V), onde a confiança é estimada de acordo com a opinião dos agentes sobre o serviço prestado por um parceiro. Além disso, tal abordagem fornece um mecanismo para avaliar a credibilidade dos agentes que atuam como fontes de informação. Como resultado, demonstramos como nossa abordagem de cálculo de confiança pode ser empregada para detectar agentes mentirosos capazes de caluniar ou promover outros agentes.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.