Milking the Reputation Cow: Argumentation, Reasoning and Cognitive Agent

Informatica: cuestiones generales

Computational trust and reputation models have been recognized as one of the key technologies required to design and implement agent systems. These models manage and aggregate the information needed by agents to efficiently perform partner selection in uncertain situations. For simple applications, a game theoretical approach similar to that used in most models can suffice. However, if we want to undertake problems found in socially complex virtual societies, we need more sophisticated trust and reputation systems, that not only focus on the construction and inference of social evaluations (epistemic decisions), but on their role in the practical reasoning performed by the agents (pragmaticstrategic decisions) and on communications and dialectical processes (memetic decisions). Most of the current state-of-the-art models struggle with epistemic decisions, on how agents evaluate other agents according to certain criteria. Curiously, pragmatic-strategic and memetic decisions are traditionally left apart, either because they are implicit in the model or because it is too dependent on the domain. This work explores this gap, arguing that in complex scenarios where more cognitive approaches are needed, both pragmatic-strategic and memetic decisions are as important as epistemic ones. Firstly, we construct an ontology of reputation and a reputation language that captures the information that most of the current state-of-the-art computational trust and reputation models manage. This starting point serves, by the one hand, to define a belief-desire-intention (BDI) agent architecture that integrates reputation information. Then, desires and intentions interact with beliefs that contain information coming from reputation models. The architecture is flexible enough to model a wide number of agents’ families and precisely determines the practical reasoning process that leads to the best reasonable action. On the other hand, we exploit the reputation language and use it to define an argumentation-based protocol that allows two parties to engage in dialog processes and exchange reputation-related information. The protocol permits agents justify their social evaluations, endowing them with the capability to intendedly decide whether communicated social evaluations are reliable according to their own knowledge.