Shared Experiences and Intentionality

Introducing the notion of experience as a primitive construct, we illustrate how the BDI cycle is completed: experiences generate beliefs, which in turn generate desires, which in turn generate intentions, which in turn generate new experiences. Furthermore, the notion of experience is used to ground social BDI agents in the physical world. We do not assume an agent’s BDI model already exists, but we illustrate how the agent’s mind, along with its environment (or its perceptions of the environment, to be more precise), helps shape and evolve its BDI belief base. The notions of shared experience and shared intentionality then naturally arise, helping us better understand and motivate social behaviour: the foundation of multiagent systems. Copyright 2011, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. As a consequence, we then build the notion of shared experience on top of the notion of single individual experiences. This paper’s proposed model helps provide a formal specification, with a computational semantics, of an abstract agent architecture for social BDI agents. We argue that such a model motivates social behaviour. For instance, it helps explain why would one agent approach another agent; why do agents speak in the first place. Furthermore, shared experiences may be used as the bases for shared intentionality, which drives group behaviour, as opposed to individual intentionality, which drives individual behaviour. Introduction In this work, we do not assume an agent’s BDI model already exists, but we try to explain how it is built through the agents’ capabilities of perceiving the environment, imagining and desiring, manifesting beliefs based on percepts (and possibly desires), and so on. We basically provide a trivial definition of what an environment is, followed by a set of definitions of the required agent’s capabilities for manifesting perceptions, experiences, beliefs, desires, and intentions. In summary, this paper essentially provides a grounding framework for social BDI agents in real environments where an agent’s perception of its environment is formally modelled, resulting in the novel notion of “experience”. We then illustrate how this notion of experience helps complete the BDI cycle resulting in a dynamically evolving BDI. Last, but not least, the framework can then be used to design agent-based technologies for enabling shared social experiences in human societies. Naturally, social BDI agents can then be designed and tested to evaluate the claim that social agents produce a more flexible and robust society. With such a framework, one can also evaluate the impact of technology-mediated shared experiences, which we currently leave for future work. This can lead to increased refinements of our model but also provide an evaluation of its ability to model social human behaviour.