Integrating Virtual Reality, Chatbots, and BDI Agents: VEsNA Goes Fast!

In the era of Generative Artificial Intelligence and the Metaverse, agents should be designed with social interaction involving humans in mind, and their situatedness should account for Virtual and eXtended Reality (VR/XR) environments. The social and the environmental contexts, where humans interact with agents in natural language and/or via virtual proxies in the form of avatars, should drive the agents’ decisions and actions. According to the strong definition, agents are conceptualized not only in terms of autonomy, situatedness, social interaction, proactivity, and reactivity but also through mentalistic attributes such as beliefs, desires, goals, and intentions. The Belief-Desire-Intention (BDI) model and architecture stick to the strong definition and are suitable to provide an explicit context to the agents’ reasoning and decision-making, which allow for explainability, transparency, and accountability. VEsNA has been conceived as the integration of the BDI architecture with chatbot-like interaction (possibly—although not necessarily—boosted by Generative AI, in particular Large Language Models) and situatedness in a Virtual Reality environment: VEsNA is an agent-based framework for managing Virtual Environments via Natural language Agents. In this chapter we motivate and describe VEsNA, we position it in the current scientific panorama, we discuss the challenges we faced since its first release was developed in 2022, and we present its foreseen evolution in the medium-long period.