Inhabited interfaces: attentative conversational agents that help

We discuss the role of attentive agents in virtual reality interfaces. This discussion is guided by experiences and experiments with a virtual reality environment we designed and implemented. In this environment we have introduced agents, sometimes embodied, with which the users can communicate using different input modalities. These agents provide information or are able to perform certain transactions or they help the user in finding her way in the virtual environment, allowing a mix of user exploration and guidance. Among the input modalities that are considered are speech, natural language, mouse and keyboard and gaze. Output includes natural language, visual speech, changes in the virtual environment, animations and menus. Gradually this environment evolves to an environment where multiple users and agents live and communicate with each other. Apart from offering different input modalities and attentive agents, in the near future we also hope to be able, based on current experiments, to offer suggestions to the users based on preferences obtained from their user profile and their visit history. 1