C-Roids: Life-like Characters for Situated Natural Language User Interfaces Anders Green IPLab, Nada Royal Institute of Technology (KTH) 100 44 Stockholm, Sweden E-mail: green@nada.kth.se Abstract In a future scenario where many devices can be controlled using the voice, easy and intuitive access and will be crucial for avoiding cognitive overload when faced with many different systems and interac- tion models. We propose a model for interaction with spoken language interfaces with heterogeneous tasks, based on the idea of using a family of life-like charac- ters. The properties of these characters are based on psychological and linguistic principles of human natu- ral language interaction. The general idea is that hu- manlike characters afford for human natural language behavior such as the ability of understanding and pro- ducing speech. 1 Introduction The purpose of this paper is to discuss how multi- ple life-like characters can be used for providing non- intrusive behavior for heterogeneous spoken language agents installed in the home or in the work-place. We are especially interested in what kind of metaphors that could provide viable models for straightforward and intuitive access to devices where the underlying system has diverse tasks. 2 Diversity in NL interfaces We argue that we can signal important features of the speech interface by using certain visual cues. By using the same kind of design elements for different instances of a speech interfaces, the user will be able to use the experience of using one interface to learn another one. In the coming years, the cost of hardware will be- come low enough to equip any kind of device with a spoken language interface. In a normal household, most appliances contain microprocessors that are used for control and user interface. These appliances can be found in the kitchen (e.g. microwave ovens, refrigera- tors) or in the living room (e.g. television, video and Figure 1: The CERO fetch-and-carry robot with the interface character placed on top of the robot cover. stereo equipment). Interaction with these interfaces will probably take on many different forms. The ideas presented in the following emanate from our work on Human-Robot Interaction. We are ex- ploring ways of providing non-intrusive visual feed- back by means of a life-like character, CERO, who is attached to the robot cover (Figure 1). The CERO character is integrated with the speech system so that it is both capable of issuing conversational gestures reactively, based on system states, and displaying co- expressive symbolic gestures. Natural language interfaces are normally used to perform a set of predefined discrete high-level actions. The work described in the following is based on the assumption that the mental models possessed by in- terlocutors involved in spoken interaction can be used in the design of interfaces for the ubiquitous computer. We argue that artificial agents possessing human-like