established describing detailed requirements concerning the design of agent based systems (tasks, platform, target user). Out of this, a first set of recommendation for the derivation of design principles according to the approach for the allocation of agents in the new classification pattern is given. Finally, an illustration of the method applied for home-shopping in a 3D world is described.
In the office environment and also on the internet, the agent area has already begun. But, transferring these concepts into the home, and into the environment of a residential user, new questions arise which were underlined through trends concerning virtual stars like Kyoto, Lara Croft, or Max Headroom. When agents will be employed in our houses, will this move along the lines shown by Negroponte ( Negroponte 1990, 1997), i.e, having lots of specialized agents, or will we have one almighty nanny? Should these agents be anthropomorphic, or should they rather be abstract? Should they differ in their appearance? Should they materialize, e.g. as robots, or just stay virtual? Of course, these are just some questions which need to be answered. But the most important questions are: What does the user want? What is expected by the residential user? Which user groups are the most promising ones in respect of this new emerging market?
Looking at current research and state-of-the-art literature, experts agree that some kind of agents will play a significant role in future services. However, trying to give a better understanding what they can really do for the private user in the home environment, how the agent will present itself and how to interact and communicate with agents is a topic which is still heavily discussed ( Shneiderman 1997). In principle it can be said that the results from previous experiments and ongoing projects show high potential in applying anthropomorphic agents, at least in the entertainment domain, which can also be extended to the private consumer domain ( Koda 1996). Assuming that UI requirements in this field are also highly demanding, we believe that the focus of future agent research envisaging the private user should shift from the question "should we apply anthropomorphism" into "how should we apply anthropomorphism" for agents.
In order to answer this question, we suggest to adapt and extend the relevant scope and classification of intelligent agents by means of stressing the visualization and interaction domain in the field of consumer products and services in