Cube-G: CUlture-adaptive BEhavior Generation for interactions with embodied conversational agents
About the Project
The verbal and non-verbal behavior of embodied conversational agents (ECAs) becomes more and more sophisticated. But this behavior is primarily based on a Western cultural background due to the available agent systems and their predefined animation sequences. But according to Ting-Toomey (1999) the most profound misunderstandings in face-to-face communication arise due to misinterpretations of non-verbal cues. Thus, culture-adaptive behavior in embodied agents can further cross-cultural communication in two ways. (1) Employing an agent that adheres to culturally determined behavior programs will enhance the efficiency of information delivery. (2) Agents capable of changing their cultural programs can serve as embarrassment-free coaching devices of culture-specific behaviors.
Based on a theory of cultural dimensions (Hofstede, 2001), we investigate in this DFG-funded interdisciplinary project whether and how the non-verbal behavior of agents can be generated from a parameterized computational model, i.e. it will no longer be necessary to develop a new agent for each culture but specifying a culture's position on the basic dimensions will allow the system to generate appropriate non-verbal behaviors for the agents. The project combines a top-down model-based approach with a bottom-up corpus-based approach which allows to empirically ground the model in the specific behavior of two cultures (Japanese and German).