PUB - Publikationen an der Universität Bielefeld

representational gestures are an integral part of human interaction, which helps to shape the conversation and facilitates the recall of essential information. Although this is true of human interaction, artificial systems exhibit difficulty in creating meaningful representational gestures that are able to convey additional information. In this project we will develop an adaptive non-verbal gesture system that that enables the generation of goal-directed gestures for collaborative tasks. As part of the development, we are exploring the amount of non-verbal expressiveness and fluidity that a system needs to exhibit, in order for gestures to be acknowledged as representational gestures. In addition, we want to understand if and how much an interaction benefits from the use of representational gestures between a human and agent and especially if this differs from a human-human context. After finishing the main components of the system, we will test the influence that intention detection and memory retention has on the perceived information in artificially created gestures. Here a special focus will also be on erroneous information and the deterioration of the interaction. We are currently working on adaptive non-verbal behaviour generation and have already developed an approach for detecting dis-/agreement and confusion events.