When communicating with a machine, we'd like the interaction to be as
easy and intuitive for us as possible. To achieve a more or less
natural communication behaviour, a robot must use similar
communication patterns to ours. Since these typically also involve
modalities beyond language, we need to investigate how and when humans
use others like gaze and gesture. Based on knowing how language
combines with visual information, we can hopefully equip a robot with
multi-modal means to express itself within and about the world.

In this talk, I will first present previous work in which we developed
an approach to modelling gaze production for robots based on insights
about human gaze behaviour. Secondly, I will sketch and discuss
problems and benefits of this approach. And finally, I will outline an
idea for a human-robot interaction scenario in which the presented
approach is in general refined and extended by another modality -
gesture.