We present work in progress on (verbal, facial, and gestural) modality selection in an embodied multilingual and multicultural conversation agent. In contrast to most of the recent proposals, which consider non-verbal ...