Observation scheme for interaction with embodied intelligent agents based on Laban notation

Abstract

Designing more intuitive interaction with robots and other intelligent agents relies on knowledge gathered through measurement methods that offer a limited perspective, such as questionnaires and naive observers. To broaden the toolset for designers of interactive robots, we developed an observation scheme based on the framework of Laban/Bartenieff Movement Analysis. Instead of limiting the evaluation to certified movement analysts, we have set and translated parameters that can be evaluated by an uncertified observer. The proposed observation scheme aims to provide more objective measures for the unconscious interaction patterns so designers can use it alongside, or instead of self-report, physiological or other measurements. The development was carried out in the context of a study about participants' interactions with an expressive walk-in closet that was able to take on a more dominant or submissive attitude by using dynamic lighting behaviors. Two certified movement analysts (CMA's) determined and translated the main categories of the Laban framework to measurable qualities for the particular experimental setting. The outcome was validated by measuring the inter-rater agreement and the evaluation of 3 CMA's, who evaluated the videos from the experiment. The observation scheme is not explicitly validated with non-certified LMA observers yet, however, the set-up is such that the observer does not require specific prior knowledge on Laban/Bartenieff Movement Analysis.