A WOz Framework for Exploring Miscommunication in HRI

The aim of this paper is to investigate management of miscommunication in spontaneous interaction between a human and a speech-enabled robot. The paper describes initial results of an exploratory WOz study, in which pairs of naive participants interacted and collaborated in a navigation task. The study is motivated by how humans achieve mutual understanding as well as previous research conducted in the area of spoken dialogue systems. The dialogue and error handling capacity of the wizard is incrementally impaired towards the capabilities of a system in three experimental conditions. Preliminary analysis of the data reveals the necessity to endow the robot with richer error management resources and point to the efficacy of less explicit error handling strategies. Analysis of the data and further experimentation are currently performed and expected soon to shed light to some of the intricacies and unique characteristics of HRI.