Autonomous Robots

, Volume 37, Issue 4, pp 335–349

Multivariate evaluation of interactive robot systems

Article

DOI: 10.1007/s10514-014-9415-y

Cite this article as:
Huang, CM. & Mutlu, B. Auton Robot (2014) 37: 335. doi:10.1007/s10514-014-9415-y

Abstract

In designing robot systems for human interaction, designers draw on aspects of human behavior that help them achieve specific design goals. For instance, the designer of an educational robot system may use speech, gaze, and gesture cues in a way that enhances its student’s learning. But what set of behaviors improve such outcomes? How might designers of such a robot system determine this set of behaviors? Conventional approaches to answering such questions primarily involve designers carrying out a series of experiments in which they manipulate a small number of design variables and measure the effects of these manipulations on specific interaction outcomes. However, these methods become infeasible when the design space is large and when the designer needs to understand the extent to which each variable contributes to achieving the desired effects. In this paper, we present a novel multivariate method for evaluating what behaviors of interactive robot systems improve interaction outcomes. We illustrate the use of this method in a case study in which we explore how different types of narrative gestures of a storytelling robot improve its users’ recall of the robot’s story, their ability to retell the robot’s story, their perceptions of and rapport with the robot, and their overall engagement in the experiment.

Keywords

Multivariate evaluation Interactive robot systems  Human-robot interaction Storytelling Narrative gestures 

Supplementary material

10514_2014_9415_MOESM1_ESM.pdf (195 kb)
Supplementary material 1 (pdf 195 KB)

Copyright information

© Springer Science+Business Media New York 2014

Authors and Affiliations

  1. 1.Department of Computer ScienceUniversity of Wisconsin–MadisonMadisonUSA

Personalised recommendations