The Interaction Between Voice and Appearance in the Embodiment of a Robot Tutor

Abstract

Robot embodiment is, by its very nature, holistic and understanding how various aspects contribute to the user perception of the robot is non-trivial. A study is presented here that investigates whether there is an interaction effect between voice and other aspects of embodiment, such as movement and appearance, in a pedagogical setting. An on-line study was distributed to children aged 11-17 that uses a modified Godspeed questionnaire. We show an interaction effect between the robot embodiment and voice in terms of perceived lifelikeness of the robot. Politeness is a key strategy used in learning and teaching, and here an effect is also observed for perceived politeness. Interestingly, participants' overall preference was for embodiment combinations that are deemed polite and more like a teacher, but are not necessarily the most lifelike. From these findings, we are able to inform the design of robotic tutors going forward.

Can you Tell the Robot by the Voice? An Exploratory Study on the Role of Voice in the Perception of Robots

“…Although some of the default robot voices had a specific accent (e.g. Flash in [16] has a Scottish female voice), here we only report people's selection upon hearing the voices we recorded or resynthesised, which had either an American, Irish, or Italian accent of English. Since the vocal resynthesis process might have altered some of the nuances of the various accents, we performed chisquare tests of independence for the natural and resynthesised accents separately.…”

The Effect of Multimodal Emotional Expression and Agent Appearance on Trust in Human-Agent Interaction

“…In fact, in [Tinwell et al 2013], uncanny virtual characters were also attributed more negative personality traits -including untrustworthiness. Similarly, in Human-Robot Interaction, previous studies found that people did not prefer or were engaged by more anthropomorphic robotic systems [Hastie et al 2017;Lemaignan et al 2015].…”

Audio-visual predictive processing in perception of humans and robots

“…Although there have been studies that examined the congruence of appearance and voice in perception of robots, most of these studies use subjective measures in characterizing people's experience with these agents (Mitchell et al, 2011;Hastie et al, 2017;Cabral et al, 2017;Stein and Ohler, 2018;Tsiourti et al, 2019;Mara et al, 2020). We extend this work by characterizing perception with more objective measures such as reaction time and accuracy.…”

“…Other studies explored mismatches in multisensory context including the congruity of appearance and voice (Mitchell et al, 2011;Hastie et al, 2017;Cabral et al, 2017;Stein and Ohler, 2018;McGinn and Torre, 2019). One drawback of these studies is that they usually use subjective measures such as fear and eeriness (Mitchell et al, 2011), credibility or attractiveness (Stein and Ohler, 2018), politeness and lifelikeness (Hastie et al, 2017), likability, expressiveness, and understandability (Cabral et al, 2017), drawings (Mara et al, 2020), or emotion labeling (Tsiourti et al, 2019) to evaluate artificial agents rather than more objective measures such as reaction time or accuracy, which would be more informative about the basic perceptual processes.…”