In modern society, voice-based systems intended to assist humans in day-to-day tasks are populating both industrial and domestic environments. This increasing use of robots is making users more likely to be exposed to robot failures, often requiring costly human intervention. The influences of robot embodiment and failure severity on people's subjective perception of robots have been studied separately. However, it seems that no previous research has studied how robot embodiment influences users' perception of failing robots, in particular with different levels of failure severity.
Additionally, user extroversion is known to affect human-human communication but the corresponding effect on different aspects of human-computer interaction is relatively unknown. In order to improve the use and acceptance of robots, it is important to investigate how different personality traits influences user performance and experience while interacting with failing robots.
This thesis assessed the effects of robot embodiment, failure severity, and user extroversion, on people's behaviours and perception of failing robots. In a mixed methods study, participants with measured levels of extroversion were instructed to cook spring rolls by a smart-speaker and a human-like robot. Both robots occasionally delayed the participants by performing intentional robot failures. Failure severity was manipulated with the introduction of a time pressure condition.
The results indicate that robot embodiment generally does not affect people’s task performance. Additionally, a human-like robot is generally preferable over a smart-speaker and is perceived as more competent, predictable, and intelligent. Smart-speakers tend to elicit weaker feelings of co-presence and behavioural interdependence compared to human-like robots. We also found that robot failures (especially with severe consequences) caused participants to perform incorrect actions more frequently. Failure severity was found to not have a main effect on the general perception of failing robots. However, participants performing tasks under time pressure asked less clarification questions to finish faster. Lastly, introverts tend to ask more clarification questions and therefore be slower in task-oriented interactions. They also experience a less perceived sense of co-presence while interacting with robots.