Evaluating the Effect of Language Fluency and Task Competency on the Perception of a Social Robot

Recent advancements in robot capabilities have enabled them to interact with people in various human-social environments (HSEs). In many of these environments, the perception of the robot often depends on its capabilities, e.g., task competency, language fluency, etc. To enable fluent human-robot interaction (HRI) in HSEs, it is crucial to understand the impact of these capabilities on the perception of the robot. Although many works have investigated the effects of various robot capabilities on the human’s perception of the robot separately, in my M.S. research, I present three large-scale HRI study (total n = 120) to investigate the combined impact of both language fluency and task competency on the perception of a robot.

For the first study, we enlisted monolingual participants who only spoke English (n = 60) in order to control for the participants’ linguistic background. Then, to establish a social interaction while also incorporating a task to complete between the human and the robot, we designed a simple guessing game modeled after the children’s game called ‘What Am I?’ for the participant and robot to play together where the robot assumed the identity of an animal and gave the participant hints to guess the animal. To accommodate the differences in language proficiency and task competence of the robot, we equipped it with the ability to be both fluent and disfluent in English and correctly match or fail to identify the animal, respectively. Finally, we developed four distinct combinations of language fluency and task competency (fluent-competent, fluent-incompetent, disfluent-competent, and disfluent-incompetent) which made up the varying robot conditions in the study. Participants were randomly placed into one of these four groups, and to understand how the robot’s condition impacts human perception of the robot, we collected and analyzed participants’ perceptions of the robot’s verbal competence, intelligence, and reliability along with the robot meeting expectations, being a good teammate, and if the participant is willing to work with the robot again.

From the monolingual perspective in Study I, the fluent-competent robot was rated higher than the disfluent-incompetent robot in every perception category except willingness to work again. The results indicate that participants found both the fluent robots (fluent-competent and fluent-incompetent) to be significantly more verbally competent than the two disfluent robots (disfluent-competent and disfluent-incompetent). Participants also perceived the fluent-competent, fluent-incompetent, and disfluent-competent robots to be more intelligent than the disfluent-incompetent robot. Participants in the fluent-competent, fluent-incompetent, and disfluent-competent conditions all perceived the robot to be significantly more reliable than how the participants in the disfluent-incompetent condition perceived the robots to be. Regarding perceptions of expectations being met, participants in the fluent-competent and disfluent-competent conditions perceived the robot to have met their expectations more than the participants in the disfluent-incompetent condition did. Lastly, participants in the fluent-competent and disfluent-competent conditions rated the robot more highly for being a good teammate than participants in the disfluent-incompetent condition did. There was no statistical significance of the effects of varying robot conditions on participants’ willingness to work with the robot again.

In the second study, we aimed to maintain the same experimental setup and design as the first study, but instead of requiring participants to be monolingual in English, we sought out participants (n = 60) who were multilingual, with English as one of their fluently spoken languages. The results of this study suggest that participants found the fluent-competent robot to be significantly more verbally competent than the other three varying robot conditions (fluent-incompetent, disfluent-competent, and disfluent-incompetent). Participants perceived both the competent robots (fluent-competent and disfluent-competent) to be more intelligent than the incompetent robots (fluent-incompetent and disfluent-incompetent robots). The results also indicate that both the competent robots (fluent-competent and disfluent-competent) were perceived to be more reliable than the incompetent robots (fluent-incompetent and disfluent-incompetent robots). Additionally, participants rated the fluent-competent robot significantly higher in reliability than the disfluent-competent robot. The participants also perceived both the competent robots (fluent-competent and disfluent-competent) to have met their expectations more than the incompetent robots (fluent-incompetent and disfluent-incompetent robots). Participants in the fluent-competent and disfluent-competent conditions rated the robot significantly higher for being a good teammate than participants in the fluent-incompetent and disfluent-incompetent conditions did. Finally, the results suggest that the participants are more willing to work with both the competent robots (fluent-competent and disfluent-competent) than the incompetent robots (fluent-incompetent and disfluent-incompetent robots).

The third study entailed comparing the perception of robots from the monolingual perspective with the perception of robots from the multilingual perspective, and this was done by combining the data from both Study I and Study II (n = 120). The purpose of this third study was to investigate potential interactions between language groups and varying task conditions on perceptions of the robot. We found that there were statistically significant differences between the ratings of monolingual participants and that of multilingual participants in certain varying robot conditions. Multilingual participants perceived the verbal competence of the disfluent-competent and disfluent-incompetent robot to be significantly higher than monolingual participants did. Multilingual participants also rated the fluent-competent robot more highly as a good teammate than their monolingual counterparts. However, the monolingual participant group rated the fluent-incompetent robot significantly higher than the multilingual participants in intelligence, reliability, meeting expectations, and willingness to work with the robot again.

The results suggest that both language fluency and task competency may impact certain perceptions of the robot at different scales. For example, in Study I, while language fluency may play a more significant role than task competency in the monolingual perception of the verbal competence of a robot, both language fluency and task competency contribute to the perception of the intelligence and reliability of the robot. On the other hand, task competency played a more significant role than language fluency in the perception of meeting expectations and being a good teammate for monolingual individuals. In Study II, multilingual participants were more impacted by task competency in their perceptions of the robot’s intelligence, ability to be a good teammate, meet expectations, and their willingness to work with the robot again. In Study III, monolingual participants prioritized fluency more than task competency in the perception categories of intelligence, reliability, meeting expectations, and willingness to work with the robot again as they rated the fluent-incompetent robot significantly higher in these conditions than multilingual participants did. Although the studies in our research did not investigate why monolingual or multilingual perceptions may have varied across certain conditions, these results may serve as a foundation for future research to explore additional relationships between monolingual and multilingual participants’ perception of robots while also understanding the factors that may cause these perceptions to differ. Overall, the findings of these three studies highlight the relationship between language fluency and task competency through the lens of linguistic background in the context of social HRI and will enable the development of more intelligent robots in the future.