Introduction: The increasing integration of digital tools in education highlights the potential of embodied pedagogical agents. This study investigates how task-related errors and language cues from a robot influence human perception, specifically examining their impact on anthropomorphism and subsequent empathy, and whether these perceptions affect persuasion. Methods: Thirty-nine participants interacted with a NAO robot during a quiz. Employing a 3 × 2 mixed design, we manipulated the robot’s error rate (above average, human-like, below average) between subjects and language style (humble, dominant) within subjects. We measured perceived anthropomorphism, empathy, sympathy, and persuasion. Data were analyzed using multilevel modeling to assess the relationships between manipulated variables and outcomes. Results: Our findings indicate that human-like error rates significantly increased perceived anthropomorphism in the robot, which in turn led to higher levels of empathy and sympathy towards it. However, perceived anthropomorphism did not directly influence persuasion. Furthermore, the manipulated language styles did not show a significant direct effect on perceived anthropomorphism, empathy, sympathy, or persuasion in the main experiment, despite pretest results indicating differences in perceived personality based on language cues. Discussion: These results have important implications for the design of embodied pedagogical agents. While strategic implementation of human-like error rates can foster empathy and enhance the perception of humanness, this alone may not directly translate to greater persuasiveness. The study highlights the complex interplay between perceived competence, likability, and empathy in human-robot interaction, particularly within educational contexts. Future research should explore these dynamics further, utilizing larger samples, diverse robot designs, and immersive environments to better understand the nuances of how errors and communication styles shape learner engagement with pedagogical agents.
Rehren et al. (Wed,) studied this question.