| The present study examined the efficacy of enhancing human-agent interaction through the use of nonverbal behaviors. A taxonomy was created, which organizes nonverbal behaviors into functional categories, including adaptors, regulators, affect displays, illustrators, and emblems. The taxonomy further divided these behaviors by the manner in which they can be embodied (i.e., through gesture, posture, paralanguage, eye contact and facial expression). The influences of demographic variables (i.e., ethnicity, age and gender) and physical appearance (i.e., bodily and facial attractiveness, clothing and artifacts) were also considered. Forty participants were presented with computer characters that varied in terms of their ethnicity, age, and gender and asked which they would prefer to cooperate with on a computer task. The results indicated that in general youthful agents of similar ethnicity were preferred but there was no clear gender preference. In the main empirical study anthropomorphic computer characters were endowed with different levels of nonverbal behavior known to engender a trusting persona. The first level provided facial expressions, eye contact, and paralanguage. The second level added to these behaviors bodily gestures and posture. A third embodiment engendered an agent with non-trusting behaviors. Based on the results of the pilot study, the characters were youthful and selected to match each participant's ethnicity, while the gender of the agent was left open to participant's choice. It was hypothesized that those agents with the first level of nonverbal behaviors would be perceived as more credible than an anthropomorphic computer character without such attributes. Adding the second level of nonverbal behaviors was expected to result in further gains, while the non-trusting agent embodiment was suggested to be ineffective. A between subjects experimental design was used, in which forty-eight female participants interacted with one of the four computer characters that assisted them in sorting photographs. The results indicated that participants that interacted with computer characters endowed with facial expressions, eye contact, and paralanguage perceived these characters as being more trustworthy and satisfying to interact with than those who interacted with an anthropomorphic character with no nonverbal behaviors. (Abstract shortened by UMI.)... |
