dc.description.abstract | Behavioral mimicry has been claimed to be a nonconscious behavior
that evokes prosocial e ects | liking, trust, empathy, persuasiveness |
between interaction partners. Recently Intelligent Virtual Agents (IVAs)
and Immersive Virtual Environments (IVEs) have provided rich new possibilities
for nonverbal behavior studies such as mimicry studies. One of
the best known e ects is the \Digital Chameleons" in which an IVA appears
to be more persuasive if it automatically mimics a listener's head
nods. However, this e ect has not been consistently replicated. This
thesis explores the basis of the \chameleon e ects" using a customized
IVE integrated with full-body motion capture system that support realtime
behavior manipulation in the IVE. Two replications exploring the
e ectiveness of the virtual speaker and head nodding behavior of interaction
partners in the agent-listener interaction and avatar-listener interaction
by manipulating the virtual speaker's head nods and provide mixed
results. The rst experiment fails to replicate the original nding of
mimicry leading to higher ratings of an agent's e ectiveness. The second
experiment shows a higher rating for agreement with a mimicking avatar.
Overall, an avatar speaker appears more likely to activate an e ect of
behavioral mimicry than an agent speaker, probably because the avatar
speaker provides richer nonverbal cues than the agent speaker. Detailed
analysis of the motion data for speaker and listener head movements
reveals systematic di erences in a) head nodding between a speaker producing
a monologue and a speaker engaged in a dialogue b) head nodding
of speakers and listeners in the high and low frequency domain and c)
the reciprocal dynamics of head-nodding with di erent virtual speaker's
head nodding behavior. We conclude that: i) the activation of behavioral
mimicry requires a certain number of nonverbal cues, ii) speakers
behave di erently in monologue and dialogue, iii) speakers and listeners
nod asymmetrically in di erent frequency domains, iv) the coordination
of head nods in natural dialogue is no more than we would expect by
chance, v) speakers' and listeners' head nods become coordinated by
spontaneous collaborative adjustment of their head nods. | en_US |