Event Location: NSH 1507

Abstract: Believability of characters is an objective in literature, theater,
animation, film, and other media. In human-computer interaction,
believability of on-screen agents improves perceptual and behavioral
responses to the character. Social scientists refer to this phenomenon
as homophily—humans tend to associate and bond with similar others.

In this thesis proposal, we first focus on achieving homophily through
ethnic congruence (i.e., similarity). We devise a methodology for
designing ethnically believable robot characters by (1) selecting
candidate behaviors, or rich points, from qualitative anthropological
studies, quantitative human corpora analyses, literature on second
language acquisition, etc., (2) evaluating perceptual response to the
on-screen renderings of the behaviors via crowdsourcing, (3) using
these evaluations to combine behaviors in a robot character, and (4)
evaluating ethnic attribution, perceptual and behavioral responses
evoked by the robot character. Our results to date show the
feasibility of steps (1) and (2), namely, of using crowdsourcing to
evaluate cross-cultural differences in perception of verbal behaviors
that correspond to rich points between native speakers of American
English and Arabic. The evaluation of the response evoked by an
ethnically believable robot character planned in step (4) shall shed
light on the currently open question: Do ethnically congruent robots
evoke homophily, just like ethnically congruent humans and on-screen
agents do?

In the second part of this thesis proposal, we address the following
question: Can homophily be achieved when ethnic congruence is not
possible? In particular, we attempt to explore the profile of
perceptual and behavioral responses to a robot character that evokes
incongruent ethnic attribution. The estimate of this profile shall
result in a model of inter-cultural alignment: a behavior shift aimed
at improving homophily under the ethnic incongruence constraint.

Committee:Reid Simmons, Chair

Justine Cassell

Illah Nourbakhsh

Michael Agar, Ethknoworks LLC, University of Maryland

Candace Sidner, Worcester Polytechnic Institute