Computer Science Thesis Oral

  • Gates&Hillman Centers
  • Traffic21 Classroom 6501
Thesis Orals

Designing Animated Human Characters as Conversational Partners: The Effect of Facial Expressiveness on Social Outcomes

Animators follow common animation practices to make characters engaging, funny, and highly effective as entertainment.  The research question posed in this thesis is whether similar practices can be used to create successful animated characters whose main purpose is education and therapy rather than entertainment. The former are domains in which trust and respect for the character are likely to be essential.  In this dissertation, I evaluate the standard practice of exaggerating character expressiveness and suggest design options for animated characters used in education and therapy. My investigation focuses on the practice of exaggerating expressive movement in animated human faces. I show that small differences in facial movement can create important differences in the inferences of viewers and the trust that people have in animated characters.

I began by investigating the perceptual effects of modifying the spatial motion of animated characters' faces. I determined participants' sensitivity to motion changes in the faces of characters rendered in a cartoon and a comparatively more realistic style.  I then conducted experiments to determine how facial motion magnitude influenced viewers’ impressions of characters’ social traits, emotional intensity, and naturalness.  Next, I investigated people’s responses to interactive animated characters that exhibited different amounts of facial motion. In two studies, I examined adults’ and children’s interactions with a confederate whose motion was tracked in real time to animate characters known as avatars. Adults collaborated on a task with a confederate-controlled, realistic avatar. Children, aged 4 to 10 years, spoke to a confederate using videoconference and avatar-mediated communication. I examined participants’ beliefs about the avatar and their interactions.

My research suggests that damped realistic characters may be more appropriate for education and therapy than exaggerated cartoon characters.  Adults found the damped realistic avatar more appealing than the exaggerated avatar, and girls attended more to the damped avatar than the exaggerated avatar. Additionally, my experiments empirically demonstrate that exaggerating realistic characters makes them appear unnatural instead of believable and lifelike.  Therefore, the practice of exaggerating animated characters may not have its desired effect when applied to characters meant for education and therapy. To extend and validate my results, future work could use different animation methods, characters, and tasks.

Thesis Committee:
Jessica K. Hodgins (Co-Chair)
Sara KIesler (Co-Chair)
Nancy Pollard
Carol O'Sullivan (Trinity College Dublin and Disney Research)

For More Information, Please Contact: