Learning with Virtual Peers

Justine Cassell, Northwestern University

Friday, March 3, 2006

Justine Cassell presented her pioneering and intriguing research using Embodied Conversational Agents – digital “virtual peers,” displayed on large computer screens – to help children improve literacy skills during language play and storytelling. Professor of Media, Technology, and Society at Northwestern University, Cassell is interested in the ways that peer performance may stimulate children to perform more robust and more complex narratives.

Cassell analyzes the interactions between children and virtual peers on a micro level, studying slices of behavior such as gesture, posture, and intonation, and how they lead to learning. To incorporate context and body movement – both key to young children's storytelling – the study uses play objects such as castles that become part of the narrative setting.

While most research on pedagogical agents focuses on the look and feel of the on-screen character, which tends to result in stereotyped depictions, Cassell focuses on verbal and non-verbal cues that influence children's perception of virtual peers. She has struggled to create virtual peers that look ambiguous – i.e., that can be perceived as either male or female; Caucasian, Latina/o, or African-American. In one study she presented a picture of a virtual peer that children did not think was African-American. However, when the same virtual peer told a story using syntactical features of African-American Vernacular English, children thought the character was African-American.

Professor Cassell ended her talk by sharing a fascinating glimpse of her latest research, on how students with Asberger's syndrome can learn social skills from interacting and playing with virtual peers.
-- Jody Clarke, Ed.D. candidate