Location

MIT Media Lab, E15-341

Description

In this thesis, DePalma presents an interactive robotic agent that follows pointing gestures and gaze cues and, through a bidirectional exchange of referential gesture, directs its internal indexing system toward both well-known recognizable objects as well as novel configurations of stimuli. This research investigates phenomena in two major domains of human-robot joint attention: 1) an analysis of gesture toward foreground shaping in static scenes and 2) an analysis of the sustainability of joint attention as a behavioral phenomenon over time. This research contributes novel observations on participant fatigue, gaze manipulation, as well as finding evidence that novel, compound deictic action is used by participants within an interaction between a human, a robot, and a dynamic scene. This work critically probes our understanding of systems designed for joint attention in human-robot interaction, our understanding of gaze guiding with a synthetic agent, and memory imprinting on human partners in similar architectures and makes recommendations that may improve human-robot peer-to-peer learning.

Host/Chair: Cynthia Breazeal

Participant(s)/Committee

Brian ScassellatiJulie Shah