Brian Scassellatti, Yale University
Most work on robot learning takes place in complex, hostile environments (like the surface of Mars) or in competitive, adversarial environments (like a robot soccer match). Learning results in these environments are both very difficult to achieve and are unlike the learning environments most humans face. In this talk, we look at the problems (and simplifications) introduced by an environment that is both social and supportive. What can a robot learn in the presence of a helpful human instructor?
The engineering challenges for operating in a social environment are significant. A machine must recognize and respond to the complex array of social cues that humans effortlessly produce. Perception of social cues (including pointing gestures, gaze direction, facial expression, and tone of voice) must be paired with the appropriate generation of cues by the robot.
We will discuss ways in which the social interaction can be exploited to allow for faster, more efficient robot learning.
Social machines can also be tools in the study of human social development.
Three short vignettes will provide examples of how these machines provide information about the development of human hand-eye coordination, self-other discrimination, and pronoun learning will be presented. Finally, we discuss how these machines are being used to improve diagnosis and therapy of autism, a social developmental disorder.
Loading more stuff…
Hmm…it looks like things are taking a while to load. Try again?