Highlight
Infant social modeling
infant social modeling
Achievement/Results
NSF funded researchers Joshua Lewis and Gedeon Deák at the University of California, San Diego, along with Dr. Hector Jasso and Dr. Jochen Triesch, have developed a simulation environment for investigating interaction and learning between infants and caregivers. The study takes an end-to-end approach to infant social modeling.
We start in the homes of our subjects by collecting hours of audiovisual data from infant/caregiver interactions. These data consist of semi-naturalistic free play sessions (see Fig 1), where caregivers are instructed to play with their infants using a supplied set of toys while the infant is seated in a tray chair. The interactions are recorded with audio from multiple camera angles. The lab has amassed many terabytes of this audiovisual data, which are coded frame by frame for both infant and caregiver actions. The codes of caregiver behavior are stored in a database and used to drive realistic behavior for our simulated caregiver (see Fig 2). If real caregivers make certain behavioral transitions frequently, then the simulated caregiver will be likely to make the same transitions. Our simulated infant agent watches the caregiver using computer vision techniques, locating the caregiver’s head and determining its direction of gaze. Over time the infant agent learns to look in directions that correspond to the caregiver’s head pose, a process called gaze following that is crucial to normal infant social development.
By better understanding how gaze following develops, we can better assist when the developmental process goes awry for real infants. In the greater context of understanding infant social development, from modeling to robotics to experimental work, we see this as occupying a productive niche between 2D models and robotic agents. We open computer simulations up to complexities that mirror those in the real world, but our learning simulations are more convenient and we can have complete control over the agent and environment. Joshua Lewis is an NSF IGERT (Integrative Graduate Education and Research Traineeship) fellow in the Vision and Learning in Humans and Machines Traineeship program at UCSD run by Professors Virginia de Sa and Garrison Cottrell. This work will be presented at the 32nd Annual Meeting of the Cognitive Science Society in August 2010.
Address Goals
Infant social development is a critical area of research and is likely important for understanding autism. Graduate students are involved in this research.
