NCS-FO:Tracking social behavior and its neural properties in a smart aviary

University Of Pennsylvania, Philadelphia PA

Investigators

Abstract

Advances in technology, mathematics, computing and engineering are making it possible to quantify behaviors within complex naturalistic environments and to relate them to underlying neural mechanisms. For social animals, which have evolved to perceive and evaluate signals within a community context, the ability to link neural function with the precise social environment is especially important and challenging. Little is currently known about how the brain integrates complex social information and how such information might be encoded. This stems in part from the experimental challenge of measuring and assessing the variables that determine a social context and then linking the state of a social network to precise neural events. This project has assembled an interdisciplinary team of engineers, neurobiologists and computational scientists to create a platform to record and evaluate brain dynamics in individual animals navigating a complex social environment. In addition to the challenge and opportunity of using sophisticated engineering and computational approaches to study how brains encode social information, this work will generate a complex dataset that will offer unique opportunities for developing novel mathematical methods to quantify and visualize social networks that can be applied to other disciplines. In this study, a "smart aviary" is equipped with arrays of cameras and microphones to create a fully automated system for tracking moment-to-moment behavioral events for each individual songbird within a social group. The songbirds are of a highly gregarious species (brown-headed cowbird, Molothrus ater) that uses vocal communications to form and maintain a complex social system. As a variable, social context needs to be mathematically constructed over multiple timescales from the sequence of all behavioral events. This entails the development of new mathematical approaches and statistical models for quantifying social network state so that individual neural events can be linked back to the precise social contexts. The project will develop new machine learning approaches for automated capture of social interactions over months-long time periods. In addition, an articulated mesh model enables visual signals to be captured in unprecedented detail, allowing the quantification of shape-mediated social signaling. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.