Interdisciplinary Institute in Artificial Intelligence

The socio-communicative complexity hypothesis states that living in groups should increase the need for efficient communication in animals. Thanks to “Rookognise,” a multitask neural network designed to automatically recognize which rook is vocalizing and when, we plan to test this hypothesis comparatively—that is, in social and non-social corvids, as well as in human and non-human primates. Rookognise needs to be adapted for non-bird species and enhanced with visual recognition of individuals and behaviors. This project should help identify key factors in the evolution of language (sociality, cognition, vocal learning) that are shared by corvids and primates. It will also help identify key communicative behaviors in children and their role in the formation of social networks and, more generally, the link between vocal and social integration. Finally, this project also holds significant innovative potential for education, animal conservation, and animal welfare.