Dynamic Network Models for Baboons and the Wikipedia

• Activity Number: 379
• Type: Invited
• Date/Time: Tuesday, August 3, 2010 : 2:00 PM to 3:50 PM
• Sponsor: General Methodology
• Author(s): David Banks*+
• Companies: Duke University
• Keywords: latent space ; social networks ; bag of words

Abstract

Latent space models for dynamic networks have the potential to provide new insight into critical mechanisms for growth and change. This talk uses them in two situations: grooming interactions for baboons in Amboseli, and the evolution of the Wikipedia. For the baboon data, we want to understand whether, after accounting for kinship and dominance, there is residual structure---this would imply the existence of something like baboon politics. Also, baboon troops split every 15 years or so---a good dynamic model should predict the post-split memberships. And, in a different context, we use a dynamic network model with a lexical latent space to predict where new links and new entries will appear within the Wikipedia. Although one might hope that similar models might apply in both situations, our experience is that each application demands one-off research.