[Colloq] Hiring Talk - Drew Margolin - Words as Bonds: Linking Social Networks and Semantic Behavior - February 28th 2:50pm

[Jessica Biron]

Words as Bonds: Linking Social Networks and Semantic Behavior 
Drew Margolin 


Thursday, February 28th 
2:50-4:00 pm 
Lake Hall – 206 


Abstract : 
Words, authors, and audiences are closely related in communication research, which studies the 
complex practice of individuals’ attempts to form relationships, establish credibility, and build 
consensus. The increasing availability of textual archives and the computational techniques 
for analyzing them has created an enormous opportunity for communication research. To take 
advantage of this opportunity, however, we must find ways to link the processes described in 
communication theories to the observable artifacts and patterns presented by computational 
analyses. More precisely, this means deriving implications of these theories for the archival data 
that is available. Through such derivations, new findings drawn from these data can connect 
to ongoing studies within established frameworks. At the same time, computational social 
scientists can gain insight into the more general processes and mechanisms that produce the 
textual patterns they observe. 


In this talk I will demonstrate two such derivations. The first study applies theories of 
sustainable cooperation and social network structure to explain word choice by a community 
of physicists studying a topic in string theory. As predicted from the theory, it is shown that the 
extent to which a community draws on its historical semantic usage patterns corresponds to the 
cohesiveness of its social structure. The second study focuses on different kinds of semantic 
overlap that individuals can demonstrate. We consider three broad processes that might induce 
similarity: sharing a topic, sharing a frame, and copying from a shared source. Using theory 
from the organizational communication and framing we identify the length of the phrase or 
n-gram at which similarity occurs as a key point of variation across these similarity-inducing 
processes. Support for this n-gram model is found in an examination of the public statements of 
the U.S. Congress. 


Bio: 
Drew Margolin is a post-doctoral research associate in the College of Computer and 
Information Science at Northeastern University. His research emphasizes the co-evolution of 
community network structures and the individual behaviors that comprise these structures. 
In his dissertation, he examined the evolution of social and semantic networks constructed 
from papers published in string theory. His current work applies these community-network 
approaches to authors of and audiences for political speech. Drew holds a BA in Economics from 
Yale University and a PhD in Organizational Communication from the Annenberg School for 
Communication, University of Southern California.