top of page


Collective behavior, social networks, & information flow in complex societies

How do societies organize themselves? I am interested in how individual-level behavior and interactions can result in system-level organization. Moreover, I am interested in exploring possible universal laws of societies—both human and animal. After all, humans are not the only social animals on the planet, and it seems implausible that the dynamics that govern other animal societies suddenly stop at the doorstep of humanity . Therefore, I believe a free exchange of ideas between the social and biological sciences will lead to a more complete understanding of social life.

As you will see below, my work has spanned fields—starting in the fields of collective behavior and behavioral ecology in animals to computational social science focused on social media. My work primarily uses agent-based models, where we simulate simple behavioral rules at the level of the individual and examine how this results in social organization. However, I always aim to test my model's predictions with real data. 

Political polarization & echo chambers in online social networks

What if the information ecosystem can fundamentally alter social organization? Echo chambers and politically sorted social networks are a topic of much interest in the modern political era, but explanations for how they come to be are mixed. To explore this question, I have constructed a computational model showing that organic information cascades resulting from news coverage—people reacting to the news and the reactions of their friends—can restructure the network they pass through. We find that when news coverage diverges (i.e., news outlets become more biased

and partisan), social networks naturally become structurally polarized, even without typical partisan cues. This new network structure in turn makes information spread inefficient. Moving beyond these theoretical findings, we monitored thousands of Twitter users who consume news and found that people who follow more partisan news outlets do end up in more politically sorted networks over time.

Measuring and preventing the spread of misinformation online

Fake news sometimes seems to be a increasingly prevalent problem that threatens to undermine a shared understanding of reality. While we know that fake news typically spreads farther and faster than true news, who actually sees these articles? In collaboration with NYU's Center for Social Media and Politics, I have been using Twitter data collected from over 130 top trending news articles (real and fake) to estimate (a) how many people likely saw each article, (b) how many people likely believed each article, and (c) the ideological composition of and b. In the end, we can answer questions like "are mostly liberals or mostly conservatives seeing fake news?"

Moreover, we conducted data-driven simulations to

estimate the effectiveness of platform interventions that Facebook and Twitter are taking to try to fight misinformation. Does adding fact-check labeling actual prevent people from being exposed to fake news? Does Twitter's new crowd-source fact-checking offer a promising alternative to using (slower) professional fact-checkers?

Division of labor & social networks in social insect colonies

Social insects live in complex societies possessing thousands of individuals and division of labor (DOL). While we may call the main egg layer a queen, she does not actually dictate what individual workers do; instead, workers organize themselves. So how do workers self-organize and specialize in different tasks? And how do social insect social networks emerge—networks that resemble social networks across animals, whereby individuals who are similar tend to interact more often?

To answer these questions, I built agent-based models that tested how simple behavioral rules could result in social organization. I have worked on two main projects:

  1. In collaboration with empiricists at Rockefeller University, we examined how DOL could emerge in small, simple ant societies. Moreover, we showed how such rudimentary DOL could provide early fitness benefits to group-living and provide a stepping stone to more complex societies.

  2. Borrowing dynamics known to drive political polarization, we showed that social interactions—specifically, a combination of social influence and interaction bias—could also result in self-organized DOL and "polarized" social networks. This work suggests that a common social dynamic may be organizing societies in several dimensions at once, including DOL, political polarization, and even personalities.

bottom of page