Identifying the mechanisms behind online diffusion

Name of applicant

Jonas Lybker Juul


Cornell University


DKK 1,148,929



Type of grant

Reintegration Fellowships


Identifying and limiting the spread of harmful content online is a most pressing challenge of our time. A major obstacle in mitigating the spread of misinformation and other harmful content is that we do not understand how online content diffuses between users in social networks. What makes content spread efficiently online? Does some content spread further or faster than other kinds of content? Why? With this project, I will seek an in-depth understanding of online diffusion and how misinformation and other harmful content can be mitigated online.


Every day many different kinds of content spread virally online. Some viral content is fun. Some is silly. Some is devastating. Although viral content can have a large impact on individuals and society, there is much we do not know about content that diffuses successfully online. What makes some content spread far and whether all viral content - fun, silly or devastating - spreads similarly remains unclear. It is also unclear how well the large body of existing theoretical models for online diffusion matches actual data. These 3 questions: What makes content spread far, whether different content spreads similarly, and whether existing models match reality are some of the questions I will confront in this project.


Epidemiologists can study chains of infection in order to understand how a virus is transmitted between people. In the same way, I will decipher how content spreads among online users by studying online chains of infection. If two kinds of content spread inherently differently, their chains of infection will look different from each other. By statistically analyzing how millions of tweets were retweeted by other Twitter users, I will test whether tweets about different topics spread differently. If some content does indeed spread further or faster than other kinds, I will use careful mathematical modeling to understand why this is so.

Back to listing page