Artificial Intelligence

Branching process models for information transmission and centrality

Friday, July 07, 2017, 11:00am - 12:00pm PDTiCal
6th floor large conference room
This event is open to the public.
AI Seminar
Professor James Gleeson, University of Limerick

The spreading of information on social networks has often been modelled in a similar fashion to infectious diseases, with each infected agent passing the infection to a susceptible neighbour with some fixed probability (e.g., Goel et al., 2015). However, empirical evidence has shown that the probability of successful information transmission (e.g., a tweet being retweeted by a follower on Twitter) has a strong dependence on the connectivity of the receiving node, due to cognitive limits and the need to divide attention among incoming messages (Lerman, 2016). In this talk, we use branching process models to investigate how the differences between the disease spread and divided-attention models lead to distinctive observable features in data. We show that the structure of the network (in particular the existence of correlations) plays an important role in differentiating between the two types of contagion, and we apply our findings to create a time-dependent measure of node influence for cascade dynamics on given networks.


« Return to Events