Happy Flu
In July 2008, I launched Happy Flu with Matthieu Latapy and Jean-Philippe Cointet, a viral experiment to measure the information diffusion on social networks.
The Happy Flu experiment aimed to observe how an information may spread among web pages. Anyone who encountered the experiment on a blog page they visited had the opportunity to relay the widget on their own blog. We then registered the date at which this new participant was involved, and on which page they had discovered the widget, thus tracking how the experiment spreaded.
Visualizing the spread of Happy Flu.
In the animation above, each node is a web page involved in the experiment. The counter shows the number of seconds elapsed since the beginning of the experiment, and the nodes appear accordingly. The size of a node is proportional to its degree (number of other nodes linked to it), and the width of links is proportional to the degrees of its extremities. Moreover, the color of nodes is a function of their arrival time: the first ones are white, and the last ones are dark gray.
We observed a counterintuitive result as there was no apparent correlation between the popularity of a blog and its influence. However, the influence was correlated to the duration the resource appeared on the website.
Fellows
Since the beginning of 2010, I have been studying the notion of (overlapping) communities in social networks, their automatic detection and how information can be extracted from structural properties in order to consolidate semantic data.
Recently, I have introduced a new metric on subgraphs, the cohesion, which quantifies the community-ness of a set of nodes in a network. In order to evaluate the quality of the cohesion, on February 8th, 2011, I launched Fellows with Éric Fleury and Guillaume Chelius. Fellows is an experiment on Facebook which automatically computes social communities from users' friends.
Slides introducing the cohesion and presenting Fellows.
In September 2011, Éric and I finally managed to prove that finding a group with maximum cohesion in a network is NP-hard. Current work focuses on heuristics to attempt to approximate the optimization problem. In parallel, we are also looking at possible extensions of the cohesion to weighted networks, uses of the cohesion in mining interests from communities and finally exploring links between cohesive communities and influence.
Who did I forget?
In January 2012, I launched Who did I forget? in order to showcase an algorithm based on the cohesion. It is an app which connects to Facebook and allows the user to generate the guest list for an event by only entering a rough sketch of who they want to be present.
Screenshot of Who did I forget?.
I was lucky enough to be able to convince a great designer and friend of mine, David le Pichon to design the interface. He came up with a vivid atmosphere which echoes perfectly the idea of the app: simple and festive.