Sometimes You Need to Know When to Shut Up

by Gabrielle DeMarco on September 23, 2010



As anyone who has played a good game of “Telephone” understands, the more people you play with, the more likely “Jimmy likes to eat glue” becomes something horrifying like “Jimmy loves to beat you”. The same communication failures also occur on much larger and more sophisticated networks, to much less entertaining results as researchers within the new Social Cognitive Network Academic Research Center (SCNARC) uncovered.


Gyorgy Korniss and Bolek Szymanski  are taking a deeper look into how and when communication in large networks fails. These networks could be anything from Facebook to a flock of birds.  

What they have found is that for a network to stay in synch and therefore to continue functioning, all members of the network have to immediately receive, process, and pass on the information provided to them. This seems reasonable and perhaps obvious to anyone who has impatiently waited for someone to read an important e-mail. But their findings, which used computer science and statistical physics, provide starling information on how devastating delays in communication can be.


Their models show that if there are continued delays between just two or three people in the crowd, the entire network will always eventually teeter and collapse. To top that, they found that once the network is on this path to destruction, there comes a point when no matter how hard and fast you work to make up the delays, the system will remain on autopilot to complete failure.


A cool representation of the carnage caused by delays is in these two videos.  They are an actual visualization of the mathematics they used for the research.


First, we have a system that is completely in synch. Watch as information moves through the branches of the network.  The red and green bubbles indicate a deviation from average performance. In other words, they show the addition and processing of information through the system. Watch as small bubbles percolate across the network. They never get too big. In fact, you could watch this movie until the end of time.

Next, watch a system with delays. It takes a moment for the first bubble to appear and spread and that delay causes other bubbles to grow bigger and bigger until the whole screen looks like a bomb exploding over and over. The delays quickly become too much for the system to handle.

 Check out my full article on the research here.