Small-world networks

Everyone is familiar with the small-world phenomenon: soon after meeting a stranger, we are often suprised to discover that we have a mutual friend, or that we are somehow linked by a short chain of friends. In this talk, I’ll present evidence that the small-world phenomenon is more than a curiosity of social networks — it is actually a general property of large, sparse networks whose topology is neither completely regular nor completely random. To check this idea, Duncan Watts and I have analyzed three networks of scientific interest: the neural network of the nematode worm C. elegans, the electrical power grid of the western United States, and the collaboration graph of actors in feature films. All three are small worlds, in the sense that the average number of “handshakes” separating any two members is extremely small (close to the theoretical lower limit set by a random graph). Yet at the same time, all three networks exhibit much more local clustering than a random net, demonstrating that they are not random. I’ll also discuss a class of model networks that interpolate between regular lattices and random graphs. Previous theoretical research on complex systems in a wide range of disciplines has focused almost exclusively on networks that are either regular or random. Real networks often lie somewhere in between. Our mathematical model shows that networks in this middle ground tend to exhibit the small-world phenomenon, thanks to the presence of a few long-range edges that link parts of the graph that would otherwise be far apart. Furthermore, we find that when various dynamical systems are coupled in a small-world fashion, they exhibit much greater propagation speed, computational power, and synchronizability than their locally connected, regular counterparts. We explore the implications of these results for simple models of disease spreading, global computation in cellular automata, and collective locking of biological oscillators.