Collaboration on Social Networks

New research on collaborative behaviour in social networks provides insights on strategies related to coalition forming, contractual agreements and negotiation.

The research, titled "Collaboration in Social Networks", appeared in the Proceedings of the National Academy of Sciences (PNAS) Online Early Edition (20 February 2012). It was carried out by ICTP physicist Matteo Marsili of the Centre's Condensed Matter and Statistical Physics section, along with Luca Dall'Asta of the Politecnico di Torino and Paolo Pin of the Universita degli Studi di Siena.

"Why do people collaborate in situations where defection or free riding on others' efforts would be the optimal course of action for selfish individuals? This is a central issue in game theory, and one of the key insights is that, when two individuals repeatedly interact over time, collaboration might be the optimal choice, given that the opponent may credibly punish non-collaborative behaviour by defection, or collaborate conditionally on the collaboration of the opponent," said Marsili.

The extension of this insight to more complex situations where many individuals interact on a social network is non-trivial. Marsili et al. found that players can credibly threaten only a subset of the players they interact with and that punishment must be reciprocal.

The researchers found that the possible collaborative agreements turn out to be subgraphs of the social network that can be quantitatively studied with tools of statistical physics.

This shows, in particular, that when the cost of collaboration is small, the subgraphs that sustain collaboration have a local nature, but when they are high a critical mass is required for collaboration to emerge. This provides a theoretical foundation for several empirical findings in the sociology.