In the wake of the 2011 wave of civil unrest, which was widely relayed and coordinated via social networks, one option being considered by the authorities is to restrict social media use in future. However, this option might be entirely the wrong approach. Recent simulations indicate that the more people are censored, the more violent they become.

Can Social Media Help Calm Civil Unrest?

At the time of the London riots last summer, the UK government was actively considering
regulating or filtering social networks if the violence were to reoccur. According to the results
of a study by researchers from Telecom ParisTech, France and the University of Greenwich
in London, a version of which has also been published by Sage in the quarterly journal
Bulletin of Sociological Methodology, this approach would only lead to greater violence.
Using computer simulation and building on Joshua M. Epstein's (2002) agent-based civil
violence model, the researchers compared the number of peaceful citizens - called ‘agents’
in the study - with the number of those actively protesting or who have been jailed, on a
scale of censorship from 0 to 10, where 0 signifies total censorship – i.e. zero freedom of
information – and 10 where people are free to exchange information without any censorship.
The simulation shows the number of ‘agents’ who actively protest rising with the level of
filtering, going from 60 when there is no restriction to 650 under total censorship.

No censorship preferable to moderate censorship

The findings are however not just about comparing the two extremes on the scale; they also
strongly suggest that no restriction is preferable to even moderate censorship. Results show
that at level 0 half the ‘agents’ are peaceful and half are actively protesting. At level 10 –
i.e. in the absence of censorship restrictions – 64% of agents are peaceful and only 6% are
active. The remaining 30% are in prison, since the model also incorporates action taken by
the authorities, but the model also assumes that those in jail will be released after some time
and become free to start protesting again. Jailed agents appear at level 2 and grow in number
up to level 7, where they represent 42% of the whole agent population in the simulation. One
might conclude from this that the lower the level of network censorship, the more people
there will be in jail. However, between levels 8 and 10 the jailed population falls to 30% of
the whole group, indicating that moderate censorship is less effective in quelling unrest than
total absence of censorship.

A modified model

NetLogo, the modelling system used in the research, is a multi-agent programmable
modelling environment. Here the behaviour of an individual, or ‘agent’, is influenced
by a series of variables – political dissatisfaction, government legitimacy, etc – within
his environment. One of the basic variables is 'vision,' a person’s ability to scan his/her
neighbourhood for signs of police presence and/or active protesters. Greater ‘vision’ means
higher awareness of one's surroundings and a wider range of potential actions. In the modified
model used for their study, the researchers took ‘vision’ as an inverse proxy for censorship
– i.e. maximum vision implies an absence of censorship. Results show that when censorship
levels are high, agents are blinded to their context and tend to act sporadically, but if there’s
little censorship, agents will know what is going on around them and can act accordingly.