In the last years we have witnessed an explosion of popular uprisings in many countries. It is not the first time that the world witnesses popular unrest, but what is new is the avenues for civil mobilisation used. Online social platforms have been at the centre of protests as channels for free expression, sources of information and tools for coordination.
Social media must be important force if governments strive to curb its freedom. And yet, revolutions and turmoil do not happen in Twitter, nor because of Twitter, just as the printing press did not cause the Protestant Reformation. But the online world is a faithful correlate where scientists mine information within which can gives important clues about these social movements.
This is where Complexity Science (CS) comes in. At the crossroads of Physics, Computer Science and Sociology, CS helps us capture some of the mechanisms and dynamics that have fed the so-called Twitter Revolutions around the globe. To do so, CS exploits the unprecedented availability of data from our ever-growing digital footprint, especially social media sites.
CS tools have given us the opportunity to look at social movements in Spain, Egypt and Brazil from a different perspective. Using Twitter data, we are able to follow how these movements grow, develop, interact and solidify.
The first lesson we learn from different examples of social activism on Twitter has to do with their growth and evolution. Unlike top-down designed systems, these networks evolve towards increasing complexity without the intervention of external forces. This means in practice that heterogeneity emerges in such structures, where a few users concentrate most attention (hubs).
|People & Power – The Indignant|
Such highly unbalanced scenario results in non-linear behaviour and an extremely efficient scheme. All of these features show up very clearly in the case of Spain’s 2011 protests.
Three interesting consequences follow. First, this growth pattern and final structure can be found in different means of communication, such as social media, mobile connections, the internet, etc. That’s the reason why an interdisciplinary approach is necessary.
Second, elites appear even in grassroots (horizontal) movements that strive for a leader-less organisation. The fact is that hubs do emerge, in the case of social networks, because the self-organised system needs to become increasingly efficient in terms of communication dynamics. With a small set of influential users, the rest of the individuals are just two or three hops away from each other. That is, the system becomes a “small world” in which reaching the global scale is feasible.
Finally, the emergence of super-connected users also hints at well-known biological and cognitive constraints: Attention is a scarce resource, and therefore only a few users get the attention of the majority and are able to coordinate information and actions.
The downside to hub-dominated structures is their weakness in front of attacks: the removal of a few selected users can lead to communication breakdown. And yet, networks stand as adaptive and versatile structures, with great plasticity to rewire connections and regain robustness.
Information flow, cascades and the brain metaphor
But political protest networks in Twitter (or elsewhere) can be studied beyond the question of how they grow and settle. The revolutionary power of social media stems from the ease with which “everybody is an influencer” (or can potentially become one). We are witnessing the transition from broadcasting communication (mass media: one-way, one-to-many flow) or personal (telephone: bidirectional, one-to-one flow) towards an unprecedented many-to-many paradigm.
This evolution has a number of deep consequences when a protest is brewing. It drastically reduces the cost of communication in terms of time, money, energy, etc. But most importantly, it facilitates the existence of cascading/viral events (anything from a single tweet to a shared video to an actual protest), which can be triggered from just about anywhere .Obviously “central” users (those with many connections) can trigger them, but so can “ordinary” users – the true “hidden influentials“.
Modular structure and polarisation
Protests and demonstrations point of course to some popular discontent, which in turn is a signal of some type of polarisation in society. Can this be measured with social media data? And more importantly, can we extract any new understanding of it? The answer is yes.
|Activity volume for specific hashtags in Egypt, from Jun to Sept 2013. Above, dominant hashtags in the pro-military intervention group. Below, dominant hashtags for those opposing it. Source: QCRI
A general observation when scientists study complex networks is that these are highly modular(i.e. they are organised in groups or communities), and political turmoil in social media is not an exception. With the aid of costly algorithms, it is possible to detect dense, cohesive communities: groups of users who connect more often within the community, than across groups.
This information is useful not only to provide an interpretable map of data, as outlining communities can tell you a lot about dynamic barriers between them. It is also important because it provides a useful key to understanding complex systems: Dynamics affect structure (the activity of similarly minded individuals tends to bring them together in the network), and structure constraints dynamics (since connections are organised in groups, activity happens within their boundaries).
In the case of political communication networks, the existence of modules can be interpreted as partisan preferences. Let’s take Egypt as an example. Both during the Arab Spring (2011) and President Mohamed Morsi’s ousting in 2013, analysis of the Twitter stream showed that (online) society was split in two: secularist vs Islamist sides in 2011 and on; pro- vs anti-military intervention in 2013. Remarkably, this can be inferred from the content of millions of tweets, but also from the network structure these tweets build up.
There are also traces of such modular organisation in Spain’s “Indignados” movement, with many communities representing partisan interests but also other factors: geographical proximity (protests were organised as camps in main cities) or opinion alignment (for instance, with many users following some preferred newspaper or TV channel because they agree with the editorial line). This can be explained by the fact that the protests attracted a rather heterogeneous crowd, from all walks of life, who were, nonetheless, united by their anger at the government’s spending cuts and high unemployment rate.
The geographic ingredient, diffusion and sensitivity
A different approach to political protests is to introduce the geographical component. We know from past experience that collective action progressed from one place to another like a wave front (the advancing front of an event, substance etc that spreads out in space and time) at a relatively slow pace. This behaviour can be well captured by the mathematical equations in diffusion models. Is that the case anymore?
|“Who-steers-whom” summary in the Brazilian protests of 2013. Red colour indicates dominance. Circles size is proportional to the activity recorded in that area.The largest demonstration took place on June 13(shaded area). Source: QCRI
With next-to-nothing time-scales (events and reactions to events happening in minutes) and reach not limited by geography, turmoil may appear anywhere, at any time. One cannot pursue anymore the “protest front”, but can instead focus on the intertwining of multiple sources of disturbances. Thus we can now tracehow a clash with the police in Rio de Janeiro elicits a strong reaction in social media in Rio Grande do Sul and provokes a street protest.
This geographical information is complemented with chronological data, such that it is possible to track the evolution of these event-and-reaction chains across time. Whole cities can be followed instead of individuals, and inter-area “who-leads-whom” relationships can be disentangled. From the resulting “movie” (as opposed to a frozen snapshot) we learn that typically protests begin in a rather centralised way – a minority of areas fuelling a yet-to-grow protest – and become highly decentralised and fluctuating as soon as they explode. This temporal approach gets closer to, but leaves yet unsolved, the possibility of forecasting or prediction.
As protest waves continue to sweep through the world, many questions about them remain unanswered. Complexity science can help us get a closer and more comprehensive look at them and perhaps can gives us the opportunity to observe in real time the rise of a new kind of social unrest that the world has not seen before.
Javier Borge-Holthoefer is a member of the Social Computing group at the Qatar Computing Research Institute (QCRI). Founded on interdisciplinary Physics, his research is focused on complex systems ranging from cognitive dynamics to social networks.