Nettime mailing list archives

<nettime> Why censoring social media would be the such a bad idea: a soc
Antonio A. Casilli on Thu, 11 Aug 2011 20:55:56 +0200 (CEST)

[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

<nettime> Why censoring social media would be the such a bad idea: a social simulation experiment about #UKriots

Dear Nettimers,
building on Christian Fuchs's excellent post on social media mobs, with my
colleague Paola Tubaro we have designed a social simulation experiment to
show why Internet censorship (as just proposed by David Cameron) would be
such a bad idea.
The complete post (including figures, tables, and code) is available here


By Antonio A. Casilli & Paola Tubaro

	?It's time we heard a little bit less about the economic and sociological
justifications for what is in my view nothing less than wanton
criminality?. (Boris Johnson, public speech London, Aug 9, 2011)

	?We are not social scientists. We have to deal with urgent situations?
(Paul McKeever, Police Federation Chairman, SkyNews Aug 11, 2011)

	?Nowadays sabotaging the social machine involves reappropriating and
reinventing the ways of interrupting its networks?. (The Invisible
Committee, The Coming Insurrection, Semiotext(e), 2009, p. 112)

--Why social media bring democracy to developing countries and anarchy to
rich ones?

O sublime hypocrisy of European mainstream media! The same technologies
that a few months ago were glorified for single-handedly bringing down
dictators during the Arab Spring, are now at the core of an unprecedented
moral panic for their alleged role in fuelling UK August 2011 riots. In a
recent post, Christian Fuchs rightly maintains:

[Quote: "Social Media and the UK Riots: ?Twitter Mobs?, ?Facebook Mobs?,
?Blackberry Mobs? and the Structural Violence of Neoliberalism | Christian

And, o! exquisite refinement in the ancient art of double standard: the
same conservative press that indignantly deplored dictators? censorship of
online communication, now call for plain suppression of entire
telecommunication networks ? as unashamedly exemplified by this piece in
the Daily Mail.

Fact is, moral panic about social media is the specular reflection of the
acritical enthusiasm about these very same technologies. They both spring
from the same technological determinism that acclaims new gimmicks and
buzzwords to smooth away the economic and social roots of unrest.
Having said that, what can we, as social scientists, say about the role of
social media in assisting or even encouraging widespread political
conflict? Very little indeed, insofar as we do not have data on actual
social media use and traffic during riots. It would take months to gather
that data ? and who can wait for so long in a media environment that spits
out ?quick and dirty? analyses by the hour?

The best approach is a more innovative one, relying on social simulation.
This is a new methodology that compares alternative computer-generated
social scenarios to detect what variables come into play in specific
social processes1.One of these variables is the use of social media to
organize flash-mobs in order to build field-awareness in urban uprising
settings. We aim to demonstrate that the more we repress and censor social
media in a situation of civil unrest, the worse the situation gets for
everybody in a given society.

--Epstein?s civil violence model (revisited)

Social scientists have been modelling civil violence via agent-based
simulation for almost a decade now2. One major contribution ? on which we
will build upon for our model ? was presented by Josh Epstein in a 2002

Agent-based simulations are like games based on very simple rules ? and
bringing forth complex results. The model basically describes a society
where there is only one type of social agent (represented by circles in
figure 1). (Before you scream at oversimplification, just ask yourself if
you feel more comfortable with the political characterization that
conservative media have been pushing in the last few days, where there are
two types of citizens: ?looters? and ?those who are ready to defend their
communities?. At least, Epstein?s standard social agent reminds us that
anyone can become a looter, according to the situation).

The agent?s behaviour is influenced by several variables. The first one is
the agent?s personal level of political dissatisfaction (?grievance?,
indicated by lighter or darker green colour in figure 1). That can lead
this person to abandon his/her expected state of quiet and become an
active protester (red coloured circles in figure 1). However, the decision
to act out ? whether it is to go on a looting spree or to burn the
parliament and overturn the government ? is conditioned by the agent?s
social surroundings (?neighbourhood?). Does s/he detect the presence of
police (blue triangles in figure 1) in the surroundings? If the answer to
this question is no, s/he will act out. If the answer is yes, another
question is asked: is this police presence counterbalanced by a sufficient
number of actively protesting citizens? If the answer to this second
question is yes, then the agent acts out. Sometimes, in an utterly random
way, one citizen gets caught by the police and is sent to jail for a given
period of time (black circles in figure 1). (Again, if you are marvelling
at the simplicity of this rule, just bear in mind how tricky it is for the
UK police to really distinguish who did what in a riot ?and how many of
these days? arrests will eventually turn out to be arbitrary).

[Figure 1 - Circles represent social agents, whose level of grievance is
indicated by lighter or darker shades of green. Active protesters are
colored red and jailed protesters black. Blue triangles represent police

Of course the model takes into account other mitigating factors, such as
the perceived risk of being arrested and government legitimacy. And of
course there is the possibility of moving from one place to another to
team up with other protesters and run havoc. We will come back to this
point because this seems determinant for the use of social media.

The main result of Epstein?s model is that, in a typical situation, civil
violence does not look like a linear process. The naïve vision of
political conflict as cumulative processes where confrontation escalates
until the government comes tumbling down is fallacious. Civil or political
unrest is what Epstein calls a ?punctuated equilibrium?. Long periods of
stability where rebellion is smouldering are followed by short violent

[Figure 2 - A typical civil unrest pattern: outbursts of violence (red
curve) punctuate long periods of stability when political tension is
building up (blue curve).]

There is another variable which is, to us, crucial for understanding
social media use to create flash mobs to deploy in a civil violence
situation: this variable is called ?vision? in Epstein model. Vision is an
individual agent?s ability to scan his/her neighbourhood for signs of cops
and/or active protesters. The higher the vision, the wider the agent?s

What we have done here is to modify the consequences of ?vision?. In the
original model4, agents and police officers move to randomly chosen places
within their vision range. We have introduced a new rule that makes agents
move to places in their vision range that are surrounded by the maximum
number of active protesters. The result of the modified simulation (if you
wish to download the code, just click here) is consistent with the
tactical use of mobile technologies by protesters in order to gain a
cognitive advantage against police forces and have a better awareness of
the field, its resources and possible weak spots.

This simple change simulates the behaviour of individuals involved in
civil unrest, using BBM or Twitter to detect, and to converge in, hot
spots. If the value of « vision » is higher (like in a situation where
online networking tools are widespread and not censored), each agent has
complete information as to what?s going on even in remote locations. If
social communication is censored, the value of « vision » is lower, and
agents have partial or non-existent awareness of their surroundings and
tend to move randomly.

--Internet censorship: a source of protracted high-level violence

Our social simulation code reproduces the functioning of a certain social
system (let?s say a city like London) over a significant period of time
(in our case 1000 time steps), for different values of the parameter
?vision? caeteris paribus (that is: while leaving the others parameters
unchanged ? cf. Table 2 in Annex 1 at the end of this post). Running the
model again and again and generating alternative scenarios, shows us the
outcomes of lower or higher values of ?vision? ? indicating the effects of
more or less censorship of social media.

Let?s have a look at the results in figure 3 (click to expand the image):

[Figure 3 - Red patterns represent number of violent protesters over time
within different levels of social media censorship: from 0 vision (total
censorship, upper left) to 10 vision (no censorship, lower right).]

As we can se, different values of ?vision? generate different patterns of
civil unrest over time. All scenarios display an initial outburst ? pretty
much what we have been experiencing in the last few days. What happens
next is influenced by the level of censorship government applies. In the
case of total censorship (vision = 0) the level of violence stays at its
maximum virtually forever. Think Egypt Internet kill switch incident in
January 2011 ? and remember its consequences on violence escalation in the
country and, ultimately, on Mubarak?s regime? The other cases correspond
to less and less censorship. Values between 1 (almost complete censorship)
and 9 (almost no censorship) correspond to different levels of protracted
civil unrest: the stronger the censorship, the higher the average level of
endemic violence over time (best linear fit represented by the black lines
in figure 3).

The last case, corresponding to perfect social agents? vision (and thus to
no censorship at all) deserves a little more comment. Apparently this
situation is characterized by incessant high-level outbursts of violence,
with peak active levels that seem to be even more significant that in
other cases. Yet, the average trend of violence over time (black fitting
line) remains low. Moreover, if we want to measure the size of violent
outbursts, looking at their peak active level is not sufficient. Looking
at time intervals between outbursts, at the duration of outbursts, and at
the level of ?social peace? between outbursts, help us discover that this
scenario is actually the best for everyone. In the absence of censorship,
agents protest, sometimes violently, nevertheless they are able to return
to significant levels of quiet (green line in figure 4), when social
unrest is halted.

[Figure 4 - In the absence of censorship, high levels of social unrest are
possible (see peaks in red line), but between uprisings social system is
able to come back to significant levels of quite (green line)]

This is the only scenario where active protest drops down to zero for
extended and repeated periods of time (cf. Table 1 in Annex 1 at the end
of this post): exactly what Epstein described as ?punctuated equilibrium?
in his initial model of civil violence. And although that does not seem to
match our wildest dreams of social harmony, it still is a situation where
citizens are free to voice their dissent on social media, to coordinate
their efforts and act about it ? albeit in confrontational ways ? while
still enjoying a higher level of quiet over time (see figure 5).

[Figure 5 - Levels of civil violence over time as function of levels of
censorship. (Higher vision means less censorship and less civil

In the absence of online censorship, social agents have vision= 10. This
corresponds to the lowest levels of civil violence over time.

--Some concluding remarks

It is not our role to pass judgments on politicians and police officers?
disapproval of ?sociological justifications? of the UK riots or on their
dismissal of social sciences as ? at best ? a luxury we can?t afford in
times of unrests. Their shoot-first-ask-later stance, although probably
moved by good intentions, can lead to ill-advised policy choices, like in
the case of Internet censorship that we have chosen to discuss here.

Of course other factors have to be taken into account to use a civil
violence model inspired by Epstein. As shown in a recent paper by Klemens
et al. (2010) rebellious outbursts are more likely given increased
hardship (the recent financial crisis does seem to come into play here).
Civil violence is also influenced by loss of government legitimacy ? which
in this case seems consistent with the unpopular budget cuts promoted by
David Cameron, not to mention the recent Murdoch/NoTW phone hacking
scandal. Finally, protest outbursts are less likely given increased
repressive capacity5. Which does not equate to the naïve argument that
?what we need is more cops? ? routinely conveyed in situations of civil
unrest. Repressive capacity, in the case of UK riots, has been about
adapting police procedures to compensate for the clear tactical advantage
rioters showed over the first few days of August ? an advantage that
seemed consistent with the increased level of ?vision? and field awareness
allowed by mobile communications, as discussed here. The growing presence
of MET police and law enforcement initiatives on Facebook, Flickr, and
Google groups can actually account for the subsequent limitation of
violent outburst by reducing the rioters? communicational advantage.

Other studies have applied social simulation to censorship in situations
of civil violence. Garlick and Chli (2009), for example, insist that
restricting social communication pacifies rebellious societies, but has
the opposite effect on peaceful ones6. Our intention is to show that the
choice of not restricting social communication turns out to be a judicious
one in the absence of robust indicators as to the rebelliousness of a
given society.

What we have tried to do is to demonstrate how, even in the absence of
empirical data, social sciences can still help us interpret how social
factors come into play, and possibly avoid trading democratic values and
freedom of expression for an illusory sense of security.


Vision levels	% time spent in quiet (no civil violence)
0	0
1	0
2	0
3	0
4	0
5	0
6	0
7	0
8	0.3
9	10.2
10	32.5
Table 1 ? % of time without riots ? corresponding to levels of vision.

Parameter	Values
Initial cop density	4%
Initial agent density	70%
Number of cops	64
Number of agents	1120
Government legitimacy	80%
Max jail term	30 time steps
Vision	0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Table 2 ? Parameters used in the model


	?1? We have presented our approach to this method in this article, in
case you wanted to give it a shot. Otherwise just go on reading this
post. &#8617;
	?2? A summary if these researches is in Amblard, F, Geller, A, Neumann,
M, Srbljinovic, A and Wijermans, N (2010) Analyzing social conflict via
computational social simulation: A review of approaches. In Martinás K,
Matika D, and Srbljinovic A (Eds.) Complex Societal Dynamics ? Security
Challenges and Opportunities. Amsterdam: IOS Press. pp. 126-141.
http://iros.morh.hr/_download/repository/Gelleretal1.pdf &#8617;
	?3? Epstein, J. M. (2002) Modeling civil violence: an agent-based
computational approach. Proceedings of the National Academy of Sciences
of the United States of America, n. 99 Suppl 3, pp. 7243-7250. &#8617;
	?4? Here we use the NetLogo version presented in Wilensky, U. (2004).
NetLogo Rebellion model.
http://ccl.northwestern.edu/netlogo/models/Rebellion. Center for
Connected Learning and Computer-Based Modeling, Northwestern University,
Evanston, IL; Id. (1999). NetLogo. http://ccl.northwestern.edu/netlogo/.
Center for Connected Learning and Computer-Based Modeling, Northwestern
University, Evanston, IL. &#8617;
	?5? Klemens, B., Joshua M. Epstein, J. M., Hammond, R. A. & M. A. Raifman
(2010) Empirical Performance of a Decentralized Civil Violence Model, The
Brookings Institution, Center on Social and Economic Dynamics Working
Paper No. 56,
	?6? Michael Garlick and Maria Chli (2009) The effect of social influence
and curfews on civil violence. Proceedings of The 8th International
Conference on Autonomous Agents and Multiagent Systems ? Volume 2, AAMAS
?09, Budapest, Hungary: International Foundation for Autonomous Agents
and Multiagent Systems, pp. 1335?1336,
http://portal.acm.org/citation.cfm?id=1558109.1558281. &#8617;

To cite this post: Casilli, Antonio A. and Paola Tubaro (2011) Is a social
media-fuelled uprising the worst case scenario? Elements for a sociology
of UK riots. Joint post Bodyspacesociety/Paola Tubaro?s Blog, August 11,

#  distributed via <nettime>: no commercial use without permission
#  <nettime>  is a moderated mailing list for net criticism,
#  collaborative text filtering and cultural politics of the nets
#  more info: http://mx.kein.org/mailman/listinfo/nettime-l
#  archive: http://www.nettime.org contact: nettime {AT} kein.org