The document discusses using an agent-based computer simulation model to test how different levels of censorship through social networks, represented by the "vision" parameter, impact the pattern of civil unrest; the model is based on Epstein's civil violence model and tests how unrest emerges from the interactions of agents with different grievance levels and behaviors; the results show that lower censorship through higher vision allows for longer periods of social peace compared to higher censorship with lower vision.
Changing the pattern of unrest: Social media and social networks in the UK riots
1. Changing the pattern of unrest:
The role of social media and social
networks in the UK riots
Antonio A. Casilli, Yasaman Sarabi
and Paola Tubaro
9th UK Social Networks Conference
London 28 June 2013
3. Introduction
The political role of online social networks
● Did use of online social networks (Twitter, Facebook, BBM) fuel the riots?
● Ambiguity in public discourse: networks as instrument of democracy (Arab
spring) or criminality (London)?
● Cameron: shall we shut the Web to stop the violence?
⇒ Would censorship work?
(assuming it is technically, legally and economically possible).
4. Method
Agent-based computer simulation
● Generate socially consistent scenarios on a computer;
● Compare their outcomes;
● Detect and assess variables coming into play within the social process under study;
● Identify sufficient conditions for a macro phenomenon to emerge from interaction
of micro behaviours.
● An aid to perform a thought experiment
5. Method
The logic of an agent-based model
● Generating an artificial population of agents in an environment;
● Endowing them with basic rules of behaviour;
● Letting them interact for a certain time and step aside;
● Observing outcomes at the system level at the end.
6. Epstein's civil violence model
● An environment (a city?);
● One type of agent (circle);
● Agent have different levels of “grievance” towards
government (shades of green);
● Some decide to go active (red circles);
● Depends on level of grievance + presence of other
actives vs. cops (blue triangles) around them;
● Cops pick up randomly an active agent in their
surroundings and arrest it (black circles). Source: Wilensky 2004.
7. Epstein's civil violence model
● Model also takes into account:
○ Government legitimacy;
○ Individual perceived risk of arrest.
● Agents can move on the social grid.
● “Vision” variable: agent’s ability to scan his/her neighbourhood for signs of cops
and/or actives.
⇒ The higher the vision, the wider the agent’s range.
8. Epstein's civil violence model
Epstein’s main result
Civil violence is not a linear process:
⇒ Punctuated Equilibrium: periods of stability followed by short violent outbursts
(red curve), while political tension builds up (blue curve).
Source: Epstein 2002, Fig. 8.
9. Epstein's civil violence model
Revisiting Epstein’s model
● Epstein’s simulation: agents move randomly to an empty place within their
neighbourhood (vision range);
● Our idea: allow agents to scan their neighbourhoods more effectively and move to
areas where there are more actives;
● Idea that online social networks give a competitive advantage against the police
(awareness of the field).
⇒ With agents endowed with greater moving power, what are the effects of
censorship?
10. Epstein's civil violence model
Revisiting Epstein’s model
Censorship represented through different values of “vision” parameter:
Low vision = high censorship;
High vision = low censorship.
⇒ How do model outcomes vary with different levels of vision?
11. Results
Tests
● Run the simulation for different values of parameter vision (1 to 10);
● Do so over a significant period of time (1000 time steps) for each parameter value;
● Observe results at the end.
12. Results
Red patterns represent number of active protesters over time with different levels of censorship:
from 0 vision (total censorship, upper left) to 10 vision (no censorship, lower right).
14. Results
How to interpret these results
● The pattern of violence changes with censorship;
● Absence of censorship does not totally eliminate violence;
● But it allows for periods of social peace to appear, which is never the case with
censorship:
⇒ It may not pay to trade civil freedoms for security!
15. Future developments
Work in progress
● Implement the model on a network rather than a grid topology.
● Taking out the cops
○ The estimated arrest probability of each agent depends on the number of
active (A) and jailed (J) agents in its range of vision - instead of the number
of A and cops (C).
16. Future developments
Work in progress
● The notion of "distance" and how it is related to "vision".
? Computationally heavy?
● How the network structure changes over time.
● Testing the effects of different values on different network structures to see what
happens.