Social Network Analysis of Political Change -- from Elections to Revolutions

8/3/2019 Social Network Analysis of Political Change -- from Elections to Revolutions

1/59

Social Media and the

Revolution in EgyptMaksim Tsvetovat


2/59

Collaboration


3/59


4/59

A good psychological warfare executive is no different from amarket research executive, except that the PSYOP officer also

likes to jump out of planes.

Colonel Layton Dunbar, Commander

4th Psychological Operations Group

Operation Desert Storm


5/59


6/59


7/59

Internet more

ubiquitous thenelectricity

87% of world population are within reachof a cell phone tower (and GPRS/3Ginternet)

80% of world population have some formof electricity accessible


8/59

Ubiquitous Social

Media Twitter has very low barriers to entry

Fast information propagation Fluid communities Largely decentralized

Works on every device including old cellphones (via SMS)


9/59

Freewheeling Media

No established channels or pages #hashtagsused to organize information

-- but are voluntary and freeform

RT @attag:retweet -- spreadsmessages while keeping attribution


10/59

Information

Cascades

Information can cascade throughretweets, and grow exponentially


11/59

Some Statistics

Number of followers on Twitter ispower-law distributed Size of information cascades is power-

law distributed (most go nowhere)


12/59

Result....

A social media outlet that is accessibleand completely uncontrollable byauthorities.


13/59

...only a matter oftime...


14/59

Moldova, 2009


15/59


16/59

First Twitter

Revolution... Protests organized using #pmanhashtag Leaderless, real-time, free-form Up to 10,000 people on the streets of

Cisenau

100,000s participating online Results in resignation of president and

parliament and new elections


17/59

Lesson learned...

Being in the right place, in the right time,and somewhat well-connected beatscareful planning


18/59


19/59

We were in the rightplace, in the right

time


20/59

Captured Data

Analyzed 25,000,000 relevant tweets over one week Identified communities, sub-communities and mass

influencers related to events in Egypt

Analyzed who actively influenced discussionscausing messages to cascade or go viralCorrelated activity to events in the physical (offline)

world

Highlighted profiles of mass influencers whodemonstrated that when they spoke, other peoplereacted.

1


21/59

Going Viral Influence in ActionThe larger the node, the greater the reaction to messages

These people influenced action they spoke, people acted

1

UNCLASSIFIED


22/59

Mubarak

ResignsMubarqk speech

doesnt resign

EventsTracked reaction online to events as they happened in the real worldCorrelated VOLUME TIME OF DAY / WEEK / MONTH

1

UNCLASSIFIED

Daily Cycle


23/59

Online Activity vs World

Events

Correlated VOLUME SPECIFIC EVENTS

1


24/59

Justin Bieber

Twitter Followers: 7.4 Million

Reactions Generated: 294

Reactions per 1K followers: 0.04

Most Influential vs Most Followed

Wael GhonimTwitter Followers: 86,000+

Reactions Generated: 3,291

Reactions per 1K followers: 38

MASS INFLUENCER

@Weddady Mass Influencer

Twitter Followers: 6,900+



1


25/59

dd dUNCLASSIFIED


26/59

weddadyTwitter Followers: 5,500



1

UNCLASSIFIED

UNCLASSIFIED

UNCLASSIFIED


27/59

DimaKhatib

Twitter Followers: 17,396



2

UNCLASSIFIED

UNCLASSIFIED


28/59

Deena

Twitter Followers: 888



21UNCLASSIFIED

UNCLASSIFIED


29/59

@5orm


Reactions Generated: 256Reactions per 1K followers: 132

2UNCLASSIFIED

UNCLASSIFIED


30/59

SaloumehZ


Reactions Generated: 990Reactions per 1K followers: 334

2UNCLASSIFIED

UNCLASSIFIED


31/59

Farida AE




2UNCLASSIFIED

UNCLASSIFIED


32/59

Maged Butter


Reactions Generated: 487Reactions per 1K followers: 1,733

2UNCLASSIFIED

UNCLASSIFIED


33/59

Ahmed Fouad


Reactions Generated: 1,229Reactions per 1K followers: 2,323

2UNCLASSIFIED

UNCLASSIFIED


34/59

You say you want a revolution Wael Ghonim

Twitter Followers: 86,000+



2UNCLASSIFIED


35/59

How do we do this?


36/59

NCLASSIFIED

Identify Activate Measure

Gather massive amounts ofdata from postings,

comments, and tweets in

real-time (or any other

web/enterprise data

source).

Identify relevant Mass

Influencers

Engage Mass Influencersthrough:

Information Operations

(government)

Marketing Campaign

(commercial)

Grassroots Activation

(political)

Measure the impact of the

campaign engagement,

sales, sentiment, conversion,

donations, etc..

Measure message cascades,

diffusion, volume andcorrelate to offline events

UNCLASSIFIED

DeepMile uses Agent209, a proprietary

social influence tool to identify, activate and

measure influencers

UNCLASSIFIED


37/59

Looking Up the Magicians Sleeve

Continuous Data Collection EntireTwitterverse and other Targeted SocialMedia (Blogosphere, etc.)

3UNCLASSIFIED


38/59

Continuous

CollectionNarrow-Deep100% of narrow slices

Wide-Shallow3-5% of

EVERYTHING


39/59

What do we have?

~ 500 million tweets, wide-shallow Up to 12 months of history in wide-

shallow

3-5 million tweets a day, deep-narrow(for short periods of time)

~50 million user profiles (30% of activeTwitter users)


40/59

Social out-reachers Where conversationsoccur

Provide targetedcontent Drive to action

But there are millions of

conversations!

Agent 209

2

M ti t i d th i
http://www1.istockphoto.com/file_thumbview_approve/692929/2/istockphoto_692929_comics_word_and_thought_bubbles_vector.jpghttp://www1.istockphoto.com/file_thumbview_approve/692929/2/istockphoto_692929_comics_word_and_thought_bubbles_vector.jpghttp://www1.istockphoto.com/file_thumbview_approve/692929/2/istockphoto_692929_comics_word_and_thought_bubbles_vector.jpghttp://www1.istockphoto.com/file_thumbview_approve/692929/2/istockphoto_692929_comics_word_and_thought_bubbles_vector.jpghttp://www1.istockphoto.com/file_thumbview_approve/692929/2/istockphoto_692929_comics_word_and_thought_bubbles_vector.jpghttp://www1.istockphoto.com/file_thumbview_approve/692929/2/istockphoto_692929_comics_word_and_thought_bubbles_vector.jpg


41/59

Map conversation topics and theirrelationships to each other extracting onlyrelevant conversations

31


42/59


Map information diffusion processesspread of ideas influence

3



43/59


Advanced geo-location of tweets andcorrelation to in-market events and news


44/59

Diffusion Curves


45/59

Diffusion of Hashtags


46/59

NCLASSIFIED

There are but two powers in

the world, the sword and the

mind. In the long run the

sword will always be beaten

by the mind.

- Napoleon Bonaparte


47/59

We have the

best congressmoney can

buy(Mark Twain)


48/59

Campaign Finance People can donate money to directly to

candidates, but not more then $2500

People and corporations can donatemoney to Political Action Committees(PACs)

PACs spend money to supportcandidates

There is not limit, but they have to reportit


49/59

What are PACs

Same or almost same as lobbyists

Some PACs are single-candidate, butmost big ones support many candidates

National Rifle Association

AFL-CIO PAC National Right to Life


50/59

2-mode network PAC ----spends-$-for----> Candidate Valued network, value of links = $$

Bipartite graph In-degree centrality = received more

support

Out-degree centrality = spends a lot toadvance a certain agenda, or ishedging bets

F 2 d t 1


51/59

From 2-mode to 1-

mode

PAC

Candidate

PAC

Candidate


52/59

Methodology

Federal Election Commission recordsfor 2000

900 candidates, 4000 PACs, 2.5 mildonations

Use SQL to compute 1-mode networks


53/59


54/59


55/59

No more business as


56/59

No more business as

usual

#TeaParty -- 2009 #occupywallstreet


57/59


58/59


59/59

Social Network Analysis of Political Change -- from Elections to Revolutions

Documents