Political Manipulation on Twitter

Download Report

Transcript Political Manipulation on Twitter

Vocal Minority vs. Silent Majority
Eni Mustafaraj
Samantha Finn
Carolyn Whitlock
Panagiotis Metaxas
IEEE SocialCom 2011 @ MIT Media Lab
October 10, 2011
Gaming
How humans can use technology to
“influence” the political process.
Examples of gaming
Vocal Minority can:
1. Frame the conversation
2. Put pressure on the media
Implications
Not all user-generated content is equal. The fact
that we have lots of it, doesn’t mean we should
use it as it is.
2008: Election of President Obama
2009: The Healthcare Reform
2009: The rise of Tea Party
Aug 2009: Senator Kennedy Dies
Special Election in MA
Martha Coakley
Scott Brown
Democrat
Attorney General
Republican
State Senator
Pre-electoral Polls
Boston Globe
01-04-2010
50%
35%
Rasmussen Reports
01-04-2010
50%
41%
Rasmussen Reports
01-11-2010
49%
47%
Twitter Data Collection
 January 13 – 20, 2010
 234,697 tweets
 56,165 unique Twitter users
Vocal vs. Silent
Silent users
(1 tweet)
Vocal users
(50+ tweets)
User Volume
Silent users
Vocal users
Tweets Volume
Political Activism
Friendship graph for 574 users with more than 50 tweets
Tweets are not made equal
Tweets are not made equal
What’s in a Tweet?
Silent Users
Vocal Users
Only Text
42.0%
8.0%
Hashtags
14.1%
53.0%
Links
29.7%
49.4%
Retweets
29.6%
60.3%
Framing the Conversation
Social Media Campaigns
Social Media Campaigns
Social Media Campaigns
115 tweets to report progress on Facebook fans
54 tweets to report progress on Twitter followers
Campaign Amplification
Targeting the Media
Political Tweetbots
Human Bots?
30 lists with tweets
2758 tweets
180 media
accounts targeted.
Pre-Fabricated Tweets
The Tweets-Factory against Media
DO YOUR JOB SHINE THE LIGHT ON ACORN
http://bit.ly/DoYourJob @ACORN Nat @SEIU
@GlobeSenateRace @wwlp #masen
WE THE PEOPLE WANT A FAIR ELECTION
http://bit.ly/acRNFraud @ACORN Nat @SEIU
@GlobeSenateRace @wwlp #masen
Conclusions
 Tweets are not equal.
 Opinion mining and predictions based on
aggregated data cannot be trusted.
Tweets Factory 2.0
Thank you
#questions