Trump, Brexit, Cambridge Analytica... In the last few years, we have had to confront the consequences of the use and misuse of data science algorithms in manipulating public opinion through social media. The use of private data to microtarget individuals is a daily practice (and a trillion-dollar industry), which has serious side-effects when the selling product is your political ideology. How can we cope with this new scenario?
Report
Share
Report
Share
1 of 36
More Related Content
Data Natives Munich v 12.0 | "Political Data Science: A tale of Fake News, Social Bots and the post-truth era" - Juan Carlos Medina
1. Political Data Science: A tale of
Fake News, Social Bots and
the post-truth era
Juan Carlos Medina Serrano
4. • Automated social media accounts that try to look like real
users by copying their behavior
• They decide automatically which users they follow and which
tweets they retweet.
• Generate texts, some could be original
• Purpose: Manipulate public opinion
• Part of computational propaganda: Data driven influence ops
5. • Automated social media accounts that try to look like real
users by copying their behavior
• They decide automatically which users they follow and which
tweets they retweet.
• Generate texts, some could be original
• Purpose: Manipulate public opinion
• Part of computational propaganda
Data driven influence ops
6. What are their effects?
• Massive spread of news
• Amplification and manipulation of topics
• Spread false information: Fake news and conspiracy
theories
• Shape the environment we are
engaging with
7. How to recognize them? (TWITTER)
• Only retweets https://twitter.com/e_pitzky
• No retweets, but copied headlines https://twitter.com/wurmreiter
• Tweets per day https://twitter.com/VonSchwer (~ 185)
How to recognize them? (TWITTER – Data Science)
• Anomaly Detection!
1 ) Machine Learning Methods (Classification)
2 ) Heuristic Methods
• Botometer:
Features: Network, user, temporal, text features
8. How to recognize them? (TWITTER)
• Only retweets https://twitter.com/e_pitzky
• No retweets, but copied headlines https://twitter.com/wurmreiter
• Tweets per day https://twitter.com/VonSchwer (~ 185)
How to recognize them? (TWITTER – Data Science)
• Anomaly Detection!
1 ) Machine Learning Methods (Classification)
2 ) Heuristic Methods
• Botometer:
Features: Network, user, temporal, text features
17. Russian Bots?
• 2752 accounts closed by Twitter that “were related” to
Russia’s IRA (Internet Research Agency). List compiled and
released by the U.S. congress
• Badaxy, Ferrara and Lerman (2017) showed that conservatives
retweeted Russian trolls 31 more often than liberals.
18. Russian Bots?
• 23,595 tweets from 458
accounts in our German
data
• 98 accounts were
tweeting in German
29. • Propaganda and conspiracy
theories
• Hate and false information
propagates faster on social media
• Disinformation campaigns
• Types:
• Deceptive News
• Serious Fabrications
• Large-scale Hoaxes
• Satire
30. • How to define the problem?
- Content-based approaches
- Visual
- Text
- Headline/Text coordination
- Social context approaches
• How to find good training datasets?
Fake News Detection
31. • How to define the problem?
- Content-based approaches
- Visual
- Text
- Headline/Text coordination
- Social context approaches
• How to find good training datasets?
Fake News Detection
• Classification Algorithms
• Visual: CNN
• Text: RNN
• Text: Knowledge Graph
• Propagation: Graph Algorithms
Edgar Welch, An oddly disproportionate share of the tweets about Pizzagate appear to have come from, of all places, the Czech Republic, Cyprus and Vietnam,