狠狠撸

狠狠撸Share a Scribd company logo

Techhub Riga misinformation meet-up, 15 March 2018

?
?
M
Mārti?? Vaivars

My presentation for Techhub Riga meet-up, 15 March 2018. I was talking about: (1) mapping out the misinformation problem and the chances to tackling in a scalable, automatised way, (2) a proposal for how corruption watchdogs (like Korupcijas Novēr?anas un Apkaro?anas Birojs) could try monitoring political spending on social media.

1 of 29
Techhub Riga misinformation meet-up, 15 March 2018
A leader in voice analytics and security company winners
Techhub Riga misinformation meet-up, 15 March 2018
It’s often neither fake nor news.
Demonstrably false factual claims in articles (headings, body) and social media content. Or claims without any sources or evidence. 01 - false factual claims in text.
Edited content that is presented as authentic. Generated (sometimes deep fake) content that is presented as authentic. 02 - doctored or deep fake images, video or audio.
It’s often neither fake nor news. 01 - false claims in written content 02 - doctored or deep fake images, video or audio
Content presented without required context (time, more information, other opinions). Clickbait. Cherry- picked content giving a false sense of a clear narrative. 03 - content presented in a misleading context.
Techhub Riga misinformation meet-up, 15 March 2018
Content presented without required context (time, location, additional information). Click-bait. Cherry-picked content giving a false sense of a clear narrative. 03 - content presented in a misleading context.
Value-judgements presented as news. Includes emotional language, statements of opinion and satire. 04 - opinion disguised as news.
Techhub Riga misinformation meet-up, 15 March 2018
Value-judgements presented as news. Includes emotional language, statements of opinion and satire. 04 - opinion disguised as news.
It’s often neither fake nor news. 01 - false claims in written content 02 - doctored or deep fake images, video or audio 03 - content presented in a misleading context 04 - opinion disguised as news
The real problem for automatic detection - what do we do with results? CONTENT ! RESULTS ?
Techhub Riga misinformation meet-up, 15 March 2018
(1) Regulate Facebook - make public broadcasters more visible, mess up the business model of thrash creators (2) Throw money at high quality journalism (3) Throw money at evidence-based critical thinking, media literacy training My personal opinion about this: ?
Coming in a few months. Stay tuned and leave your details at http://4facts.org/
Techhub Riga misinformation meet-up, 15 March 2018
We have a problem with political spending control when it comes to social media.
We have a problem with political spending control when it comes to social media. Mārti?? " TV, newspapers, outdoor ads # “I can audit this”
We have a problem with political spending control when it comes to social media. Mārti?? " Targeted, paid social media content $ % “Damn, can’t really see this”
The best solution by far would be cooperation by social media companies. Mārti?? " Targeted, paid social media content $ % “Damn, can’t really see this” “Here’s some data required by law”
Alternative solution - let’s capture relevant data with the help of social media users (1/3).
Alternative solution - let’s capture relevant data with the help of social media users (1/3). Text data Image, video data Links to external sources Timestamp Any other interesting meta- data
Alternative solution - let’s capture relevant data with the help of social media users (2/3). Any relevant user data that Facebook takes into account when targeting their ads Participant "
Alternative solution - let’s capture relevant data with the help of social media users (3/3). “OK, here’s something we can look into” Content data Targeting data ?
(1) We need a representative sample. Very hard with non- paid social campaigns (2) Not only desktop, but also native apps on mobile devices (3) A lot of paid political content is not created by o?cial accounts Of course, there are challenges: ?
martins@toneboard.com @martins_vaivars

More Related Content

Techhub Riga misinformation meet-up, 15 March 2018

  • 2. A leader in voice analytics and security company winners
  • 4. It’s often neither fake nor news.
  • 5. Demonstrably false factual claims in articles (headings, body) and social media content. Or claims without any sources or evidence. 01 - false factual claims in text.
  • 6. Edited content that is presented as authentic. Generated (sometimes deep fake) content that is presented as authentic. 02 - doctored or deep fake images, video or audio.
  • 7. It’s often neither fake nor news. 01 - false claims in written content 02 - doctored or deep fake images, video or audio
  • 8. Content presented without required context (time, more information, other opinions). Clickbait. Cherry- picked content giving a false sense of a clear narrative. 03 - content presented in a misleading context.
  • 10. Content presented without required context (time, location, additional information). Click-bait. Cherry-picked content giving a false sense of a clear narrative. 03 - content presented in a misleading context.
  • 11. Value-judgements presented as news. Includes emotional language, statements of opinion and satire. 04 - opinion disguised as news.
  • 13. Value-judgements presented as news. Includes emotional language, statements of opinion and satire. 04 - opinion disguised as news.
  • 14. It’s often neither fake nor news. 01 - false claims in written content 02 - doctored or deep fake images, video or audio 03 - content presented in a misleading context 04 - opinion disguised as news
  • 15. The real problem for automatic detection - what do we do with results? CONTENT ! RESULTS ?
  • 17. (1) Regulate Facebook - make public broadcasters more visible, mess up the business model of thrash creators (2) Throw money at high quality journalism (3) Throw money at evidence-based critical thinking, media literacy training My personal opinion about this: ?
  • 18. Coming in a few months. Stay tuned and leave your details at http://4facts.org/
  • 20. We have a problem with political spending control when it comes to social media.
  • 21. We have a problem with political spending control when it comes to social media. Mārti?? " TV, newspapers, outdoor ads # “I can audit this”
  • 22. We have a problem with political spending control when it comes to social media. Mārti?? " Targeted, paid social media content $ % “Damn, can’t really see this”
  • 23. The best solution by far would be cooperation by social media companies. Mārti?? " Targeted, paid social media content $ % “Damn, can’t really see this” “Here’s some data required by law”
  • 24. Alternative solution - let’s capture relevant data with the help of social media users (1/3).
  • 25. Alternative solution - let’s capture relevant data with the help of social media users (1/3). Text data Image, video data Links to external sources Timestamp Any other interesting meta- data
  • 26. Alternative solution - let’s capture relevant data with the help of social media users (2/3). Any relevant user data that Facebook takes into account when targeting their ads Participant "
  • 27. Alternative solution - let’s capture relevant data with the help of social media users (3/3). “OK, here’s something we can look into” Content data Targeting data ?
  • 28. (1) We need a representative sample. Very hard with non- paid social campaigns (2) Not only desktop, but also native apps on mobile devices (3) A lot of paid political content is not created by o?cial accounts Of course, there are challenges: ?