,
8 tweets,
2 min read
Read on Twitter
Yesterday Google published a 30 page white paper about how they fight disinformation across Google Search, Google News, YouTube and Google Ads. I read the paper and there's lots of good information in it which I'll summarize in this thread.
blog.google/around-the-glo…
blog.google/around-the-glo…
First point, there's no silver bullet in addressing disinformation
- verifying accuracy of breaking news in real time is hard
- real people have different perspectives on seemingly simple issues
- enforcement needs to be predictable to users & enforcers which is hard given above
- verifying accuracy of breaking news in real time is hard
- real people have different perspectives on seemingly simple issues
- enforcement needs to be predictable to users & enforcers which is hard given above
Google uses three approaches to address disinformation across it's products
1. Make quality count by ranking the most useful information highest
2. Investing in systems to counteract bad actors
3. Giving users context about content they are consuming via labels & disclosures
1. Make quality count by ranking the most useful information highest
2. Investing in systems to counteract bad actors
3. Giving users context about content they are consuming via labels & disclosures
Google also tries to to boost the entire information ecosystem by doing the following
- $300M pledged to help newsrooms over next 3 years
- partnering with fact checkers
- research to detect deep fakes
- Protecting elections by training campaign officials & with product features
- $300M pledged to help newsrooms over next 3 years
- partnering with fact checkers
- research to detect deep fakes
- Protecting elections by training campaign officials & with product features
Google News & Search anti-disinfo
- use ranking coupled with human evaluators of their results to assess site quality
- action to remove spammy sites, both via automatic detection & user reported
- don't personalize content based on user interest
News has more stringent policy
- use ranking coupled with human evaluators of their results to assess site quality
- action to remove spammy sites, both via automatic detection & user reported
- don't personalize content based on user interest
News has more stringent policy
YouTube anti-disinfo efforts
- Only remove content if it violates community guidelines
- Change recommendations philosophy from Growth to Responsible Growth
- Dedicated tabs for breaking/top news show trusted sites & wikipedia excerpts on some videos
- Demonetize bad actors
- Only remove content if it violates community guidelines
- Change recommendations philosophy from Growth to Responsible Growth
- Dedicated tabs for breaking/top news show trusted sites & wikipedia excerpts on some videos
- Demonetize bad actors
Google Ads disinfo efforts
- require user verification to run election ads
- dedicated teams working on detecting foreign political influence ops
- policies ban misrepresentation and inappropriate or unoriginal content
- "Why this ad?" label shows users why they are seeing an ad
- require user verification to run election ads
- dedicated teams working on detecting foreign political influence ops
- policies ban misrepresentation and inappropriate or unoriginal content
- "Why this ad?" label shows users why they are seeing an ad
The whitepaper was an approachable & comprehensive overview of Google's anti-disinformation effort with 49 additional references in footnotes for more detail.
Ads was a bit light & YouTube story was a bit rosier than reality but overall a good effort.
blog.google/documents/33/H…
Ads was a bit light & YouTube story was a bit rosier than reality but overall a good effort.
blog.google/documents/33/H…
