If you are a Twitter user in the US, you might have seen little notes attached to popular or noteworthy tweets. These notes come from the company’s Birdwatch program, a novel approach to fact-checking that provides more context to tweets – all done by users.
The first question that pops to mind knowing that this system is peer-reviewed is who decides (and how) that a context note is right and should be attached to the tweet. Wouldn’t such a system just amplify the division often seen online? After all, social media platforms are plagued by misinformation and thrive on divisive, polarizing, and extreme content. Whistleblowers from Facebook and many other websites have stressed how that’s how these places generate engagement.
Birdwatch does something different. It uses a system called Bridging-based ranking. This method in general is about valuing positive interactions across diverse audiences. In the specific case of the context notes, the ones attached to the tweets are the ones that people from different walks of life and with different opinions all find helpful.
“In many online spaces, especially those using engagement-based ranking, divisive content can be more likely to go viral. Bridging-based ranking systems aim to overcome this 'bias toward division'. Birdwatch’s use of bridging to elevate context found helpful by people that tend to disagree is an exciting step toward a better internet — one that supports those building common ground,” Aviv Ovadya, from Belfer Center at Harvard Kennedy School, said in a Twitter blog post.
Ovadya has explored this approach in a detailed report. It is important to know that this is not just about showing opposing opinions (something that can make the polarization issue worse) but instead rewarding and ranking higher content that is less divisive. For Birdwatch, a lot of the content has been seen as informative independently of political affiliation.
Birdwatch is now available to anyone in the US and users can volunteer to take part in it. The program has been running as a pilot for over a year, and the social media company has conducted several surveys and data analyses to see its impact. When it comes to misleading tweets, there was on average a 20 to 40 percent reduction in agreement with them once the context note was attached. Also, once a note is attached, users are on average between 15 and 35 percent less likely to choose to Like or Retweet a Tweet than someone who sees the Tweet alone.
Twitter seems to be going all-in with the transparency for Birdwatch. The algorithm that powers this approach is publicly available on GitHub.