Reducing misinformation through decentralization

Misinformation online is a fundamental issue that undermines a shared notion of truth, and platforms struggle to protect users from it with their current toolkits.  Auditing, annotating and removing content is a slow process. It can't be applied uniformly at scale, across all languages and geographies, and doesn't always earn users' trust due to the lack of transparency and unavoidable false positives.

Twitter's Birdwatch experiment explores the hypothesis that users themselves could be empowered to add context to misleading information collaboratively and at scale.

A solution like this has never been built by a social media company before. The principles we followed include:

• All our algorithms are open-source;
• Users are included in the design process;
• We design experimentally, and rigorously test each result, sharing learnings with the public as we go
• Our focus isn't on engagement, growth, or popularity metrics, but to find ways to deliver the highest-quality information that is trusted by the largest range of people.

During the past two years, Birdwatch has evolved into a system that reliably produces helpful context to misleading information online. Most importantly, all written and selected by people who use Twitter, for people who use Twitter.

We have data to show that labels applied by Birdwatch contributors:

• Reduce the virality of misleading content
• Increase understanding of potentially misleading topics
• Are found helpful by people from across the political spectrum

Finally, we believe that the approach taken in Birdwatch — harnessing the power and trust of the people through decentralization and transparency — can be applied to many other digital spaces. Decentralized systems can help improve online literacy, strengthen democracy, and bridge gaps in the polarized social fabric.

‍