Bitpress: An Open Protocol for Tracking the Credibility Of News
The only thing worse than the broken business model for news is the wholesale destruction of trust and credibility.
It took 500 days for foreign governments and unchecked social networks to destroy 200 years of hard-earned credibility of the most trusted news organizations on the planet. Today, we’re doing something about it by launching an open protocol that ranks the trust of every article by comparing cross citations between news organizations.
The internet and social media have nearly eliminated the cost of publishing content, making it possible for bad actors to spread misinformation for profit or to advance a political agenda. The business model for news has broken. Online publishing cannot support quality journalism, the institution that has historically been trusted to produce and vet credible news.
Bitpress is the first decentralized trust protocol that tracks the credibility of news by enabling journalists and news organizations to identify bad actors and misinformation. Built on a blockchain, this process is completely transparent, tamper-proof, and immune to censorship by any government or organization.
Why the Blockchain?
Blockchains are natively transparent and decentralized, both key components for any successful trust protocol.
1 Any algorithm that ranks the trustworthiness of a publisher or authenticates a source needs to be clearly understood. Facebook’s attempts at eliminating fake news have largely been inside a black box, and don’t seem to be working. Ironically, Facebook and other organizations trying to address the misinformation problem will ask us to trust that they have our best interests at heart. But incentives are a powerful motivator, and Facebook’s incentives are to maximize engagement, not maximize to truth.
2 Any successful solution also has to be immune to manipulation and censorship — requiring it to be independent of any government or commercial entity. Using the decentralization natively supported by the blockchain insures that no one person, company, or government agency can be responsible for the determining what is authentic or true.
If this all sounds like a bunch of jargon forced together in an effort to impress crypto investors and anarchists, take 30 seconds and check out the blockchain explained in about three tweets.
How Bitpress works on the Blockchain
In the same way that Bitcoin uses this blockchain technology as a way to exchange money, Bitpress uses this technology as a way to exchange trust.
Today, a link from one publisher or article to another article is typically seen as an endorsement — this is basically how Google’s PageRank works: every time someone links to a website, Google counts that link as a vote. Similarly, if the New York Times publishes an article that links to an article run by the Washington Post, it’s generally assumed that The NYT trusts that source.
Bitpress uses these links to construct a network of cross-validations between all media organizations. Through this network of validations, we established measurements of trust for both a publisher and the article that looks something like the below visualization.
This network of article links is then run through the TrustRank algorithm, which is a modified version of PowerTrust — invented to rank people exchanging files in a peer-to-peer file sharing platform like BitTorrent.
Every publisher gets a TrustRank, based on the number of credible articles that they produce. The more TrustRank they gain, the more they can give to other publishers.
Then, to get trust scores for articles, we examine how these articles link to each other like in the visualization above. In a similar way, we take that network of article links, multiply it by a publisher’s TrustRank and push it through a modified version of PageRank.
Finally, everything is written to the blockchain, so that anyone can examine our findings and get to the same result.
The same methods that the blockchain uses to protect money with Bitcoin, work for protecting TrustRank.
How Bitpress differs from PageRank
Google’s PageRank algorithm was conceived by studying the relationship between the number of times a paper in the Stanford Library was cited and its importance.
Bitpress differs from PageRank in two important ways:
1 PageRank assumes that all links between articles are an endorsement, but in the real world links between articles can mean many different things such as disagreement, newsworthiness, and novelty. In order to account for links that do not signify endorsement, Bitpress allows journalists to indicate the context of their links.
Where positive links add trust to the ledger, negative links add distrust.
We categorize distrust in seven different ways: Satire or parody, misleading content, imposter content, fabricated content, false connection, false context, manipulated content.
Once a negative link is posted and corroborated by existing trusted news sources, users on social platforms or anywhere on the web could see warnings about the false content.
Additionally, platforms like Facebook and Youtube could consume Bitpress however they like, opting to embed their own warnings to their users.
While research has revealed a certain cohort of conspiracy-focused users who seek out misinformation and believe it to be true, all platforms would have the option to automatically reject content that fell below a TrustRank threshold of their choice.
Both the transparency and decentralization of the protocol makes it open and trustworthy to users.
2 Another way that Bitpress differs from Google’s algorithm is that PageRank treats all links equally, even those from potentially untrustworthy domains. This popularity model is prone to manipulation, a flaw that eventually caused Google to publicly abandon its own PageRank score.
In contrast, Bitpress weighs the impact of each link according to the overall trust of the publisher.
Networks of bad actors attempting to build false credibility would not have trust, and their links would carry no weight, meaning the only way to manipulate the Bitpress protocol would be to create a legitimate news organization.
The resulting TrustRank represents collectively how much media organizations trust or distrust a publisher or article — a solution that has been supported by existing research, current academic thought, and months of hard work testing this process by the Bitpress team.
The CrossCheck Project
In just one example of this concept working in the real world, consider the CrossCheck project.
CrossCheck was a collaborative journalism project designed to find misinformation before the French presidential election. More than 100 journalists in 33 newsrooms monitored claims and rumors circulating on the social web, as well as fabricated images and videos. When misinformation became widely shared, a debunk was published on the CrossCheck website.
They found that having several newsrooms collaborate across political spectrums gave readers increased levels of trust in reporting.
Because CrossCheck represented the consensus of media, readers felt it increased objectivity, neutrality, and reliability of the reporting. Many CrossCheck members included local outlets which readers considered to be less partisan. Perceived impartiality by readers is one of the reasons the project appealed to a wide audience.
Similar to CrossCheck, Bitpress is built to provide consensus from trusted news sources, and as you will see, has already demonstrated the ability to identifying the most trustworthy news across all bias.
Next steps for Bitpress
While Bitpress was incubated inside of OwnLocal, a VC-backed company that works with more than 3,500 local news organizations, we acknowledge that it needs to be open source and a not-for-profit organization to be successful.
Starting with OwnLocal’s extensive network of news organizations and journalists, the Bitpress platform is currently testing initial assumptions outlined in our whitepaper.
Today, we are processing TrustRank for more than 200,000 news articles per week, representing more than 5,000 domains, and yielding over 10,000,000 links between news organizations.
We then use these links to create a newsgraph that algorithmically determines the importance of each domain, identifies power nodes, and outputs TrustRank for every article in our web.
Performing this algorithm, and using the resulting TrustRank to represent the size of the nodes, yields the following graph:
Looking at the results, its clear that certain news organizations who are typically considered untrustworthy, are frequently cited by some of the most trusted news organizations, increasing their score.
In particular, Breitbart (#21), Wikipedia (#25), The Daily Caller (#39), and The Daily Mail (#58) are all ranking high despite not being traditionally considered top journalistic organizations.
Looking at the links it appears that all are frequently cited by some of the top newspapers, like the New York Times and Washington Post. In some instances these citations appear to specifically be in a negative context (like these), but other times they are being cited as a primary source without any indication of a negative context.
When Larry Page and Sergey Brin first started working on Google, they were simply reverse engineering an existing system in an attempt to deliver the best information to users. But this assumption that all links should be positive or neutral has demonstrated to have its flaws, and for this reason we believe that negative linking is the pivotal feature missing from the web.
Allowing journalists to easily flag articles as misinformation, is the key to making this TrustRank work. And as we continue to test our assumptions, we will begin to open negative linking and other features to our news organization partners and journalists.
How you can help:
- Read our whitepaper and tell us what you think.
- Spread the word by clapping and sharing this post, or sign up for updates.
- If you’re a publisher, sign up to be a beta partner.
- If you’re a journalist, sign up to work with Bitpress.
After our initial testing phase slated to complete by the end of 2018, Bitpress will start writing to the blockchain and become completely decentralized — outside the control of any one organization.