Over the last few years with Facebook's dominance over our consumption of current affairs, fake news has changed our political landscape and the society we live in. It's terrifying when you really think about it 🙀.
@frankyoung + @jay_khurana have created CrossCheck to give each of us the ability to check the accuracy of a post / article and discover similar suggested content.
Please tell us more about it @frankyoung@jay_khurana, I'm particularly curious about how the accuracy checking actually works.
@jay_khurana@abadesi A big difference for CrossCheck is that we work article by article, post by post, rather than whole labeling websites as fake or not. Also, you can use us on any text-based content whether its a tweet or a New York Times article.
Using machine learning our algorithm collects millions of articles from across the web in real time, extracts their content, and sorts all of them into clusters. Then these clusters are sorted into good or bad piles.
When users highlight a Facebook post or an article our product matches said post/article to these clusters to produce a CrossCheck score. This score represents the percentage of how much the highlighted content matches the "good" clusters.
I tried front page articles from both The New York Times and Breitbart. Neither had any supporting articles, but their scores were wildly different. Now I'm no fan of Breitbart, but if you have no support to back up a score, what exactly is it based on?
@alexg473 Hi Alex, the CrossCheck score is calculated based on how well the quotes, facts, and statistics of an article are broadly verifiable in all of the real time content being published right now. Supporting articles are an additional metric that allows readers to go out and find additional information for themselves, perhaps related articles would be a better term!
Report
So interesting. Going to try it out and follow up with questions.
Hustle Crew Academy
CrossCheck
Play
CrossCheck
CrossCheck
Dancing Dogs app