Services like Twitter, Google, and Facebook use what are called 'collaborative filtering algorithms', which create predictions about what you are likely to find interesting based on your online identity. These algorithms are designed to prioritize information which is relevant to you, however, they do so at the cost of limiting your exposure to new ideas. This is the Filter Bubble, and it stunts your ability to grow your mind and challenge your understanding of the world.
In February of 2014, Facebook launched a smartphone app called Paper: Paper is a digital newspaper constructed using information gained through monitoring your Facebook activity. While it is one of the most overt examples of a filter bubble, it is far from one of the first. Google, Bing, and Yahoo have been filtering your search results and news in more subtle ways for years. Facebook filters what appears on your wall and Twitter uses a similar formula to suggest who you should follow. Ebay and Amazon employ filtering algorithms to recommend products. Netflix, Hulu, and Youtube try to predict which videos you'll want to watch next. All the while, ad networks follow your actions on many of these sites, collecting data to decide what advertisements to show you.
Bubbles like these are all over the web, so I'm developing dfiltr to pop them.
dfiltr is a discussion platform which uses collaborative filtering to pair you with people who will constructively challenge your views, pitting the Filter Bubble algorithm against itself.
When you hit the front page you're given a choice between discussing a completely random topic, restricting the topic to a set of tags, or writing your own.
When using tags, entering a + before the tag will require that the selected topic contains the tag in question, entering a - before the tag will omit any topics with the respective tag, and entering a tag with neither a + nor - prefix will prioritize topics which have it.
Next, you engage in a discussion about the provided topic. Discussions support a modified version of markdown. This allows you to emphasize text with bold and italics, embed images and videos, and add monospaced 'code' blocks. After the discussion, you are asked to give feedback about your partner. You can give a delta, a congruence, or a flag. These ratings inform the algorithm about who you should be paired with next.
- delta: Indicates that your partner changed your view.
- congruence: Indicates that you and your partner began the discussion with the same view.
- flag: Indicates that your partner is malicious or is not attempting to contribute to discussion.
Additionally, deltas are stored on the recipient's profile page. You can share your delta score with your friends, family, and classmates, as well as check out global high scores. You can view previously held conversations, and give those deltas, allowing for the most insightful discussions to rise to the top of the global stats.
Finally, you can earn badges for completing specific tasks, such as awarding someone else a delta, coming up with a good idea for a discussion topic, or finding a bug in our code. Badges come in blue (easy), yellow (medium), and red (hard). There also might be some super secret badges in other colors... don't tell anyone about those, though. Only people who read all the way to the bottom of this page are allowed to know about those.