Digg Experiments with Changes to Top News Algorithm

Digg announced another series of tweaks to its Top News algorithm in its continuing attempt to refine the content that appears on users’ homepages.

More from software developer Will Larson via a post on the Digg Blog:

Our Top News algorithm has a number of levers that can be tweaked, and for several hours, we’re going to be running with a new configuration. We are looking to see if we can increase both the quantity and diversity of stories on Top News without reducing overall quality. After running this test, we will study the data and your feedback. We will probably conduct a few more tests based on the results.

Over the past couple of months, we have noticed a major uptick in comments, which we are keen on seeing continue. Toward that end, we have added a new sort for comments: the “Best” sort. (Note: An update on the Digg Blog post said this feature was temporarily disabled, pending a bug fix.)

Conceptually, this sort is a little bit similar to the existing Most Dugg sort, but rather than using the raw score, it tries to figure out a confidence interval for each comment based on its number of votes up and down. If you’re interested in more detail, we recommend reading this excellent blog post from Evan Miller.

Another change we have made to comments is how we display comments that been voted down. Previously, we completely hid the content of comments with scores below your viewing threshold (you can change that threshold in your settings). This worked well for hiding low-quality comments, but it also meant that controversial statements quickly became invisible.

We have always felt that a bit of controversy is where great discussions happen, and we have seen a number of conversations that never got off the ground because one viewpoint’s comment got voted to -1 and suddenly it had to be expanded to see if it was worthy of an upvote (it is just a little click, but hey, who likes extra clicking?).

To address that concern, we no longer hide the entire comment when it is only slightly down-voted — rather, we show a truncated version of the comment in a lighter color. Hopefully this will give rise to more great discussions!

We have also launched a number of improvements to user suggestions. In particular you can look forward to seeing fewer inactive users being suggested.

Related Stories
Mediabistro Course

Online Marketing

Online MarketingStarting November 4, learn how to launch a successful digital marketing campaign for your brand! In this course, you'll learn how to create an online marketing strategy, identify the appropriate distribution platforms, and use the web to successfully deliver your brand's message. Register now!