Yesterday, the New York Times added “related content” as a test feature. The move was generally panned across the blogosphere, and for good reason: it’s not particularly useful. The interface is cluttered, and the related content isn’t really interesting.
Nonetheless, the NYT continues to move in the right direction. Related content is going to play an increasingly important role in the content world, it just needs to be structured correctly.
The amount of content being generated online is growing at an exponential pace. The very nature of having more content, means that readers should get better information and a more complete picture of any subject they read about. At the same time, because there are now many more ways for a reader to find an article on any subject, publishers that have the best related content should generate massive amounts of traffic.
In an ideal world related content would provide a list of the best articles (blog posts, comments, mainstream articles) from anyone in the world writing on the specific subjects being discussed.
The reason that no one offers this today, is that it’s virtually impossible for algorithms alone to understand and group articles in detailed and specific topics. You can find generally related articles from the last few days, but not articles about specific topics. For example, an article about the Supreme Court upholding second amendment rights should link to the best articles about this ruling, as well as the best perspectives about “Gun Control” (which might have been written three hundred years ago).
The other key issue is that “popularity” algorithms cant distinguish quality. What this means is that most times you see “related content,” it’s just a bunch of mediocre related articles, not specific and high quality articles that provide additional or more complete information.
Overall, these reasons are a big part of why related content engines haven’t taken off.
We’ve spent a lot of time working on this at Veritocracy, and have found that a big component of solving these issues involves optimally combining humans with algorithms. We use algorithms to recommended “topics” to each author, and then let the author approve the specific classifications. In return, authors get “other perspectives” for their content, as well as a whole new channel for readers to find their articles. All of this, of course, is layered on top of our personalization system, which learns which topics a given user or publisher will find interesting, and which articles with in that topic the user will consider useful, interesting and high quality.
Regardless of how it ends up being implemented (or whether it’s us or somebody else, for that matter), we are ultimately moving towards a more open, integrated Web. Related content will be a large piece and driver of this new organization. Like any marketplace, by increasing order in the market for content, everyone will end up with a better result. Better, more complete information, and more traffic for the people producing that information, are all around the corner.
Add New Comment