[Update: Of course, right after I publish this, I see the news that my6sense is launching a Chrome extension that will prioritize your native Twitter stream on the web. Looks cool, and I think it provides an answer to one of my criticisms below. If something like my6sense can sit on top of all my browsing, it will learn more about the diversity of my interests.]
I’ve been using an Android app called my6sense. You hook it up to your Twitter, Facebook, and Google Reader accounts. As you use it to browse those feeds in a single stream, it starts to learn what you’re interested in. Over time, it reorders your stream based on what it thinks you’ll be most interested in. It also starts to feed in content from other sources that might not necessarily be in those original streams, but is related to your interests. It’s pretty magical.
This sort of system is often predicted to be the future of news consumption: knowing our interests and reading behavior, tools will start to tell us what we should be reading on those subjects. But there’s a problem. By design, the longer you use it, the more the technology relies on you to teach it that you’re interested in new or other things. In fact, it encourages you to isolate yourself into certain interests by feeding you more and more content from things you’re already interested in. For example, my6sense has picked up on the fact that I read a lot about Google. So, usually, the first 10-15 links it gives me relate to Google. That’s not all I’m interested in, of course. A related problem is that my6sense can only use my activity on its own service to learn about me. It doesn’t know that I’ve been reading tons of articles about Egypt on the New York Times website, and it doesn’t know the articles I read via Techmeme on lots of different tech topics besides Twitter.
What I want is a system that not only knows what I’m presently interested in based on all of my online activity, but also knows what I might be interested in if prompted. Some believe that social connections are the best hook to solving this problem, the idea being that you’ll be interested in things you’re friends have read because you likely share some interests. This is a perfectly valid theory and approach. But there has to be a technological solution as well. Shouldn’t there be a way to look at my content history, figure out what I’ve been interested in, and then deduce from all that information something in a whole new category that I will likely be attracted to?