Exploration vs. exploitation is a tradeoff found in machine learning but it has applications to RSS readers, news and Twitter.

Specifically (and quoting from wikipedia)

“[Exploration and exploitation] is related to the problem of balancing reward maximization based on the knowledge already acquired with attempting new actions to further increase knowledge”

Another way of looking this is through music. Pandora is really good at recommending songs that I like. But eventually, it starts repeating the same music over and over again. It is exploiting what it knows about me. But I need diversity and I want it to explore other types of newer music rather than repeating Sting’s “Fields of Gold” over and over.

Traditionally, this balance between discovering new things that are interesting and playing the oldies but goodies has been maintained by people. In news and content, it was journalists.  In photos, it was the Life magazine editors. In music, it was the radio djs.


Anyone, (including yours truly) can create a blog with opinions or a upload a song to imeem or a photo to Flickr. And hundreds of millions of people are doing this in the process creating billions of pieces of content that is impossible to wade through – without filters.

The noise to signal ratio is so high that data takes an enormous effort to sift through and we’re less efficient as a result, but more people are able to have a voice.

Society is currently valuing the democratization of media higher than the value of curation and filtering. If you need proof, take a look at the revenues of the NYTimes.

Eventually, however, the pendulum will swing the other way: the noise will become so large and burdensome as to drown out the signal and people will value curation of content again. Of course, the forms of curation will differ dramatically from the previous forms. I’m excited to see what they look like.

What do you think?