Facebook provides more this-like-that instead of this-oh-that! (Or relevancy, interestingingness, and serendipity)
Facebook has announced that it’s going to start adding to your newsfeed stories that you don’t know about but that are on the same topic as ones you follow. As their post puts it:
Now, when a Page tags another Page, we may show the post to some of the people who like or follow the tagged Page.
In the late 1990s and early Oughties, the size of material being indexed by search engines busted the main metrics. Precision measured how many results of a query pertained to that query — how “noisy” the results are. Recall measured how many of the pertinent results were missed by the results list. But when you are indexing hundreds of billions of pages, total recall results in a noisy list because there are so many results that you can’t find the one that’s most relevant. Thus relevancy became much more important than before.
But even relevancy doesn’t cut the mustard when you are browsing the hay more than looking for the needle. Thus, over the past ten years or so we’ve seen interestingness become important in some environments. Sorting Flickr search results by interestingness turns up some of the most striking photos.
Search for “needle,” sorted by “Interesting” at Flickr (cc-by-nc-sa dmelchordiaz)
Reddit‘s community upvoting mechanism results in a front page that reflects not precision, recall, or relevancy, but interestingness. Reddit’s front page also illustrates that when we ask for results sorted by interestingness, we apparently tolerate far more noise than with any of the other three metrics.
These four criteria obviously each have circumstances in which they have value. If you know what you’re looking for, precision counts. If you need to do a complete review of the literature, or just need to cover your backside — an “Oh crap I didn’t come across that” moment is not permissible — then recall is your friend. If you are finding your way through a new topic, then relevancy will give you a feel for the terrain. But if you want to find something that will stimulate and amaze you, click on the interestingness button.
Facebook has opted for relevancy. This makes sense for them from an economic standpoint: You will be a happy Facebooker if you are shown stuff you didn’t know about that conforms to your existing interests and values. In their blog post explaining the change, Facebook takes as their paradigmatic example showing you a post of a photo captioned “James Harden and Dwight Harden throw down some sick dunks during practice” because you “follow or like Dwight Howard.” Highly relevant. And if Facebook started showing its users posts as noisy as what you get on the Reddit homepage or from a Flickr stream sorted by interestingness, its users would likely revolt.
So, I understand how this new move makes for happier users and thus makes Facebook richer and safer.
It’s a missed opportunity for helping to break us out of our “filter bubble” — Eli Pariser’s term for always being shown items that too closely reflect our existing interests and worldview, and that therefore confirm that worldview rather than expanding it. (See Eli’s excellent TED Talk.) It would have been far more helpful if Facebook had chosen to expand our worldview through interestingingess rather than reinforce it through relevancy.
Interestingingness is the key to serendipity, a term that, like precision and recall, doesn’t scale very well. Those who call for greater serendipity are trusting too much in the randomness now that the domain of possibilities is so huge. For example, one could create a site (which means that it’s already been created) that uses truly random ways to create a set of links to Web pages. Randomized Page Roulette. But how long do you think you would spend visiting those pages if they’re truly random? The list would be serendipitous but highly unlikely to be either relevant or interesting.
So, instead of serendipity, think about how Facebook could provide us with interesting links instead of links it knows we’ll like. It could use its awesome Social Graph to guess at enticing content that is outside our normal interests. These links would would have the sort of appeal that Reddit does, especially if it were marked as a stab and what you’ll find interesting rather than as stuff FB is confident you’ll like.
These links would be a powerful addition to Facebook’s value, for nothing is more stimulating to us than the discovery of something unexpectedly interesting or, even better, the discovering a new unexpected interest.
Most important from my point of view as a non-shareholder in Facebook, it would use what Facebook knows about us to expand our vision rather than adding another brick to the walled garden of our existing interests.