Just Because Your Friends Like It -- Will You?

As Facebook's "Like" button gains traction and other social media such as Twitter help us follow what our friends and other people we care about are "into" at the moment, we have to ask if you will necessarily "like" the same things that they do?  Interesting question.

It's not yet clear that your friends can point you to content or music or items to buy that you will like any better than the anonymous collective wisdom of crowd-sourcing plays like Amazon's collaborative filtering approach. In fact, a recent study published by MIT found that hearing about songs that my friends like might tease me to try them -- but not necessarily to like them or download them. Songs need to meet my taste, not just my friend's.



While social recommendation encourage the notion that I should be interested in what my friends are interested in, the "Amazon approach" solves a totally different problem, which is "How can we know what we or our friends don't know?" That means that while it's likely that my friends and I share something in common -- and that's why I follow them -- we are not exposed to vast amount of content that we may love, but are often just not aware even exists.

This discovery problem is a real global pain in multiple market segments, from music to textual content --  especially video,  where it's difficult to figure out what the content is about because there is  little meta-data around it. This is something we've been researching  for the past three years.

When a user has 5 minutes to watch a video with tens of thousands of available videos  (or even more if it's on a site like YouTube with millions of  videos) predicting what users want to watch in real time is a very difficult mathematical problem to solve. But by looking at millions of video views and even more millions of users, we are beginning to see clear patterns of behavior that make recommending what to watch next far more accurate. Not that your friend's aren't worth listening to, but with a sample base of millions, the hit rate for you "liking" what the anonymous crowd "likes" is pretty high.

Often the "discovery" process is subtle. How many times were you introduced to a song for the first time, and as soon as that happened you started hearing it everywhere: your car, at work, on TV. Do you think the song wasn't there before? It was. You were just not aware of it until you heard it and decided you liked it.

It's all about being exposed to things you really like. I think social recommendations will be a great layer to adjust "Amazon recommendations," but I'm not yet convinced it is a sustainable system on its own

Your friends simply don't have the time and/or inclination to search endlessly for videos they know you'd like. But machines do  -- and they are getting better by the minute.

So just because your friends like something -- will you?  I'm not sure yet.

1 comment about "Just Because Your Friends Like It -- Will You? ".
Check to receive email when comments are posted.
  1. Adam Singolda from Taboola, December 9, 2010 at 1:49 p.m.

    Paula, thanks for the comment.

    I believe I agree with you, and this is why:

    Things are happening all the time (and again and again, and in the mass). By definition those pieces of content were not aimed at ut originally. True #1.

    However, we may really would love to engage with them once we found out that they actually exist and available to us. True #2.

    The big challenging question is how can we discover them in practice. You've mentioned actually a technique called in mathematics / Recommendations-Engine Discipline - "Exploration". This means that some algorithms would offer you content just for the sake of seeing if you're in an exploration mood, and would like to engage with something. That is very similar to the "flip to the end of the paper" you've mentioned. True again (#3).

    This article covers different methods of discovering content, focusing on -- social versus machine based.

    I hope that helps, and thanks again for your comment.

Next story loading loading..