The limitations of social discovery

Social discovery

“Discovery” is a hot topic these days. The curse of a new buzzword is that it’s difficult to come to a shared mental model in the early stages. Instead of tackling that large problem, I’ll start with something simpler: defining “social discovery” and suggest that social discovery is a stepping stone on the way to algorithmic discovery.

“Social discovery” has two definitions. On one hand, it’s used to mean services like Highlight that help you to find other people. However, the broader definition is services that help you find just about anything by using recommendations from friends.

Friends are, in general, good at recommending things, since they know you so well. However, this technique suffers from two limitations. First, as social networks grow to include people outside of one’s close circle of friends, discovery won’t be as accurate. Second, with a diverse stream of friends, you’ll probably have foodies who can help you discover new and interesting restaurants, but those same people may not be audiophiles who can help you discover great music.

The second problem can be overcome by adding a curation element; that is, a user can select a subset of friends who know about the topic in question. Though you avoid the high-school-friend-on-Facebook problem, social discovery+curation also has issues. The burden is left to the user: It’s difficult to curate each social discovery application to have the right set of people to generate good recommendations. Also, I may not have friends who have the knowledge I need. For example, I could certainly get good restaurant recommendations from friends in New York City, but none of my friends may ever have been to Tulsa, Okla.

What’s the solution for this? I’d like to suggest that we’ve already entered and are accelerating in an era of algorithmic discovery. Algorithmic discovery generates intelligent recommendations based your social network, broader social networks, and your own personal tastes and preferences:

  • Instead of having to worry about which of my friends are accurate predictors of what I like, an algorithm should be able to curate that list for me.
  • Discovery can happen outside of my social network. There may be someone I’ve never met who is actually an excellent predictor of my tastes.
  • Everything I’ve liked or disliked in the past will help to inform the algorithm, which goes beyond potentially limited recommendations from friends.

Think of it this way: you’ve got your own personal concierge who knows you very well, is an expert in what you like, and also knows what everyone else has liked or disliked.  That would be an incredibly powerful tool, because it harnesses me, my friends, and the broader collective knowledge.

Social discovery is a necessary step on the way to algorithmic discovery. Data scientists can only start to create algorithms when they have a large set of data about people’s likes and dislikes. With so many people sharing on Facebook, Pinterest, and Twitter, that’s a foundation on which discovery engines powered by actual user preferences can be built.

Social discovery is exciting right now, but I don’t believe this will be an endpoint. And algorithmic discovery isn’t just a pipe dream. Movie recommendations on Netflix, news recommendations on Zite, and many other services with similar recommendation power are popping up. Personally, I look forward to more and more of these applications helping me with recommendations in all areas of my life, so that I can finally begin to discover what I’ve been missing.

Mark Johnson is CEO of Zite, a personalized magazine for the iPad and iPhone. You can follow him on Twitter at @philosophygeek.

[Top photo credit:  jokerpro/Shutterstock]