Jul 20, 2011

Social Networks Done Right

I haven't spoken publicly about social networks because Google was still working on their product and I'm an ex-Google engineer. However, my prior work related to social networks is now covered by at least one patent and Google's social network is released, so I feel ok discussing it.

This post is my opinion on what a social network should be. Many things that I'll post as my idea are actually concepts that come from many conversations with various colleagues involved in social networks spanning 6 years.

What do I mean by "done right"?

IMO, a correctly designed social network would remove the need to have multiple accounts on difference services. Even with the most advanced social network, Google Circles, there still is a need for LinkedIn and Facebook because it's still the only way to isolate work colleagues from college buddies. See my other post about Google Circles.

What follows is the short version of a social network model that is roughly what I and a few others were pitching at Google back in 2005. But, it's probably more accurate to say that it is what I recall and like all memories, I've probably rewritten it in my own head.

Common over-simplification when thinking about social networks.

Part of my pitch back in the day, was to first bring everyone up to speed on social networks. One key point was that for all but the most trivial social networks, it's a good practice to consider the 3 person use-case, not just two.

My examples went something like: Person A and B are connected by some relationship rule and User O is the observer. Now, this User O may still be connected to Person A and/or B. Iin fact, that's where things get interesting and the complexity happens, but it's import to keep it all straight.

Rule 1 - Permission flow out from a user

This seems like a "well duh", but it's still important to point out. A user can lock down their profile to an extreme point where just the username is the only thing publicly visible. But when they "invite" someone to be their friend, they place them in a cluster (or circle) of friends and that invited person now sees whatever non-public information that circle is allowed to see... even before the friendship link is reciprocated.

This is important because invited users must decide if they know the inviter. If the default, locked down profile is all that's displayed, there will not be enough information for the "are we friends" test.

Rule 2 - Only symmetric relationships show up as friends.

The asymmetric relationship is also known as "fanning" (one person being a fan of another). To keep the value of a social network high, fan relationships should not be displayed like true symmetric relationships. What do I mean by "high value network"? Well, from orkut.com we learned that people decide if they know someone by looking at who they are friends with. If fan relationships seems like a mutual friendship, then how can anyone quickly derive who's really in their network of friends?

A social exploit I recently witness on Google Circles used a profile named "Marissa Mayer" and a picture she commonly uses on social networks. Once the fake account's profile looked legit, the only way an average person acquainted with Marissa Mayer to determine it's validity is to examine her friends. However, because of the way "fan" relationships to Larry Page, Surgey Brin, etc. sort of appear like friends, a couple ex-Googlers were fooled into accepting friend requests. And with each new Google friendship, the fake account seems more legitimate. Keep in mind, this approached worked to fool a few ex-Googlers into friending this fake account until it was deleted.

A more theoretical example showing why it's important to keep your social currency exchange rate high uses famous figures. Anyone can say they are friends with President Obama; however, it's something completely different if the President agrees that you are friends. Both parties verifying they know each other is significantly more valuable than one person claiming to know the other.

While it's not true for me, how much more cool would it be to answer the question, "You know the president?" with "Yeah, but more importantly, the president knows me."

Rule 3 - Friendship information needs protecting

Currently, this is one area where Circles doesn't go far nearly enough. Considering that the point of a social network is the network data itself. Google already had profiles... Circles is about social connections. So, it really should focus on privacy of that information as much as sharing posts.

In real life, your various circles don't know each other. My family doesn't know or interact with my colleagues. Likewise, my co-workers can't browse my rock climbing friends. There might be some individuals who overlap, but overall, my mom (or boss) shouldn't be able to peruse my friends and follow links to their profile and see pictures of me doing stuff.

A phrase to sum up this behavior is insulating circles from each other. Currently, it is done by using multiple accounts or completely different services (LinkedIn vs Facebook); however, there is an attention cost to maintaining multiple accounts and as API and services integrate with a social network, there's a user benefit to having it all in one place.

Granted, things get tricky if you really want to insulate circles of friends from each other, but the concept can be summed up with the following rule:

Rule 4 - Both friends must agree that the observer can see their relationship.

More on this rule below...

Warning: social consequences ahead

Say President Obama and I are friends. I'm proud of this fact, so I make our friendship public by putting him in a circle of friends that has no privacy limitations. But the President is a little concerned that reporters are going to harass me, so he puts me in a circle of friends logically called "Geeks" and sets the permissions to insulate this collection of relationships.

This means anyone else in the same Geeks circle would see the President and I are friends, but nobody else could. If I was alone in Obama's Geeks group and the Insulate This Circle was enabled, then nobody but the two of us would see our connection.

You might be thinking, "Whoa there! I'm proud of my associations and I want the world to know!!!" This is called "social currency" and it's greatly devalued when relationships in a network aren't required to be mutual and increased when both parties agree the relationship view-able. You may want it to be public knowledge, but there are two of you involved and if one person is more private than you, then that relationship should not show up to observers... in either of your friend list.

Why excluded from both friend lists? Well, a search engine could derive a list of the President's friends by crawl the site looking for "back links." (Now, the President could show up in my "fan" or "following" list until he adds me to a Circle).

There are numerous consequences from this design... limited network growth effect, friends seemly missing from lists when others are looking at you, and maybe some way to infer what others think your peer groups is.

Some of the concerns are mitigated by the fact that people can be placed in multiple circles and Circle's feature where you can view your own profile from another user's POV.

Other rules that don't seem as important now

My original model I pitched included Communities and non-person entities (like an Amazon). The idea was that non-Google companies can access the relationship data using the same permission mechanism as other users. If I want an external service to see a subset of my friends, I'd add that service to the appropriate circle.

For example, if you wanted a group of friends to see your Picasa album, then you'd drop the Picasa Album Agent into that circle of friends.

My point at the time was that Social Networks aren't a product in-and-of-themselves. They just made every other product much more powerful. And to prevent market fragmentation, an open standard should be established so that other big companies don't have to create mirrors of the Google network, or worse, a competing network. (Remember, this was 2004.)

We saw fragmentation of the browser toolbar market. Every company wanted one like Google's Toolbar and Google delayed opening up the toolbar API so smaller companies could leverage it. The result was a zillion toolbars all competing for space.

No comments:

Post a Comment