Rethinking Lists, Groups and Circles

Editor’s note: Yoav Shoham is professor of computer science at Stanford University and co-founder of Katango, which organizes Facebook friends into groups

The recent introduction of Google+ has been fodder for much Google-versus-Facebook discussion. At the center of the discussion has been the Circles component of Google+, which allows users to arrange their contacts in meaningful clusters (for example, “family” and “work”) and share different content with different clusters. Circles play a role that’s almost entirely analogous to Facebook’s lists, which have been around (if somewhat buried in the Facebook UI) for a long time. Facebook of course also has the notion of groups, separate from (and more recent than) lists. Here are some basic observations on lists, groups and circles that seem to have been glossed over in the various recent articles.

  • This recent discussion has focused on the differences between the Facebook and Google offerings, but misses what I think is a more basic common – and striking – feature. They both ask the user to create groups/lists/circles manually. This works fine for groups with a small and stable membership; family, for example. But it’s a non-solution for large and/or fluid groups.
  • About six months ago I went through the exercise of sorting my then-321 Facebook friends into lists. It was excruciating. It took me over an hour to do a halfway decent job, and I wasn’t fun to be around when I was done. I’m now up to 388 friends; you couldn’t pay me to go through that exercise again. Facebook statistics confirm that I’m not alone (only 5% of the users have created lists, for example).
  • Facebook has had a creative solution—switch from lists to groups. The idea was that whereas only you can create and maintain your lists, a group is maintained collaboratively by all its members. In a twist on the familiar newsgroup self-subscription, in Facebook groups one needs an invitation; only existing members can add new members. (Both lists and groups are still supported on Facebook, though most users are not aware of these subtleties.)
  • But this does not solve the problem. Groups are not lists. My lists define me socially; they are my social mirror. They are mine alone and I’ll be damned if I let you touch them. I’ll decide who’s in my family circle, my AI cohorts, my tech-guru list, my college-friends cluster.
  • Of course, my lists overlap with yours. Indeed some are pretty darn similar. My wife’s family list has an 80% overlap with mine. This can be confusing, especially when we start using these lists to communicate. When my wife shares a photo with her family list and I with mine it can be hard to tell the two lists apart; both because the names of the lists may be identical (“Shoham family”, say, with apologies to the Eliasaf brand), and because the two sets of people are almost identical. For this reason we see a natural dynamic in which people with fairly similar lists tend over time to “standardize” on the same list. An informal rule of thumb I use – not verified in any way – is that lists will merge if their overlap is 75% or greater.
  • This doesn’t mean that groups aren’t important. This is especially true when there is an objectively-defined membership criterion. Membership in “Stanford class of ‘02” is not a matter of taste or opinion; you either were there or not (I’m sidestepping subtle issues, such as do we mean you graduated in ’02 or started in ’99). This group may (or may not) be part of my social mirror, but it can safely be constructed by others. Even when there’s not a pre-defined membership set, a group makes sense when there is an objective, impersonal concept defining it. Anyone can add themselves (as in newsgroups) or their friends (as in Facebook groups) to the “cat lovers” group, if it’s not meant to include only my cat-loving friends; it’s fine for it to grow organically. Finally, as multiple similar lists coalesce over time around one list (perhaps following the 75% rule), at that point the list in effect has also been “untethered” and become a member-maintained group.
  • (As an aside, philosophers and logicians have a lot to say about these issues, involving concepts such as “sense” and “reference”, “intension” and “extension”, “notation” and “denotation”, but you need to like that sort of thing to spend more time on it. I do, but it’s a lonely hobby.)
  • Ok, so if groups are not the solution, how do you avoid the pain of list creation and maintenance? I believe the answer is algorithms; they do 95% of the work, and the remaining 5% is manageable and even fun. There are many algorithms that produce sets of people; clustering algorithms, of the kind powering Katango’s first product release, are an example (yes, I am completely biased here). The output of the algorithm is “almost right”. Almost always, each emergent cluster makes sense, but you need to hone it: name the cluster (reliable auto-naming still eludes even the best algorithms), add and remove a few names (usually more removing than adding; by design, the system usually over-includes friends since removing is a simple mouse-click away), create a sub-cluster (for example, create an “immediate family cluster” from an “extended family” one), merge clusters, and so on.  This final honing is fun rather than a chore, for several reasons. First, it’s quick; a matter of minutes. Second, it’s rewarding to see your social mirror emerge; one user described examining a new emergent cluster as “unwrapping a present”. And third, it’s your social universe, and at the end of the day you know it best. The algorithm did the heavy lifting, but like an expert surgeon, you stepped in and made it perfect; you feel in control.
  • In an interesting recent TechCrunch piece, Tom Anderson lauded the fact that Google+ hands the user control over his/her social experience, and lamented Facebook’s over-reliance on EdgeRank-style algorithms to decide what information the user ought to see. I think that’s only half justified. One cannot master the social torrent without some algorithmic assistance, any more than one can navigate the web without algorithmic assistance. There’s a reason we use Google more often than Yahoo to find relevant web pages. But as I say above, I agree with Tom that at the end of the day the user must be given final control of his/her social interaction.

So, the main takeaways:  Don’t confuse lists or circles with groups; and let algorithms do the heavy lifting.