Archive For July, 2011

Can We Ever Digitally Organize Our Friends?

Can We Ever Digitally Organize Our Friends?

When Google+ launched last week, one of the most discussed features was Circles. In case you haven’t read a single blog, Tweet, or Google+ post in the last week (and yet, somehow stumbled into this dark corner of the internet), Circles is Google’s way of allowing you to group people. You can put anyone into a one or more Circles such as “Friends”, “Acquaintances”, “Co-workers”, “People I Eat Brunch With”, “Cyclists”, etc. And you have to put a person in at least one group.

Many discussions have ensued about how people are organizing their Circles. Many have also praised Google’s elegant and unique implementation as a clear answer to Paul Adam’s research entitled “The Real Life Social Network“.

I’ve been thinking about grouping and organization of friends for a long time. As an information architect, it’s in my nature to want to organize and tag everyone I know. I even wrote a post in 2004 about the organization of my instant messaging list. In thinking about this for a little while, I’ve decided to try to document my thinking so far on Google Circles, as well as the larger context of digitally grouping people.

Why We Need Groups

I’ve noticed from reading a number of articles about Groups, Circles, Lists, etc. is the variety of use cases and needs. We see so many implementations because there are many needs. There are different use cases for publishers and consumers, public and private, symmetrical and asymmetrical. Before I talk about the challenges I think Google Circles, and any similar feature faces, let’s look at some problems grouping attempts to solve.


One reason for needing groups is because you’re only comfortable with certain people seeing what you’re talking about. For example, you may only want to share baby photos with family. Flickr’s Friends and Family settings are very much targeted for this use case.

Interest Context

Subtly different from privacy, you feel only these people need to know about what you’re talking about. An example might be asking for advice on what tires to buy from your car enthusiast friends. You don’t necessarily care that it’s public, you just want to make sure the right people read it. Currently, communities around interests is largely solved by services such as Yahoo!, Facebook, and Google Groups.

Local Context

Posts that are specific to your location such as “what’s going on tonight?” or “Anyone have tickets to the concert tonight?” should be targeted to friends within that geographic area.

Event Context

Similar to local context, when you’re at an event (e.g., Coachella music festival), you may want to communicate and publish only to your friends who are also at the event. Group messaging tools such as GroupMe and Beluga—and now Google’s Huddle—attempt to solve for this use case. Not surprisingly, all of these solutions are mobile-centric.

Organizational Context

Your college friends don’t know your ex-coworkers don’t know your softball team. Maybe you want to share a link about your alma mater or one on how your former company just changed CEOs.

Not Spamming Everyone

A corollary to the contexts is that one may not only want to publish a post to a specific audience, but may also feel self-conscious about spamming others with what’s deemed irrelevant material.

Targeted Consumption: Reading From a Specific Group

For any of the contexts above, you may also want to consume based on a grouping. Perhaps you wish to view only messages from friends also at Coachella. Or you want to read the latest tech news from the technology journalists and authors you follow. Twitter Lists is an example of a way to address such a use case.

What Google Circles Does Right

Anything that’s created this much discussion and buzz is clearly doing something right. While I think there are some challenges for Circles, they’ve done many things that I think are positive, innovative, or at least interesting.


Unlike Twitter Lists or Facebook’s Friend Lists (did you know you could organize Facebook friends into lists?), everyone in Google+ has to be in a Circle. This is a bold move and puts organization as the focus of the entire product. While forcing grouping may seem like a higher barrier to entry, their interface makes it as easy as adding a friend on other sites.

The Circle Interface

Circles is a delightful experience. It makes you want to add people just for the fun of it. I do feel the drag n’ drop interface is one that’s more suited for touch interfaces and may be too much effort for large collections of people. However, adding people to circles through their Suggestions is a cinch. If you haven’t tried deleting a Circle yet, you should. Delightful.

The Sharing Interface

Although not unique, the sharing interface is simple. I can easily type in the Circles I want to share a post with and it’s clear from looking at any post which Circle I shared it with. One caveat is that the Extended Circles option is a pretty confusing one, even for those of us immersed in this world.

Auto-Suggest Circle Members

In one of the first Circles I created, after adding only three people, Google started suggesting others to add and every single suggestion was correct. This made creating that Circle much easier. Unfortunately, none of the other Circles I created had suggestions. I suspect they are generated from the Google Group the members are a part of.

One Way Circles

I don’t know about you, but I’m pretty sure some people I think are my friends don’t think the same of me. Likewise, that guy who calls me his “buddy” doesn’t need to know that I put him in the “Acquaintance” Circle.

Why Grouping Sucks

When I first started using Google+, I had a sense of déjà vu as I categorized my friends. I’d done this before… on Flickr, on Facebook, on Twitter, on my instant messenger contact list, and in my address book. Shortly thereafter, I came to the conclusion that it wasn’t worth the effort to rigorously group everyone. Then I started thinking about whether it was ever worth the effort to do so. Because as much as one tries to emulate the real life social network and address Paul Adam’s research, there are some human subtleties we’re missing in the digital world.

The Soft Line

Have you ever had a Facebook or LinkedIn friend request where you weren’t sure whether to accept or not? There’s a soft line that separates a friend and an acquaintance.

It’s true that you can probably label everyone you met through an organization (school, work, etc.) but the boundaries quickly become blurred. Let’s say you met someone through your classmate but she’s not in your school. Does she belong in the school group? What about the person who sometimes hangs out with your group of friends? Or the guy you met dozens of times but only at parties?

We’re incredibly adept at knowing the right situations to include the right people. They’re not black or white rules and depend heavily on context: is it a party, who else is there, do they know any of the other people, have you talked recently, etc. Unfortunately, this skill and these implicit social rules we know are not easily translated.


I have a very close friend, Mike. We used to share an office together back in 2000. We talked about everything, went on trips, and hung out nearly daily. Today, I see him on average once every two months. We still share our thoughts and are there for each other for support, but life got in the way and our relationship is different now.

Sociologist Gerald Molenhorst has shown that we change half of our social network every seven years but there isn’t a Changing of the Guard ceremony here. It’s not entirely clear at what point Mike moved from one group to another.

Thus, maintaining digital groups has two problems. First, you don’t know when to move someone from one group to another because transitions happen gradually. Second, it’s simply a lot of effort to maintain. How often would you update the entire list? And if it’s not updated, how useful are the groupings, really?


I think I could run an open card sort for myself and probably come up with some good categorizations for my friends. However, once I’ve created these fancy Circles, will I actually remember who will see a given post? From my experience organizing my Facebook and address book, I’ve found that I don’t remember the complex taxonomies I dream up. In fact, I don’t know that I can list every person that’s in my “Family” group in Flickr even though it’s less than twenty.

When compounded with the high overhead of maintenance and likely outdated groups over time, it’s even less likely that I’ll know who I’m actually sharing a post with.

One use case where recall isn’t a concern is in consumption. If you’ve created a “Celebrities” group to read their content, it doesn’t matter if you don’t remember every individual in the group.

Can It Be Done?

The maintenance required for grouping our friends is too high and too vague. We simply don’t have the rules as clearly defined as programs require and even if we did, the parameters change. Your personal tastes change. The influential people change. Even your friends change. Keeping the groups accurate and remembering its members is a challenge.

The obvious question to ask is: what about automation? Google Buzz attempted to automatically determine social ties based on who you frequently emailed. That solution lead to disastrous results, linking a woman closely to her ex-husband.

However, Buzz’s nascent attempts and failures do not necessarily mean automation is untenable. If you’ve seen LinkedIn Labs’ InMaps, you’ll know that your network is simultaneously clustered and complex. You can almost make out the groupings but there are many nodes that overlap multiple categories or aren’t easily categorized at all.

A new app that launched last week, Katango, seems to be a technology demo of how we can use these clusters to auto-group your Facebook friends. I was impressed with how accurately they created logical groups that hadn’t occurred to me. Perhaps the smartest part of the app is that it then allows you to edit the group by removing people. This approach is smart because it’s easier to say, “this looks like my college drinking buddies… except that guy; he was only sort of with us” than it is to say, “this looks like my college drinking buddies… who’s missing?” Recognition over recall wins again.

Perhaps as we refine these patterns and technologies, we can start to not only recommend the grouping, but also recommend changes to the groups over time, thus lowering the maintenance cost. However, what I really wonder is whether we should be trying to mirror real life interactions at all. Instead of mapping, wouldn’t it be more interesting to change or create new behaviour?

Thanks to Dave Gray for his input and feedback on my draft (see his Google+ post on sharing) and to Coley Cheng for some masterful editing.

1 of 2