Brynn Marie Evans

I’ve been working on a social sensemaking project with Peter Pirolli at PARC. In the process of designing it, we realized that there were smaller pieces of the puzzle, so to speak, that could be investigated one at a time. The results I’m reporting here are from a study looking at just how virtual today’s online social networks are. (Done with Peter Pirolli and Sanjay Kairam.)

We were primarily interested in whether there was a relationship between online social network size and real life social connections. In other words, if you have lots of friends on Facebook, are you more likely to have lots of friends in real life? At least more so than someone with fewer Facebook friends? What about with Twitter, a “networking” sites of sorts where “friendships” don’t have to be reciprocal. We wanted to look at online and real life network sizes to see if we could develop a quick metric for representing one’s social dynamics for future studies we are planning.

Methods:
To do so, we created a short web-based survey (hosted at wufoo.com) asking users to report:

1. Their primary (or preferred) social network: Facebook or Twitter.
2. The number of friends they have on that network.
3. The number of people in that network who they had a social interaction with in the last month.
4. An estimate of the number of friends they had in real life.
5. The number of people in this network who they had a social interaction with in the last month.

We defined “social interaction” as any sort of exchange between two people, occurring either online or offline. These could be email exchanges, chatting over IM, messaging within a networking site, talking face-to-face or on the phone. We also tried to provide some boundaries for estimating the total number of friends people have in real life: we said this could include friends, acquaintances, and other people that you would consider to be part of your extended social network.

We distributed the survey among friends, who shared it with their friends, etc. We took out a Facebook ad and posted under “Volunteers” on Craigslist in a number of cities around the country (Honolulu, Atlanta, Philadelphia, Minneapolis, etc.)

Data:
In total, we collected 123 replies from Facebook users (45% Male) and 78 from Twitter users (77% Male). Most were relatively young (85% of Facebook users were between 18-35 yrs; 77% of Twitter users were between 25-45 yrs). Respondents were also highly educated: 75-80% having college educations or higher. We actually didn’t find any age differences in the distributions of online or real life social network sizes.

Our data appeared to be highly non-normal. Users’ online social network sizes varied greatly from 5 to 1495 (Facebook average: 314; Twitter average: 126). Real life estimates also varied from no friends to 1000 friends (Facebook average: 168; Twitter average: 143). (See below.) (Side note: It’s interesting that real life network sizes averaged out to be very nearly Dunbar’s number of 150: or the limit of the number of active social connections that humans can maintain. There has been some discussion over how technologies can change Dunbar’s number, which I think ultimately has a complicated answer; but based on my data, we see somewhat reasonable reporting of network sizes based on distributions of known social networks.) The trend for a higher average real life network size for Facebook users may be due to their overall larger online networks, or may be an artifact of the order of survey questions. (Users estimated their real life sizes after they reported their online network size.)

Mean Network Size (Online & Real Life Network) for Facebook & Twitter

The number of people our respondents had an interaction with in the last month varied similarly. On average, their estimates were about 50 people (53 for Facebook; 47 for Twitter). The range was greater for Facebook users, extending up to 450 people, whereas the maximum number of interactions for Twitter users was 250. Again, these numbers could be an artifact of the survey ordering, although their reduced variability suggests that they may be slightly more accurate than estimates of total network size.

Analysis:
Since our data were non-normal, we used non-parametric statistical tests that did not make any assumptions about the underlying distribution of the data. The Spearman Rank Order Correlation is a good substitute for the Pearson’s Correlation, so we use this for the following comparisons:

1. Online network size vs. Real life network size
2. Online network size vs. Number of recent interactions with friends in the real life network

We found that all were significant for both Facebook and Twitter (although stronger correlations for Facebook), and that the number of social exchanges with people from one’s real life network seems to be the best indicator of online network size. (This could be because counting the number of people you’ve interacted with is easier than estimating the size of your whole network, and therefore may be more accurate.)

Positive Correlation for Facebook Network Size & Recent Social Interactions

The correlations were also positive: meaning that people with greater online network sizes have more friends in real life. In fact, this is what you would expect—you would expect there to be overlap between these social networks so that some proportion of real life friends are represented in the online network. (In retrospect, this would have been an interesting question to add to the survey; we did not collect any information about the percentage of overlap between people’s networks.)

Interpretation of Results:
After exploring the data from a number of angles and factoring in the positive correlations (above), we concluded that Facebook is most likely to have the best relationship to one’s real life social network. In other words, Facebook users’ online networks seem more representative of their real life friendships than Twitter users’. This could result from a number of different factors relating to the technologies themselves and their respective socio-cultural makeup. For example, a friend on Facebook must be bilateral—we are only friends with each other if we both confirm the friendship. Twitter “friendships” do not require confirmation (unless users are private), and as a result, a number of Twitter users add friends who 1) do not add them back and 2) who they may have never met before. Furthermore, Twitter is not as mainstream as Facebook (3 million vs. 100 million users), which means that a Twitter user’s friends may not even be on Twitter. Of course, this applies to some extent to Facebook, too; but it could help explain the differences we observed in our data.

Just last week, Facebook reportedly banned a user for adding too many friends. Why? Because “Facebook accounts are meant for authentic usage only. This means that we expect accounts to reflect mainly ‘real-world’ contacts…We encourage users to add people that reflect their real-world connections and create trusted networks.” As I suspected! …although it’s fun to see this reflected in our data.

Later that day, I searched for Twitter updates that made reference to this news (there weren’t many), when I stumbled upon these two posts which provide additional anecdotal support for this finding.

Facebook reflecting real life:

Twitter being detached from reality:

Future work:
As we mentioned previously, this work is part of a larger study of social sensemaking. The hypothesis is that users with large social networks, who also have diverse social networks, will have better access to (social) information resources and will, therefore, perform better on sensemaking tasks. Our next step is to assess diversity. Do large social networks suggest greater diversity? Not necessarily, if people surround themselves with people who are like them. However the “diversity” measure also depends somewhat on what aspects of diversity we care about. Stay tuned! ….