Analyzing Facebook Updates

December 27, 2010

The Network World site has an interesting article about an analysis of Facebook status updates, performed by the Facebook data team.  (For those readers who are not familiar with Facebook, a status update is a short message that your Facebook “friends” see on their “News Feed” page.  Typical messages might be updates on a vacation, reports from a concert or sporting event, or just an interesting link to an article or YouTube video.)   The analysis first categorized word usage in a large sample of updates posted by US English speakers; the categorization attempts to identify characteristics of the message such as subject matter and emotional content (positive or negative).

Facebook analyzed the word usage for about one million status updates from its US English speakers. The social network said all identifiable information was stripped from the status updates before they were analyzed  …

Once the updates were anonymized, the words were organized into 68 different word categories based on the Linguistic Inquiry and Word Count (LIWC)–a text analysis software program created by James W. Pennebaker, Roger J. Booth, and Martha E. Francis.   Some examples of word categories used in the study include past tense verbs, prepositions, religion and positive feelings.

The analysis measured correlations between user characteristics, such as age, message length, time of day, and number of responses with the LWIC measures.  The  results were interesting, although not especially surprising in most cases.  The team found that:

  • Younger users tend to express more negative emotions, talk about themselves more, and cuss more than older users.
  • Older users post longer messages, on average, and talk more about other people.
  • People post more positive updates in the morning; the proportion of negative updates rises as the day progresses.
  • Longer updates attract more responses.
  • Negative updates generate more responses than positive ones.  The positive updates get more “likes”; the negative ones get more comments.   (Flame wars have been popular on the Internet for a long time.)
  • Groups of friends tend to post updates with similar characteristics.  (Now there’s a stunner!)

There is nothing here that is terribly surprising, but it is interesting to see that at least some of one’s intuitions seem to be confirmed.

The Facebook data team has posted a report on the Facebook blog; it has more details of the results.

%d bloggers like this: