This seems like a bad-idea:
As a part of the partnership, POLITICO and Facebook will also survey voting-age users in South Carolina on a daily basis. POLITICO’s editorial team will write the questions and publish the survey results on POLITICO the following day. The results will also be available on POLITICO’s official Facebook page as well as on U.S. Politics on Facebook, Facebook’s hub for campaign 2012 information.
Using the word ‘survey’ leads one to believe that the process is opt-in, with respondents selected through some randomized process. That assumption is incorrect. The ‘survey’ is done by connecting directly to the Facebook firehose and using sentiment analysis on the results:
Most notably, the Facebook-Politico data set will include Facebook users’ private status messages and comments. While that may alarm some people, Facebook and Politico say the entire process is automated and no Facebook employees read the posts. Rather, every post and comment — both public and private — by a U.S. user that mentions a presidential candidate’s name will be fed through a sentiment analysis tool that spits out anonymized measures of the general U.S. Facebook population.
Some of the things people say publicly on Facebook are atrocious. I can’t imagine the private content is any better. That said, I’d like to get my hands on the data set. I would however be interested to see if there was a difference between public and private sentiment.