New Frontiers in Political Polling: Social Media and "Sentiment Analysis"
Whether you like it or not, your online status updates are now fueling America's political machine. Campaigns are mining social networks for data on potential supporters. Facebook is teaming up with Politico, granting the news outlet exclusive access to user data. The goal: using sophisticated computer programs to interpret our language, emotions and political dispositions. Tech Tuesday explores the technical challenges and ethical grey areas of "sentiment analysis" in politics.
The Young Turks' explanatory piece on the Facebook/Politico partnership:
"Sentiment Analysis," Real-Time Reactions, And Social Networks
Earlier this month, Facebook struck a deal with Politico, granting the news outlet exclusive access to data about its users' political preferences. Under the deal, Facebook used a sophisticated algorithm to analyze the language of its users public and private postings, flagging all mentions of Republican Presidential candidates by name, and analyzing whether those comments were positive or negative. This "sentiment analysis" data was then shared with Politico to produce reports and commentary. As Mashable's Alex Fitzpatrick reported, Politico can also use the Facebook information to provide more personal data about users, like age and location.
The Facebook-Politico deal prompted mixed reactions. Some observers said the arrangement raised strong privacy concerns. Others expressed hope that this kind of data-analysis could end up deepening our understanding of American political attitudes, even as they raised concerns about the arrangement itself.
The Future of Political Polling?
Traditionally, public opinion polls and focus groups were the only reliable source of data about politics and consumer goods. But many campaigns and corporations believe that "sentiment analysis" could someday replace traditional polling. Last year, researchers at Carnegie Mellon University found that data analysis from Twitter yielded the same results as traditional polling. In the lead-up to the Iowa Caucuses, an analysis of Twitter data by Mashable / Global Point Research correctly predicted a surge by Republican Rick Santorum before traditional pollsters.
"Sentiment analysis" exists at the intersection of computer science and linguistics. Writing on Language Log, Guest Phil Resnick recently explored the technical challenges and ethical implications of building algorithms to interpret and quantify natural human language on the web.
Resnik is currently working on an app with colleagues that would let people react to live events (like political debates) through their smartphones, and their comments would be displayed in real time on an associated web page. Resnik thinks this kind of tool will let people engage deeply with the events and participants as they watch, rather than just yelling at the T.V. or venting on Twitter. Scientists would also be able to study exactly which candidates' statements people reacted to down to milliseconds.