Twitter users do poorly at determining whether a particular tweet is true or false, and those who have more experience with the microblogging service tend to be more credible than novices, finds a new paper from Microsoft (NASDAQ: MSFT) Research and Carnegie Mellon University statisticians.

The paper (.pdf), published online Feb. 1, comes at a time when the federal government is increasingly open about its monitoring of social media, including Twitter. On Jan. 23, Customs and Border Protection officials denied two British citizens entry into the United States after one of them earlier tweeted that he planned to "destroy America," by which he meant getting heavily drunk during an upcoming trip to Los Angeles. ("They asked why we wanted to destroy America and we tried to explain it meant to get trashed and party," one of the two Brits told The Daily Mail.) In addition, search engines increasingly integrate social media into their search returns.

But while separate research shows that analysis of aggregated social media postings could reveal information about matters such as national mood, people are poor judges of the truthfulness contained in individual tweets, according to the Microsoft and Carnegie Mellon paper.

A study of 266 Twitter users conducted by report authors finds that the ability to spot truth from fiction in tweets from outside users' tweet stream has no correlation to how long users have used Twitter, how frequently they read or send tweets, or their number of followers or following. But, study participants with more Twitter experience tended to give higher credibility ratings to tweets and unknown tweet authors than those with less.

Twitter users spend about 3 seconds reading any given tweet, the paper says, leading users to make systematic errors of judgment when assessing the credibility of an outside-stream message--such as rating a tweet about a serious topic as more credible, presumably just due to the seriousness of the topic itself. Twitter users also resort to heuristics such as Twitter handle and image, giving more credibility to content and author when the handle is a topical one and the image isn't the default Twitter image, the paper says.

Information such as author bio, location, relative number of followers and following, number of tweets on a topic, number of mentions by other handles and retweets help users make more informed credibility decisions, but are not immediately available, report authors say. They recommend search engines make changes so that information is available at a glance, along with the content of the tweet.

For more:
- download the paper, "Tweeting is Believing? Understanding Microblog Credibility Perceptions" (.pdf)

Related Articles:
FEMA focuses on speed, not perfection in using social media
Social media a 'primary source' of Intel, says FBI
Swearing on Twitter could be sign of coming protests
Social media creates a democracy-challenging informational filter