Researchers Develop an Internet Truth Machine

Hugh Pickens writes "Will Oremus writes that when something momentous is unfolding—the Arab Spring, Hurricane Sandy, Friday's horrific elementary school shooting in Connecticut—Twitter is the world's fastest, most comprehensive, and least reliable source of breaking news and in ongoing events like natural disasters, the results of Twitter misinformation can be potentially deadly. During Sandy, for instance, some tweets helped emergency responders figure out where to direct resources. Others provoked needless panic, such as one claiming that the Coney Island hospital was on fire, and a few were downright dangerous, such as the one claiming that people should stop using 911 because the lines were jammed. Now a research team at Yahoo has analyzed tweets from Chile's 2010 earthquake and looked at the potential of machine-learning algorithms to automatically assess the credibility of information tweeted during a disaster. A machine-learning classifier developed by the researchers uses 16 features to assess the credibility of newsworthy tweets and identified the features that make information more credible: credible tweets tend to be longer and include URLs; credible tweeters have higher follower counts; credible tweets are negative rather than positive in tone; and credible tweets do not include question marks, exclamation marks, or first- or third-person pronouns. Researchers at India's Institute of Information Technology also found that credible tweets are less likely to contain swear words (PDF) and significantly more likely to contain frowny emoticons than smiley faces. The bottom line is that an algorithm has the potential to work much faster than a human, and as it improves, it could evolve into an invaluable 'first opinion' for flagging news items on Twitter that might not be true writes Oremus. 'Even that wouldn't fully prevent Twitter lies from spreading or misleading people. But it might at least make their purveyors a little less comfortable and a little less smug.'"

I wonder...

[Anonymous Coward]

How effective would this be on real media? I bet it'd put those bastards in their place! :)

Chile's Earthquake

[thejynxed]

It's interesting to note, that a seismology student at a university in Chile finally had enough nonsense from false information over Twitter, etc about earthquakes, that he directly wired a big batch of seismographs to directly post their results via Twitter. The last I knew, they had over 1 million followers, and this particular student has been getting big thank yous from residents of the country.

This is crap

[Anonymous Coward]

One of the criteria in their algorithm seems to be that credible tweets were

... significantly more likely to contain frowny emoticons than smiley faces.

They were evaluating tweets about a disaster; not a lot of smiley faces there.

The algorithm seems to have a bias toward bad news. So, if my buddy tweets that a rare Belgian beer will be available at the local liquor store, the algorithm will decide that it isn't credible because of the smiley face.

We just had the above case. Beer that you usually have to cross the Atlantic to get became available for about 30 minutes locally. Some of us lined up starting at 3:00 AM. I would have been really ticked off if some algorithm had made me miss the news.

Gaming Reliability/Credibility Assessment

[girlinatrainingbra]

Of course, in just the same way that spammers can game Bayesian spam filters or rule-matching pattern filters by knowing what the rules are, given a known set of rules that attempt to assess credibility of tweet allows someone to tweak their tweets in order to be assessed as having high credibility:
1 -- max out your tweet length
2 -- include an URL [doesn't say whether to use a link shrtnr ;>(]
3 -- use a Twitter account with a high number of followers
4 -- use a negative tone
5 -- no question marks or exclamation points
6 -- use 2nd person (same as don't use 1st or 3rd person)
7 -- don't use swear words
8 -- use a sad emoticon
.
Example to maximize this:
a - break into / hack a high follower account (e.g. justinbieber) and tweet: cat > finaltweet
You should know Mayan Calendar sez: world ending this week. Confirmed@ http://netcraft.calendar.mayan/ [netcraft.calendar.mayan] you go hug loved 1s now. :>( beebs
wc finaltweet
1 20 139 finaltweet

First iteration was:
gia@sodium$ cat > count2
You should know that Mayan Calendar says : world ending within week. Confirmed by http://netcraft.calendar.mayan/ [netcraft.calendar.mayan] , you should hug loved ones now. :>( -- beebs
gia@sodium$ wc count2
1 25 159 count2

Please note that the "[netcraft.calendar.mayan]" was inserted by /.'s /-code and is not part of the wc wordcount :>(