On August 25th, 2014, a 9 year old girl accidentally shot and killed her gun instructor while learning how to fire an Uzi.  It sounds outrageous enough to be a hoax–an article from the Onion–but various high profile news media outlets confirm the tragic shooting, which captures international attention for days following the incident.

Investigating this story using the Trails system reveals what we would expect to see from a true story: dozens of verified accounts tweeting about it, receiving hundreds of retweets.  However, we also find a persistent though low profile attempt on Twitter to disprove this story.  In this post, we study the emergence of a spam-like conspiracy theory surrounding the accidental shooting.

Explore this story on TwitterTrails, or read on for our observations.

We investigated this story on September 5th, over a week after the shooting occurred, with the search terms, which become our keywords: (optional) girl kills instructor, girl uzi kill, shooting instructor, girl nine kill, and (required) girl kill, with at least 1 optional keyword.  From these we collected over 12,000 tweets, 3,500 of which form the relevant tweet dataset.  Although we collected the data well after the incident occurred (and the Twitter Search API gives you data up to a week old), we ended up collecting earlier tweets by people retweeting them in the week prior to our collection.  In this way, we can recreate a meaningful picture of the activity when the story first broke, since the many influential tweets/users will continue to be retweeted.

Propagation Visualization

We can use the propagation visualization to learn how the story first broke and rose to prominence on Twitter.  This propagation graph is an excellent example of the development of a true story.  Because we collected the data far after the story broke we see only tweets which were still being retweeted a week later. Likely because of this, the timespan of this graph is very long (about a day and half); there were probably many more tweets coming in faster when the story first broke which are no longer being retweeted ans we are unable.  However, the tweets we do see are a sample of the most prominent tweets, since they were still being retweeted (which also means people still have interest in and trust the information enough to continue to retweet it).

Labelled on the graph are the verified accounts who posted some of the top tweets.

Many of the accounts are verified (indicated by the blue borders), and have many followers (indicated by the large points).  From the y-position we can see that a few tweets received around 200 to 400 retweets, one over 800, and the rest varying from 0 to 200.  The variety of colors show that there are very few tweets with identical text.  However, they all report or comment on the same story:

9-year-old girl accidentally shoots, kills Arizona shooting instructor as he was showing her how to use an Uzi: abcn.ws/1p7qsRc – @ABC
Obama looked Sandy Hook parents in the eyes and promised action on guns. Today a 9-yr-old girl killed a shooting instructor with an UZI. – @piersmorgan
9-year-old girl accidently shoots, kills instructor at Arizona gun range ti.me/1tT54DL – @ TIME
Girl aged 9 accidently shoots and kills shooting instructor during lesson with Uzi gun bit.ly/VQq6qShttp://t.co/HcnTFuPmzh  – @DailyMirror

Looking at this graph, we can be very certain that this story is true.  Many trusted and verified accounts have posted the same information, all confirming that a 9 year old girl in Arizona killed her shooting instructor with an uzi.

Timeline Visualization

Because of the many recent shootings, gun control debates and the tragic nature of this incident, this story had a long life on Twitter.  Though we collected data over a week after the shooting, people were still discussing it.  Using the timeline visualization, we can track the lifespan of various themes to this story which emerged during it’s life on Twitter.  Four are highlighted in the graph below: 

A few days after the shooting, People Magazine (@peoplemag) posted that the gun range the girl was shooting at was “closed indefinitely,” shown in the black line above.  Shortly after that, BBC wrote an article which circulated on Twitter (shown in orange) that no charges would be brought up against the girl.  About a week after the shooting, a story written by the AP was picked up by various news accounts, about how the girl involved said the uzi was “too much” for her (purple).  @cnnbrk posted a story about the family speaking out later that day, shown with the green line.  People have gone on from observing that the story happened, and are discussing different stories related to the initial shooting, which is a good indicator that overall, Twitter accepts this story as fact.

To test this, we can look at the negation in the Timeline visualization.  Based on what we have observed so far, we would expect to see no tweets, or at least very few (perhaps sarcastic tweets, or people expressing disbelief).  Instead, we see something a bit strange: low but persistent negation of this story, going on for days.  See the pink line below:

The negation series, showing a persistent denial of this story.

It wouldn’t be too surprising to see a short spike of negation after the story is initially posted.  The “negation” line collects tweets which express doubt in the veracity of the story, containing words like hoax, false and fake.  After this story first broke, people might wonder whether this is a hoax, because it sounds almost like satire.

But this “doubt” in the story persists for days, so we will take a closer look.  Here is the text of the first tweet in the series:

“BS ALERT: “9 Year Old Girl Kills Shooting Instructor With Uzi””bit.ly/1pjxwzX #MainStreamMedia#FalseFlag #Hoax – @URSheepleWaKeUP

The rest of the tweets are almost identical: claims that the story was faked by the mainstream media in order to take away people’s 2nd Amendment rights.

Conspiracy Theory

In this story, the negation we pick up is different from what we saw in the Plane in the Sea Story, where there is a legitimate correction to false information which gains in popularity and results in the decreased spreading of the initial false information.  In that story the negation was a clue to us that the story was false.

Here, the claims that the shooting of a gun instructor by a 9 year old girl is completely false is very different.  In fact, the spread of this rumor resembles, much more closely, the spammers we saw in the Oregon High School Shooting story: there are two similar messages being propagated, shown using the Similar Tweets widget.

The Find Similar Tweets widget displays tweets that are textually highly similar to a search tweet (displayed on the left).  The Cosine Similarity measures how similar the two texts are from 0 to 1, with a measure of 1 being identical.

These tweets link mainly to three websites:

• A youtube video which has since been made private:  https://www.youtube.com/watch?v=Psmnpxvvhlk
• A reddit board (in the conspiracy forum): http://www.reddit.com/r/conspiracy/comments/2ezm70/boom_9_year_old_girl_kills_shooting_instructor/ which references the same video and claims that the gun that was shot was a non-lethal airsoft gun instead of an Uzi
• A blog, beforeitsnews.com, where anyone can post news stories.  It is known for being full of often conservative conspiracy theories: http://beforeitsnews.com/alternative/2014/08/9-year-old-girl-kills-instructor-with-uzi-complete-hoax-plot-to-further-gun-grab-video-3020636.html

The blog references various youtube videos by Red Pill Revolution, claiming that the whole story was faked, including the entire identity of the gun instructor who was killed.  The whole story is claimed to be a liberal media hoax to further gun control and the destruction of 2nd Amendment rights.

 

Ignoring the political agenda behind this message, it is fairly hard to believe due to the nature of how it spread.  For one, the story never picks up in popularity.  Although it persists for a few days, the activity is extremely low; at its highest point, 8 tweets in 10 minutes.  The text of the tweets are also identical, as though there were copied and pasted. Unlike the Oregon Shooting story, they do come from different sources (source meaning the client used to post the tweet; TweetDeck is a popular example, or people posting directly from the Twitter website), but sharing the exact same text is highly suspicious.  People were not independently coming to this conclusion and tweeting about it.  The beforeitsnews post shows up in the top 10 most tweeted links, but with a slightly suspicious ratio of 65 tweets posted by only 39 users.  Overall, these are indicators that give us reason to doubt a claim, especially compared to the claim that the story is true which has much more popular support on Twitter.

Conclusion

The story of a girl accidentally killing her gun instructor is easily confirmed as true.  Trusted (by most people) news media accounts are in consensus about what happened, and the story spread with highly retweeted posts and had a long life on Twitter.  However, there was still an effort to prove the story was faked, but a study of that effort reveals a conspiracy theory, highly open to doubt because of its lack of propagation and the fact that there are only a handful of original messages reposted nearly verbatim by dozens of users.