Ebola has been making headlines since March, but it didn’t really start trending until September 30. But what does “trending” actually mean? Basically, a man was diagnosed with Ebola in the United States, and the Internet went bananas.
How do we know this? Well, partly because we pay attention to the news. But also because we asked the social media analytics company Crimson Hexagon to look into it for us. Their software analyzed 4.2 million global tweets that used the word “Ebola” or the hashtag #ebola, from August 31 through September 30.
What they found is, at first, not surprising. Until September 30—the day Thomas Eric Duncan was diagnosed—not all that many people were tweeting about Ebola. Roughly 100,000 to 150,000 tweets per day were about Ebola, despite the fact that during the same month, the number of cases and deaths almost doubled (3,069 cases and 1,552 deaths climbed to 7470 cases and 3,431 deaths). Then, within one hour after the announcement that Duncan had Ebola, the volume of tweets shot up to 200,000. For the entire day, over 550,000 tweets spread across the Internet, eclipsing the month’s averages.
A fairly large segment of that activity was generated by RTing the breaking news. CNN’s tweet announcing the Duncan’s diagnosis was the top Ebola-related RT on Sept. 30, attracting more than five times the volume of RTs from other handles with similar headlines.
“People share news early on from news organizations’ handles, and, in my opinion, famous or well-known handles tend to be the most retweeted,” said Elizabeth Breese, a senior content and digital marketing strategist at Crimson Hexagon.
But that’s not all we can know about how and why Ebola started trending. Crimson Hexagon also performed a “sentiment analysis” on the data; that’s a process that categorizes tweets based on tone—depending on the language they use, they can be classified as positive, negative, or neutral.
Here’s a chart showing their results:
So, this is the possibly surprising thing: Throughout the month, tweets about Ebola were not, as you may have expected, overwhelmingly negative. In fact, they were overwhelmingly neutral about a virus that kills an estimated 70 percent of those it infects. Why so neutral, Twitter?
“The tweets before Sept 30 are neutral probably because they are largely sharing news reports or commenting without using too much ‘emotional’ language,” said Breese.
Some language is actually quite hard to classify, no matter how common it is. Take this measurement of tweets involving both “Ebola” and “OMG” or “oh my god”:
Thing is, OMG, as exclamatory as it is, gets classified as positive language. So that blip of positive you on Sept. 30? You might as well throw that in with everyone freaking out.
All right, so that was the beginning of Ebola’s Twitter-trending run. But what happened in the week following dwarfed what had come before. Take a look:
As you may notice, the volume of activity after Sept. 30 not only increased wildly, but the proportion of neutral sentiment shrank, while the corresponding amount of positive and negative sentiment grew.
So what happened on Oct. 2 that got Twitter so emotional?
Breese said the language that people used might have gotten more animated as the news started to talk more about Duncan’s life.
On Thursday, the Centers for Disease Control and Prevention and local Texas health officials began the process of contact tracing to prevent an outbreak, which pieces together who Duncan had interacted with while he was contagious.
More details came to life almost hourly about who Duncan was, why he had traveled to the United States, his flight path, and where he had been between Sept. 24 and Sept. 28. Health officials revealed that as many as 38 people may have been exposed to Duncan while he was contagious, and some of them were children who attended public schools in the area.
Twitter blew up because of a single Ebola diagnosis in Texas. But that same week, the World Health Organization released an updated number of total cases and deaths; 7,470 people had contracted Ebola in West Africa, and 3,431 of them had died.
None of those numbers made it to any top RT lists.