Twitter Shows Ideas Are Contagious, but Only the Bad Kindby
Bad news for public health officials: Negative messages about vaccines spread through social networks, but pro-vaccine messages don’t. That’s the conclusion of a new study by Penn State researchers who analyzed Twitter messages from the H1N1 swine flu pandemic three years ago.
It gets worse. People who were exposed to a lot pro-vaccine messages—those whose social networks were basically telling them to get their flu shots—were more likely to come out with the opposite view. “If you’re intensely exposed to positive messages, we would essentially be able to predict that you would go negative, express negative sentiment,” says Marcel Salathé, a professor of biology and computer science at Penn State’s Center for Infectious Disease Dynamics.
Salathé’s team analyzed more than 300,000 tweets that mentioned the H1N1 flu vaccine during a 45-day period in 2009. The researchers, with the help of 100 students, identified about 10 percent of the messages as negative, positive, or neutral. Then machine-learning algorithms used that information to sort the rest.
The challenge of Salathé’s study was teasing out which Twitter users shared similar views and happened to be connected on the network and which ones were actually spreading their views, a phenomenon known as social contagion. “It could be that these people are actually already friends because they share the same views,” he says. “If you ignore that, then you’re always overestimating contagion.”
Maybe it’s not surprising that negative messages spread and positive ones don’t. (Recall the aphorism attributed to Mark Twain: “A lie can travel halfway around the world while the truth is putting on its shoes.”) But why would people on the receiving end of pro-vaccine messages eventually express the opposite?
The study doesn’t address the reason, but Salathé hypothesizes that “there are actually a lot of people who have latently negative views about flu vaccinations, but it’s not necessarily the politically correct thing to say.” Bombarding them with pro-vaccine messages may trigger a backlash.
Some important caveats: It’s impossible to know whether people actually got vaccinations (vs. what they posted on Twitter). Other messages and experiences besides Twitter influence how people feel about vaccines. And Twitter’s 18 million users in 2009 don’t represent the whole world.
If the findings are right, what’s a public health official to do? First, the study suggests a need to focus on “prevention and control of negative sentiments (particularly if based on rumors, misinformation, misunderstandings, etc.).” Salathé adds that it may make more sense to focus messages on the negative effects of not getting a flu vaccine, rather than promoting the benefits of getting one.
The cost of getting a disease, after all, is much more salient than the benefit of staying healthy. “Vaccination is one of those things, if it works really well, you don’t see the reason why you would need it,” Salathé says. Prior generations “were constantly surrounded by all the misery that these diseases cause, and we’re not.”