The great east japan earthquake was the largest in the country’s history. It was powerful enough to shift the Earth on its axis by several inches, with forty-metre-high tsunami waves following soon after. Then the rumours started. Three hours after the earthquake hit on 11 March 2011, a Twitter user claimed that poisonous rain might fall because a gas tank had exploded. The explosion had been real, but the dangerous rain wasn’t. Still, it didn’t stop the rumours. Within a day, thousands of people had seen and shared the false warning.[118]
In response to the rumour, the government in the nearby city of Urayasu tweeted a correction. Despite the false information having a head start, the correction soon caught up. By the following evening, more users had retweeted the correction than the original rumour. According to a group of Toyko-based researchers, a quicker response could have been even more successful. Using mathematical models, they estimated that if the correction had been issued just two hours earlier, the rumour outbreak would have been 25 per cent smaller.
Prompt corrections might not stop an outbreak, but they can slow it down. Researchers at Facebook have found that if users are quick to point out that their friend has shared a hoax – such as a get-rich-quick scheme – there’s an up to 20 per cent chance the friend will delete the post.[119] In some cases, companies have deliberately slowed down transmission by altering the structure of their app. After a series of attacks in India linked to false rumours, WhatsApp made it harder for users to forward content. Rather than being able to share messages with over a hundred people, users in India would be limited to just five.[120]
Notice how these counter-measures work by targeting different aspects of the reproduction number. WhatsApp reduced the opportunities for transmission. Facebook users persuaded their friends to remove a post, which reduced the duration of infectiousness. Urayasu City Hall reduced susceptibility, by exposing thousands of people to the correct information before they saw the rumour. As with diseases, some parts of the reproduction number may be easier to target than others. In 2019, Pinterest announced they’d blocked anti-vaccination content from appearing in searches (i.e. removing opportunities for transmission), having struggled to remove it completely, which would have curbed the duration of infectiousness. [121]
Then there’s the final aspect of the reproduction number: the inherent transmissibility of an idea. Recall how there are media guidelines for reporting events like suicides, to limit the potential for contagion effects. Researchers like Whitney Phillips have suggested we treat manipulative information in the same way, avoiding coverage that spreads the problem further. ‘As soon as you’re reporting on a particular hoax or some other media manipulation effort, you’re legitimising it,’ she said, ‘and you’re essentially providing a blueprint for what somebody down the road knows is going to work.’[122]
Recent events have shown that some media outlets still have a long way to go. In the aftermath of the 2019 mosque shootings in Christchurch, New Zealand, several outlets ignored well-established guidelines for reporting on terrorist attacks. Many published the shooter’s name, detailed his ideology, or even displayed his video and linked to his manifesto. Worryingly, this information caught on: the stories that were widely shared on Facebook were far more likely to have broken reporting guidelines.[123]
This shows we need to rethink about how we interact with malicious ideas, and who is really benefitting when we give them our attention. A common argument for featuring extreme views is that they would spread anyway, even without media amplification. But studies of online contagion have found the opposite: content rarely goes far without broadcast events to amplify it. If an idea becomes popular, it’s generally because well-known personalities and media outlets have helped it spread, whether deliberately or inadvertently.
Unfortunately, the changing nature of journalism has made it harder to resist media manipulators. An increasing desire for online shares and clicks has left many outlets open to exploitation by people who can deliver contagious ideas, and the attention that comes with them. That attracts trolls and manipulators, who have a much better understanding of online contagion than most. From a technological point of view, most manipulators aren’t abusing the system. They’re following its incentives. ‘What’s insidious about it is that they use social media in precisely the ways it was designed to be used,’ Phillips said. In her research, she has interviewed dozens of journalists, many of whom felt uneasy knowing they are profiting from stories about extreme views. ‘It’s really good for me, but really bad for the country,’ one reporter told her. To reduce the potential for contagion, Phillips argues that the manipulation process needs to be discussed alongside the story. ‘Making clear in the reporting that the story itself is part of an amplification chain, that the journalist is part of an amplification chain, that the reader is part of an amplification chain – these things need to be really foregrounded in coverage.’
Although journalists can play a large role in outbreaks of information, there are other links in the transmission chain too, most notably social media platforms. But studying contagion on these platforms is not as straightforward as reconstructing a sequence of disease cases or gun incidents. The online ecosystem has a massive number of dimensions, with trillions of social interactions and a huge array of potential transmission routes. Despite this complexity, though, proposed solutions to harmful information are often one-dimensional, with suggestions that we need to do more of something or less of something.
As with any complex social question, there’s unlikely to be a simple, definitive answer. ‘I think the shift we’re going through is akin to what happened in the United States on the war on drugs,’ said Brendan Nyhan.[124] ‘We’re moving from “this is a problem that we have to solve” to “this is a chronic condition we have to manage”.
