Editor’s note: This is one of the occasional posts we do speaking to someone with something to say on topics that we find interesting. If you are seeing this and are not yet signed up to Deepnews, click the button bellow to start receiving our blog posts every week and a Digest of quality news on an important subject every Friday.
By Christopher Brennan
Deepnews wants to get people good information, whether it be publishers using our API to learn more about their own articles, news aggregators incorporating our quality score into their own recommendation algorithms, or individuals who read the Digest.
But beyond the value of good information in and of itself, quality information can also help prevent the worst effects of misinformation. While these effects have been thoroughly studied, there is also the question of whether “inoculation” against bad information is enough to combat it.
With the world’s attention turning to the rollout of COVID vaccines, I recently spoke with Dr. Jon Roozenbeek, a researcher at Cambridge University’s Department of Psychology, who has looked into a “psychological vaccine” against fake news.
He and his colleagues have looked into the effectiveness of interactive games that teach players the strategies that misinformation uses to spread conspiracies. Previous efforts have focused on general misinformation and misinformation in politics, though recently the team has focused their inoculation efforts on misinformation about COVID-19 in a game called Go Viral!
“It kind of works differently in mRNA vaccines, but it used to be the case that the vaccine itself was a weakened version of the virus. That would trigger the body to create antibodies. Here with psychological vaccines it is more or less the same thing,” Roozenbeek said.
“It isn’t it isn’t quite the same as what you might encounter in real life, but it does work the same way, does have the same features. So, in that sense your brain will be able to pick up on the strategies that you learn about in the game when you encounter them in real life.”
His results have shown that making people more aware of how they are vulnerable to misinformation reduces the perceived reliability of misinformation in the future with “immunity” that lasts for months. Roozenbeek is now working with models to try to figure out exactly how much inoculation is necessary for a sort of “herd immunity” where misinformation, such as conspiracies about COVID or COVID vaccines, no longer has the same pernicious effect.
However, the way that misinformation spreads online is of course not the same as the way that viruses spread in the physical world. One of the differences is that fake news and conspiracy theories are often spread on social media and platforms that target bits of information based on how likely they are to make someone click.
That raises the question of whether “vaccination” against misinformation can be entirely effective when, even if there is a small group of people who refuse to be inoculated against information, their bad information may become super viral because it is outrageous and likely to squeeze engagement out of people. Misinformation has been shown to spread six times faster than true stories.
“There could be a whole lot of reasons [to spread misinformation], not only at the individual level but also at the societal level, that happened when you let these algorithms run free and optimize the way that they display information to people based on how engaged they are,” Roozenbeek said.
“My impression is that inoculations in this particular case are quite useful. Definitely not useless. But are they the only solution? I don’t know … Are solutions that target the psychological level, any solution, really is that ever going to be enough? As a solution on its own my impression is probably not.”
While the answer is yet to be determined, there are other ways to sort content online other than pure clicks, such as Deepnews’s quality score. The usual response is that social networks have business models that depend on engagement, but there are also different types of engagement than baiting someone to click.
For example, last week our blog post touched on data about Deepnews’s quality score that shows higher quality articles on newspaper sites are tied to more readers converting into paid subscribers. While it is impossible to know exactly why people subscribed, my hunch is that people prefer to spend time on a site that has good quality content. Perhaps platforms then have something to learn from newspapers, and optimize for a positive user experience (including perhaps some inoculations against misinformation) rather than pure clicks.