You can put corrections in brackets and red italics, but they still won't stick in people's minds if they don't want to believe the truth. That's one of the conclusions of a new study about the effectiveness of corrections to information online.
Not everything you read on the Internet, as the saying goes, is true. Yet what kinds of corrections will work best to convince people of the facts? A new study of 574 people examined two different ways of presenting corrections, to see which would work better and with whom.
Reading a correction on a separate website, after a delay, works moderately well with most people, the study found. But there's a split when corrections appear right inside a story, in brackets with red italics. Those whose political beliefs line up with the corrections are more convinced the corrections are true, while those with opposing political beliefs act just like a control group that saw no correction at all. [SEE ALSO: Wi-Fi Names Reveal Political Beliefs, Map Shows ]
"We see stronger bias in the response that people have to corrections when we do it in real time," said R. Kelly Garrett, a communications researcher at Ohio State University who led the study. That's too bad, he said, adding: "One might hope that approach would work well because it would allow you to correct information at the source, before it has the chance to spread."
There's one upshot to Garrett's findings. They show that separate fact-checking websites such as Snopes or Factcheck.org work pretty well for people with all kinds of political beliefs, he told TechNewsDaily. "It works for everyone. Everyone becomes a bit more accurate," he said.
Yet in the future, Garrett thinks people may see more real-time, embedded fact-checking. Companies are working now on browser add-ons that automatically find incorrect statements — things like "President Obama wasn't born in the United States" or "Vaccines cause autism" — and offer the expert consensus on the topic. "It's not a huge area, but there are certainly a number of projects focusing on this topic," Garrett said.
Many previous efforts to make such software, such as a 2009 project called Dispute Finder, failed because not enough people downloaded it, Garrett said. Nevertheless, he predicted the current generation of insta-fact-checking browsers will take on. "I still believe in the idea and I think in five year's time we will see something like this out there," he said.
He hopes his research will guide the design of insta-fact-checkers. Perhaps the browser add-ons could collect their findings in a report at the end of the day, instead of embedding comments. They might also tailor their tactics to users' personalities, Garrett said.
In an age of cable networks and endless blogs and social media posts, Garrett thought a browser that shows people what the expert consensus is could be a relief.
"People are having to rely on their own ability to sort through competing claims, more than ever," Garrett said. "One answer [to that problem] has to do with building tools that help us sift through this chaos."
Garrett and his doctoral student, Brian Weeks, will present their work in February, at a conference hosted by the Association for Computing Machinery.