Study reveals the simple way people get around Facebook's fact

Recently, Facebook has been taking a harder stance on misinformation.

The company banned content related to the QAnonconspiracy theory and cracked down on coronavirus misinformation. But it's still not enough.

According to a recent studyby the non-profit advocacy group Avaaz, Facebook is failing in a major, basic way. Facebook Pages that spread misinformation are finding their way around one of the platform’s most important tools for fighting fake news: its AI system.

When Facebook’s fact-checkers debunk a claim in a post, its AI is supposed to flag and label alternative versions of the post spreading the same misinformation. But the study says Pages are getting around these fact-checks.

SEE ALSO:Facebook's 'monopoly power' hurts user privacy, finds Congress

How? By slightly tweaking the photos and memes used to spread misinformation.

Avaaz’s researchers looked into 119 “repeat misinformers” – pages that have spread misinformation a minimum of three times – to understand how these pages get around Facebook’s AI detection.

Turns out, all they have to do is change the background color or font on the photo or meme they’re sharing. They can also change up the location of the text on the meme or try cropping it.

Mashable Light SpeedWant more out-of-this world tech, space and science stories?Sign up for Mashable's weekly Light Speed newsletter.By signing up you agree to our Terms of Use and Privacy Policy.Thanks for signing up!

Below is an example from the study showing two pieces of content spreading the same fact-checked claims. The image on the left just needed to change the format and text placement of the image on the right to avoid the fact-check label from Facebook.

A fact-checked meme could easily avoid a Facebook warning label by just tweaking some attributes.Credit: AVAAZ

Another workaround is to simply take the text from an image and copy and paste that same text on top of a different image or meme. Or they can take that copied text from a meme and simply paste it as a Facebook status. All of these methods will avoid Facebook’s AI detection, which means the content will appear without a fact-check or warning label.

Avaaz estimates that these 119 “repeat misinformer” Pages amassed 5.2 billion views from August 2019 to August 2020. Alternative versions of fact-checked posts had around 141 million views with 5.6 million interactions.

The study points out that the misinformation attacked both presidential candidates, Donald Trump and Joe Biden, and other Democrats and Republicans. However, there is a clear partisan bent on the list of “repeat misinformer” Pages that received the most interactions. Eight out of 10 of them shared right-wing content.

Alongside pages like “Conservative Values” and “President Trump Fans” spreading misinformation is a sitting Congressman’s official Facebook page: Republican Rep. Steve King.

This is particularly concerning being that politicians are exempt from many of Facebook’s rules. One such policy, which has been routinelycriticized since it was announced late last year, basically allows public servants to lie on the social media platform.

In fact, the one way a politician would be fact-checked on the platform is if they’re sharing content that has already been fact-checked. However, the workarounds uncovered in Avaaz’s study help politicians avoid even that.

Avaaz presented its findings to Facebook before publishing the study. The group says Facebook added a warning label to 4 percent of the 738 posts it flagged. It removed another 3 percent. The majority remain on the platform without a label.

最新评论