Just recently, Facebook has been banning content that is related to the whole QAnon conspiracy theory as well as coronavirus misinformation. According to a study by a certain non-profit advocacy group called Avaaz, Facebook is actually failing as certain Facebook Pages still found a way to spread fake news and go around the whole AI system.
The known way around Facebook's fact-checking AI
According to Mashable, the way to get around the AI fact check is actually by the use of either photos or memes in order to spread misinformation. Avaaz researchers were able to look into 119 different "repeat misinformed" or pages that have actually spread misinformation over three times, in order to actually understand how they were able to get past Facebook's AI detection.
It was found that in order to get past detection, all that needs to be done is to change the background color or even the font of the meme or photo that is being shared. The location of the whole text or meme can also be changed and it can also be cropped. Another technique is actually to take a text from an image and also copy paste it on the top of another image or meme.
The findings of "repeat misinformer" Facebook pages
Avaaz was able to estimate that the whole 119 "repeat misinformer" Facebook pages were actually able to gain 5.2 billion views all the way from August 2019 up to August 2020. There are also alternative versions of the whole fact-checked posts that actually gained about 141 million views as well as 5.6 million interactions.
In short, the most used way to get past Facebook's misinformation and fact-checking AI is through the use of images. Whereas texts can easily be scanned and pop up instantly, images actually cannot and Facebook still uses human content moderators in order to identify some images.
Read Also: Twitter Bans Tweets Hoping Trump Dies or Anyone; Won't Tolerate Passive Death Threats
Why do images get past Facebook's fact-check AI?
The thing about using images is that content moderators usually look at an image and attack about five different words related to that image. Once those words are searched, then the image pops up. However, when it comes to memes or text images, it's hard to point out what they mean and they could sometimes slip past content moderators.
With the use of images on Facebook, as long as the background picture is not gory or does not violate their content as well, they could very well be used to get past the Facebook fact-check AI. Although according to Daily Mail, they are extending efforts to fight against the spread of misinformation through the use of reverse image searching. But for original photos, they should still be able to bypass this.
Although using graphic content is strictly prohibited (but sometimes still get through) text images, especially those with picture backgrounds, can still be a loophole for people to get past the whole AI fact-check and might even be the reason for certain parties, groups of people, or even individuals spreading misinformation.
Read Also: 5 Google Maps Hacks to Use the App for More than Just Directions
This article is owned by Tech Times
Written by Urian Buenconsejo
