Instagram is training a “fact-checking” AI that will suppress “misinformation”

It’s part of Facebook’s continued efforts to censor information through “fact-checking” in the run up to 2020.


Facebook-owned Instagram will start “fact-checking” posts in the US today and will use the “signals” it collects to start training its artificial intelligence (AI) to suppress “misinformation.” The fact-checking program will be rolled out to all users within the next two weeks and any posts that are labeled “false” will be scrubbed from Instagram’s “Explore” and “Hashtag” pages.

According to Poynter, a Facebook fact-checking partner, from today US users “can report false content and expect certified fact-checkers to analyze its veracity.” Instagram will then collect “signals” from users and fact-checkers to train its AI which will eventually suppress what it deems to be false information on the platform.

The example that Instagram provides suggests that there’s no room for nuance and posts that contain jokes, memes, sarcasm, or anything else that isn’t intended to be taken seriously will be suppressed if flagged by a fact-checker:

“Here’s an example of how the flagging and checking will take place: Let’s say someone posts an image saying that 2+2=5 and uses the hashtag #MathExpert to promote its publication on Instagram. The moment a certified fact-checker rates this post as false, this piece of content stops being shown at the #MathExpert hashtag page.”

Poynter adds that Instagram user won’t inform users when their posts are rated “false” which suggests the program will have a similar effect to being shadowbanned – a process where the user who posted the content can still see it and isn’t aware that it’s being suppressed. Instagram already shadowbans what it deems to be “borderline” content on the platform.

Some of Facebook’s own fact-checking partners find this decision controversial, with Tai Nalon, executive director and co-founder of Aos Fatos saying: “We have been reiterating to Facebook that the most important thing about flagging fake or distorted content is to make it clear why it happened.”

Will Moy, chief executive at Full Fact, added:

“By not telling people, by not giving the opportunity to correct a post, Facebook and Instagram weaken the program and we all move to a situation where platforms control what is being said. It is important to act openly.”

The full rollout of this Instagram fact-checking program comes after Instagram started testing the program in May. Instagram hasn’t said why it decided to start the full rollout now but when announcing the test in May, Instagram spokesperson Stephanie Otway said the system has “been in the works since the US midterm election” which indicates it could be part of a wider effort to control the content users see on the platform as the 2020 US presidential election approaches.

Recently the motives and biases of various fact-checking organizations have been brought into question. Earlier this week, Politifact, a Facebook fact-checking partner, was called out by pro-Trump comedian Terence K. Williams after it started fact-checking jokes from his page a few days after he was retweeted by President Trump. And fact-checker Snopes also faced criticism this week over its decision to introduce a “Labeled Satire” rating – a decision which came after it was called out for consistently fact-checking Christian satire site The Babylon Bee.

Despite these concerns, it appears tech giants like Instagram are committed to giving both human fact-checkers and eventually a fact-checking AI more power to suppress and censor content that they deem to be “false.”


Tom Parker

Tom Parker is a head contributor for Reclaim The Net and provides news and analysis on how we can promote free speech, stop censorship, and protect our personal data online. [email protected]
Do NOT follow this link or you will be banned from the site!