How is Facebook using AI to fight fake news?

The fight against Fake News is not stopping any time soon

‘Fake news’ has been a buzzword in politics for the past 3 years as the spread of misinformation has become a big point of contention in Western democracies. Partly due to it being Donald Trump’s favourite term in the lead up to the 2016 Presidential Election, recognition of fake news spreading across social media platforms has grown. The platform criticised the most for aiding this spread is Facebook. Following the results of the 2016 US Election, many have blamed Mark Zuckerberg for making it easy for trolls and Russian hackers to manipulate voters into favouring Republicans, through made up articles and posts defaming Hillary Clinton and the Democrats. So how does Mark Zuckerberg plan on fixing the immense fake news problem?


It was recently announced that Facebook has expanded its fact-checking program to include the use of machine learning, a type of AI, to fact-check images, videos and text posts. The company’s algorithms will scan viral posts to spot any information that could be deemed untrustworthy - this includes plagiarised sentences, suspicious ads and facts that differ from trusted news sources. If the algorithm flags a post as fake, Facebook implements machine learning to recognise and delete identical stories that have been posted across different news pages. Facebook plans to introduce an improvement to this system: filters that predict which pages are more likely to share false stories, based on geographical locations and previous behaviour. Stricter enforcement of bans is also being implemented for pages that repeatedly violate community guidelines by sharing hoaxes. Additionally, the platform is working on improving technology that can analyse an image’s metadata, which shows further information on the origin of an image to be compared with the context of its use in a story.


However, machine learning is still hugely flawed. A key problem is that it requires time and practice to perfect machine learning algorithms, and there are constant inconsistencies in correctly identifying false information. The technology needs huge advancements in order to keep up with the masses of content uploaded to the platform each day (Facebook currently has 1.47 billion daily active users). Some also argue that it is simply impossible to rely so heavily on automation to tackle fake news due to false flagging. However, human moderators, the network’s previous solution, are incapable of overlooking millions of posts a day. A new challenge threatening fake news detectors is the rise of ‘deepfakes’ - videos and images that have been altered to look like a person saying or doing something they haven’t. This artificial intelligence, which is also repurposed for Snapchat and Instagram filters, is used in deepfakes to spread false political statements using the faces of influential figures such as politicians and celebrities. The realistic appearance of these creations makes it increasingly difficult for AI used by Facebook to detect what is real and fake, so a solution to fighting fake news seems increasingly implausible.


Anjale VII