How artificial intelligence on Whatsapp can help fight disinformation
Issued on: Modified:
Since the start of the Covid-19 pandemic in 2020, there has been massive dissemination on social networks of information that is false, manipulated or lacking in context. To address a surge in questions from the public, a team from the Spanish fact-checking organisation Maldita created a WhatsApp chatbot, which uses artificial intelligence to provide automated answers. You will find more information in our latest episode of Truth or Fake.
The tagline of Spanish fact-checking outlet Maldita puts readers at the centre of the team's journalistic work: the Spanish phrase "Hazte Maldito" (meaning "Be part of Maldita!") invites the public to send in potentially fake news items and ask questions about the virus.
Before the pandemic, Maldita received about 200 messages a day on their WhatsApp number, occupying a full-time journalist. After the pandemic started in March 2020 in Europe, their daily messages increased to nearly 2,000.
Maldita has launched a WhatsApp chatbot to automate and centralize their interactions with their community. After a user sends in a social media post to the WhatsApp number - either a photo, a video, a link, or a WhatsApp channel that's been sharing questionable content, the bot analyses the content. If Maldita has already published a fact check on the given item, the user automatically receives a link to the debunking article. If the content has not yet been verified, the request is sent to the team's journalists, who receive an aggregation of similar requests.
Three success stories of the WhatsApp bot
The bot has allowed Maldita's journalists to detect fake information when it is being shared on the private messaging app's groups, closed groups that are limited to 256 members. Misinformation often circulates among private messaging groups before reaching open social networks like Twitter and Facebook. Maldita's journalists say the bot allows them to detect false information at the critical point before it goes fully viral on social networks Let's see three examples of how it works.
In the midst of the US presidential election, on November 11, 2020, Maldita's editorial staff received many questions on their bot about a voice message in which the Spanish foreign minister, Arancha González Laya, supposedly confused Joe Biden with Osama Bin Laden in a congratulation message for the Democratic candidate.
In reality, the message was not issued by the minister. Maldita was able to contact her office and compare audio excerpts to promptly establish that it was false. They published a verification article that same evening, 24 hours before the false audio message hit social networks, which they hope diminished its virality.
Similarly, Maldita's bot detected a video of a session at the Danish Parliament whose subtitles were manipulated twice since March 2021. A first version featured subtitles suggesting that the MPs were making fun of the Spanish monarchy; a second version claimed they were making fun of a proposal to recognise Catalonia as an independent country. In reality, as this video shows, the Danish prime minister was not talking about Spain, but about the the Danish government's purchase of four elephants from circuses so they could retire.
Lastly, the bot also allows journalists to be reactive on health issues that can cause potential harm. On February 11, 2021, in the middle of a televised debate before the elections for the presidency of Catalonia, hundreds of messages were sent back by the bot about a positive PCR test for Covid-19 in the name of Salvador Illa, one of the candidates present on the panel.
On this very sensitive matter, journalists were able to start working while the debate was still going on, and to publish a fact-check the very next morning: the positive PCR test had been faked.
To find the other episodes of Fact or Fake, click here!
If you want to report an image or a video that caught your attention and that you want to verify, contact us by message on the Twitter account @InfoIntoxF24.