• Facebook
• Instagram
• LinkedIn
• Twitter

The Spanish elections

The Spanish regional elections are still almost four months away, but Irene Larraz and her team at Newtral are already prepared for the impact. Every morning, half of Larraz’s team from the Madrid-based media company sets a schedule of political speeches and debates, preparing to verify politicians’ statements. The other half unmasks disinformation, analyses the web for viral falsehoods and works to infiltrate groups that spread lies. After the elections in May, a national election will have to be called before the end of the year, which will probably cause a wave of online falsehoods. ‘It will be very difficult,’ says Larraz. “We are already preparing for it.”

The proliferation of disinformation and online propaganda has meant an uphill battle for fact-checkers around the world, who have to sift through and verify large amounts of information in complex or rapidly changing situations, such as the Russian invasion of Ukraine, the Covid-19 pandemic or election campaigns. This task has become even more difficult with the advent of chatbots using large language models, such as OpenAI’s ChatGPT, which can produce natural-sounding texts at the click of a button, essentially automating the production of disinformation.

Faced with this asymmetry, fact-checking organisations need to build their own AI-based tools to automate and accelerate their work. This is not a complete solution, but fact-checkers are hoping that these new tools can at least prevent the gap between them and their opponents from widening too quickly, at a time when social media companies are reducing their moderation operations.

‘The race between fact-checkers and fact-checkers is unequal,’ says Tim Gordon, co-founder of Best Practice AI, an artificial intelligence strategy and governance consultancy, and trustee of a UK fact-checking charity.

“Fact-checkers are often tiny organisations compared to those that produce disinformation,” says Gordon. “The scale of what generative AI is capable of producing, and the pace at which it can do it, means that this race will become increasingly difficult.”

Newtral began developing its multilingual AI language model, ClaimHunter, in 2020, funded by profits from its TV wing, which produces a fact-checking show about politicians and documentaries for HBO and Netflix.

Using Microsoft’s BERT language model, the developers of ClaimHunter used 10,000 statements to train the system to recognise sentences that appear to include statements of fact, such as data, numbers or comparisons. “We taught the machine to play the role of a fact-checker,” explains Rubén Míguez, chief technology officer at Newtral.

Simply identifying claims made by political figures and social media accounts that need to be verified is a daunting task. ClaimHunter automatically detects political statements made on Twitter, while another application transcribes politicians’ videos and audios into text. Both identify and highlight statements that contain a statement relevant to public life that can be proven or refuted – that is, statements that are not ambiguous, questions or opinions – and flag them to Newtral’s fact-checkers for review.

The system is not perfect and sometimes marks opinions as facts, but its errors help users to continuously retrain the algorithm. According to Míguez, the system has reduced the time needed to identify statements for verification by 70-80%.

“Having this technology is a huge step forward to listen to more politicians, find more facts to verify [and] debunk more misinformation,” Larraz says. “Before, we could only do a small part of the work we do today.”

Newtral is also working with the London School of Economics and broadcaster ABC Australia to develop a claim-matching tool that identifies false and repeated statements by politicians, saving fact-checkers time by recycling existing clarifications and articles debunking claims.

The quest to automate fact-checking is not new. The founder of the American fact-checking organisation Politifact, Bill Adair, first experimented in 2013 with an instant verification tool called Squash at Duke University’s Reporters’ Lab. Squash compared politicians’ speeches live with previous fact-checks available online, but its usefulness was limited. It did not have access to a large enough library of verified documents to cross-reference statements, and its transcripts were riddled with errors that humans had to double-check.

“Squash was an excellent first step that showed us the promise and challenges of live fact-checking,” Adair tells WIRED. “Now we need to combine what we did with new advances in artificial intelligence and develop the next generation.”

But a decade later, fact-checking is still far from being fully automated. Although large language models (LLMs) such as ChatGPT are able to produce text that appears to be written by a person, they are unable to detect the nuances of language and have a tendency to make things up and amplify biases and stereotypes.

“LLMs don’t know what facts are,” says Andy Dudfield, head of artificial intelligence at Full Fact, a UK fact-checking association, which has also used a BERT model to automate parts of its fact-checking workflow. “Fact-checking is a very subtle world of context and caveats.”

• Facebook
• Instagram
• LinkedIn
• Twitter

From Politics to Hotel Rooms

Let's bring this back to the topic of the article. Could we use the same technology to understand the quality of a hotel service? To understand whether our money will be well spent in that room, in that restaurant, in that bar?

In A.I.LoveTourism's opinion, it will all depend on how much the software will remain in 'open' form and not turn into the usual paid service.

Essentially, for two reasons: 1) gratuitousness will allow continuous data and information to flow into Bert's library, with greater accuracy in fact-checking; 2) an 'open' system guarantees the neutrality of the service itself, with objective accuracy in the search for the veracity of reviews.