New AI Tool Lets You Test Whether Politicians Repeat False Claims
Full Fact hopes its tool can help identify and tackle false information ahead of multiple general elections across the world (Alamy)
5 min read
An influential fact-checking organisation is rolling out an AI tool to try to monitor and prevent an onslaught of misinformation and disinformation ahead of the UK's upcoming general election this year.
Full Fact, a leading UK organisation which fact-checks claims made by public figures and users of online platforms, has been expanding the rollout of a new AI-powered tool which enables a quicker and more efficient way to filter and identify online misinformation.
They began to develop the tool in 2019 and in the years since have been expanding the number of clients who subscribe to access and use it. As well as in the UK, Full Fact's tool is now being used by fact-checkers from South Africa, Nigeria, Kenya, Senegal, Australia, Egypt, Iraq, Jordan, Lebanon, Libya, Palestine, Sudan, Tunisia, and Yemen.
By using AI to filter through vast amounts of online data from news websites and major social media platforms, the tool helps users find and monitor the most important things to fact-check on a daily basis. It allows you to be able to see what is being said in a public debate and who is saying it, with different features which allow you to search by topic, speaker, or political party.
"AI really allows the media monitoring and claim identification to be taken off the hands of fact-checkers, so that they can focus on doing what they do best, which is understanding context and nuance, finding experts, looking at data, analysing and understanding it and then importantly, communicating their findings to communities that are affected by misinformation," Full Fact's senior product manager Kate Wilkinson said.
"Full Fact has really invested in this space as far back as 2012, because it's really understood that we cannot counter the problem of misinformation at a human scale. You can hire all the fact-checkers you want, you can have them work as many hours as you want, but the scale of that information online is just so large, and it's generated and spread so quickly, that we need a solution that is scalable and uses tech to amplify the work that we're doing."
She explained that the tool also identifies where false information is repeated online or by public figures – a very common phenomenon where misinformation is found to spread quicker than correct information.
"This is an incredibly powerful tool for fact-checkers because when you do the work of fact-checking you very quickly learn that just fact-checking a claim once and publishing a report doesn't mean that a falsehood is never going to be repeated," Wilkinson continued.
"It often takes repeated effort and a very conscious effort to correct a misunderstanding in the public discourse about a topic. So what our tool does is every time we fact-check a claim, it automatically adds what we have fact-checked to the tool, and it starts comparing that to all the media that we're monitoring on a daily basis."
An example of how the AI tool can be used to monitor claims online (Full Fact)
Wilkinson said that they had fact-checked a claim made by Boris Johnson when he was prime minister where he had said that employment levels had rebounded and were higher than pre-pandemic levels.
"We found that that was only correct for payroll employment; that when you took into account all of employment, including self-employed people, that wasn't correct," she said.
"And we then tracked that claim using this tool and we were able to identify numerous repeats by MPs. I have a whole list of the repeats.
"A lot of our work is not just fact-checking, but actually actively campaigning, advocating and engaging with politicians, MPs, people in positions of power to correct the record when they made an error. This tool flags those repeated falsehoods, so that we are able to understand how misinformation is spreading, who is amplifying it and how it's changing, but then also act to change the ecosystem that allows that misinformation to flourish and be shared."
As the UK approaches a general election by the end of this year, alongside important elections in many other countries, Wilkinson argued it was more important than ever for fact-checkers to use technology to navigate the risks of misinformation more effectively.
"We genuinely believe that tech and specifically AI has the potential to transform fact-checking and transform the wider fight against false information," she said.
While it's mostly other fact-checking organisations that have so far used the tool, a number of research and advocacy organisations and charities have also used it to monitor misinformation around their topics. The Office for Statistics Regulation (OSR) is one such client, using it to find cases where official statistics were being misrepresented or misused so that they could intervene if needed.
There are potential gaps, however, particularly on how to monitor misinformation on encrypted messaging apps such as Whatsapp.
"I think that is a challenge, it will continue to be a challenge, but what we have seen is that misinformation does tend to leak between channels," Wilkinson said.
"So it's something that we're going to need to invest in and to try and bring as much monitoring onto the platform as possible. But I expect to be an ongoing challenge given the state of current social media platforms and new ones which will pop up."
As much of the public and political discourse over the last 18 months has been centred on the challenges and risks surrounding AI technology, Full Fact is keen to present how it can also present solutions.
However, Wilkinson said it was important to also have humans in the loop to prevent bias across AI tools.
"We were always very conscious that we didn't want to build this tool from a purely UK perspective, with UK media and with UK fact-checkers, because that would create a tool which would have a regional focus bias and it wouldn't be applicable to fact checkers around the world," she continued.
"So that's why we bought in fact-checkers from African countries and South American countries... they were all involved in producing the training data to build the algorithms that we use."
PoliticsHome Newsletters
PoliticsHome provides the most comprehensive coverage of UK politics anywhere on the web, offering high quality original reporting and analysis: Subscribe
