More human contact is needed in the UK’s automated welfare system, ministers have been warned, as it emerged 350 low-paid workers every day are raising complaints about errors in welfare top-ups, causing financial hardship and emotional stress.
The Department for Work and Pensions (DWP) handled 126,286 disputes about errors made by its decade-old automated “real-time information” (RTI) system in 2022, a freedom of information request revealed.
The technology is a key cog in universal credit (UC) – the UK’s main welfare system – and delivers a flow of earnings data from tax to benefit offices to automatically adjust workers’ welfare top-ups.
When it goes wrong, claimants have described it as “hellish” and “horrible”. Others say the “digital by default” approach is an improvement on an earlier, less responsive system.
The problems particularly affect single mothers bringing up children and working part-time, according to researchers at the University of Edinburgh, who obtained the complaints data using transparency laws. DWP has been criticised for its lack of openness about the way its automated and artificial intelligence systems work and how they affect claimants.
Emily, an administrator, said she was docked £300 when her employer filed the wrong monthly salary amount. She could not sleep as it plunged her into a financial crisis. The error was not fixed for two months.
She said: “The whole system is … an absolute shambles. They are causing more hardship.”
Jennifer, a single mother of two who works in a school canteen, suffered delayed benefits because the system could not adapt to her being paid every four weeks, not monthly. In one message she wrote to the UC system, she said: “Is there a crisis place that can help me with school uniform or school dinners? My cooker’s broken. I have £76 in my account to do me – all bills, shopping … I don’t understand why I am worse off on universal credit. It’s horrible. Never been this skint in my life.”
If each complaint came from a single claimant it would mean one in 18 working welfare claimants have raised complaints in the last year. More will have been affected without raising it formally.
The researchers, Morgan Currie and Lena Podoletz, described this as a “high level of error” with a “huge human impact”.
“Among the people we interviewed, it regularly took over two months for their disputes to be settled,” they said. “For a middle-class household, with some savings to tide them over until the dispute is resolved, this would not be a major problem. But for the people on universal credit, who are living on the breadline, this can lead to extreme hardship.”
The complaints come amid increasing use of automation and artificial intelligence in welfare delivery in Britain. Five years ago, the United Nations rapporteur on extreme poverty, Philip Alston, warned of the “disappearance of the postwar British welfare state behind a webpage and an algorithm”.
“In its place, a digital welfare state is emerging,” he said. “The impact on the human rights of the most vulnerable in the UK will be immense.”
The researchers are calling for more human contact in the system plus “a public register that describes any systems or algorithms that are used in the delivery of UC to make this process transparent”.
The RTI system is not believed to use artificial intelligence, but last year the DWP said it had been trialling “a machine learning algorithm” to detect fraud in claims for UC advances. It makes predictions based on historical fraud and error data “without being explicitly programmed by a human being”.
The National Audit Office has said the government “is aware of the potential for such a model to generate biased outcomes that could have an adverse impact on certain claimants”.
Errors would be inevitable, it said, and “if the model were to disproportionately identify a group with a protected characteristic as more likely to commit fraud, the model could inadvertently obstruct fair access to benefits”.
The DWP has so far declined to release further information about how it works, or any results of the trial including any unfair bias that may have been found. It has cited the need to protect the system against crime – most likely fraud by individuals or organised crime.
The department has been approached for comment.