Thousands of people could miss out on Universal Credit as DWP uses a COMPUTER to decide on claims
THOUSANDS could miss out on Universal Credit payments because of a new automated process to detect fraud, which is set to be trialled by the government.
The Department of Work and Pensions wants to tackle fraudulent Universal Credit applications.
To do this, it is trialing a system that uses a machine learning algorithm to detect cases of benefits fraud.
An algorithm is a process or set of rules which can be followed in calculations or other problem-solving operations, especially by a computer.
The government’s algorithm could stop payments from being sent to people if it thinks their applications are fraudulent.
While campaigners support the crackdown on fraud, there are concerns that using a machine-learning algorithm to decide on benefits claims could leave people unfairly out of pocket.
Read more about benefits
The DWP has admitted it would be “unavoidable” that some cases flagged as potentially fraudulent would turn out to be legitimate claims under the new system.
If people do not get their money they could be left in serious trouble – especially as the cost of living crisis bites.
How does the algorithm work?
The plans were outlined in the DWP’s annual report and accounts as well as a report from the National Audit Office.
The government has already trialled a system which looks at historic fraud cases to predict which are likely to be fraudulent in the future.
Most read in Money
In this model, when the algorithm decides a case is potentially fraudulent, it flags it to human caseworkers so they can investigate.
This system, which was trialled last year, works on cases where people have already received benefits payments.
But the DWP is planning to trial a new system in 2022-2023 which will try and spot fraud cases BEFORE any benefits payments are sent out.
“If successful, this could improve its ability to prevent fraud before these benefits are paid out, avoiding the need to seek recovery,” the DWP’s report said.
It means some people might not receive payments – because the algorithm has decided their claims are fraudulent.
The DWP acknowledged that sometimes the algorithm would get things wrong.
“It is unavoidable that some cases flagged as potentially fraudulent will turn out to be legitimate claims,” the report said.
“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 said it would do pre-launch testing and continuous monitoring of the algorithm to try and mitigate these risks.
It also said that although the model identifies potentially fraudulent claims, the final decision about whether a claim is legitimate would always be made at the discretion of one of the Department’s caseworkers.
Despite this, campaigners have warned the system could cause problems for vulnerable people.
“Using a machine-learning algorithm to detect fraud in Universal Credit claims creates a risk that people will be unfairly left out of pocket,” said Ariane Adam, legal director at the Public Law Project.
“In the midst of a cost-of-living crisis, this could penalise and discriminate against marginalized and vulnerable groups.”
Exactly how the DWP’s algorithm decides that an application is fraudulent is not yet clear.
Flawed systems
Back in 2020, an international watchdog warned that an algorithm used to automatically calculate Universal Credit payments was pushing vulnerable households into hunger and debt.
The calculation worked out how much people were entitled to but actually left many worse off, according to the Human Rights Watch (HRW) found.
The algorithm was supposed to adjust payments monthly based on changes to a claimant’s income but in reality it only took into account wages received in the same calendar month.
It didn’t make allowances for those who were paid multiple times in a month, such as weekly, fortnightly or every four weeks, which could leave households out of pocket.
Read More on The Sun
The fault is estimated to have affected 85,000 claimants who were sometimes paid twice in the same assessment period, for example, because payday would normally fall on a weekend or bank holiday.
There is nothing to suggest that the government’s new algorithm would be flawed – but it demonstrates how people can be affected if things go wrong.