UK Government Criticized Over Use OF AI To Evaluate Benefits Claims

The UK government is coming under fire for using AI to assess applications for social security benefits.

The Department of Work and Pensions (DWP) has said it plans to use AI to identify patterns in claims that could indicate error or fraud as part of plans to cut £1.3 billion from the benefits bill this year and nex.

“Our teams are working flat out to prevent new fraudulent claims and expose people who have been exploiting the system—with strong results,” says the minister responsible for tackling fraud, Tom Pursglove.

“But we know we need to go even further, because the fraud landscape is changing, with the tactics used by criminals evolving quickly.”

During a trial last year, the department used AI to evaluate the claims of people applying for cash advantages, for items such as emergency household goods. And this trial is now to be expanded to cover applications from people living together, the self-employed, and people seeking housing support. Claims flagged by the system are put on hold until they can be reviewed by the relevant DWP teams.

However, the use of AI is attracting heavy criticism from privacy groups concerned about the transparency of the algorithm, and anti-poverty campaigners worried about the effects on vulnerable people. Campaign group Privacy International, for example, is calling for the creation of a new outside body with an oversight role.

And now the UK’s spending watchdog, the National Audit Office (NAO) has waded into the fray with a new report.

“When using machine learning to prioritise reviews there is an inherent risk that the algorithms are biased towards selecting claims for review from certain vulnerable people or groups with protected characteristics. This may be due to unforeseen bias in the input data or the design of the model itself,” reads the report.

“DWP has established tight governance and control over its use of machine learning, with safeguards designed to assess the impact that using the model has on its different customers. However, DWP says its ability to test for unfair impacts across protected characteristics is currently limited.”

It says the DWP is working to improve its fairness analysis across a wider range of protected characteristics. The department is expected to respond to the recommendations later this year.

The Labour party, meanwhile, is also pledging to use AI to tackle benefit fraud. However the party also wants to use AI to help people prepare their resumes, apply for jobs and receive benefits more quickly.

“DWP broadly gets 60% of unemployed people back to work within nine months,” says shadow work and pensions secretary Jonathan Ashworth. “I think by better embracing modern tech and AI we can transform its services and raise that figure.”