DWP confirms that it has no plans to publish details of algorithms used in trials to detect fraud in claims for universal credit advances

Information about the nature and operation of algorithms could be used to ‘defraud the benefit system and impact the public purse’, says Minister

DWP Minister Tom Pursglove has confirmed that the Department has no plans to publish details of the algorithms used in trials to detect fraud in claims for universal credit advances.

Responding yesterday to a written question in Parliament from Shadow Work and Pensions Secretary Jonathan Ashworth, Mr Pursglove said that –

'… we currently have no plans to publish details of either the (a) nature or (b) operation of the machine learning algorithms used in trials to detect fraud in claims for universal credit advances. Similarly, we also have no plans to publish the results of trials using machine learning algorithms to detect fraud in claims for universal credit advances. It is not in the public interest to publish, as it contains information that fraudulent actors could use to defraud the benefit system and impact the public purse adversely.'

Confirming that the DWP has conducted an equalities assessment of the trials using algorithms to detect fraud in claims for universal credit advances, Mr Pursglove added that –

'The Department has robust processes to ensure ethical use and impact of data is considered, which includes Equality Impact Assessments for large-scale transformative initiatives that involve personal data, aligned with data-ethics frameworks, codes of practice, and working principles for analytical communities within the Department that work with personal data.

The DWP’s Personal Information Charter ensures that its customers are aware of the DWP’s use of Artificial Intelligence.'

Mr Purslgove also said that it is important to note that the Department does not use use algorithms to make decisions regarding fraudulent claims – which are always made by humans – and that –

'The Information Commissioner’s Office have indicated publicly that they are broadly supportive of the current use of AI within the welfare benefit system, based on sampling they have undertaken.'

Mr Pursglove's written answer is available from parliament.uk

Update (12 June 2023) – in a further written answer today, Mr Pursglove confirmed that the Department will not publish its equalities assessment of the trials using algorithms to detect fraud in claims for universal credit advances because the information it contains 'would allow fraudsters to understand our services and the way our IT systems work'.