Question to the Department for Work and Pensions:
To ask the Secretary of State for Work and Pensions, whether her Department collects data on (a) errors and (b) false positives arising from algorithmic fraud detection tools.
DWP is committed to processing data lawfully, proportionately, and ethically, with meaningful human input and safeguards in place to protect individuals. “Algorithmic fraud detection tool” is not a term we use in DWP however, the department develops, tests, and invests in advanced analytics to support the detection of fraud and error. Currently, the UC Advances model is the only machine learning model deployed at scale in live service. On the 17th July, the Department published a fairness assessment of the UC Advances model, which includes consideration of the model’s performance. The model remains an effective fraud prevention control, performing approximately three times better than a control group in identifying high-risk advances.