Recommendation 23

We asked the Department about the degree of transparency that the public can expect to have about how its data analytics and machine learning tools will work. The Department told us that this was a “challenging balance”. It cautioned that it did not want to make public details about its techniques or what characteristics or behaviours it was looking for that could tip off fraudsters about how the Department identifies fraud. But it agreed that it should look at publishing what it can to give the public confidence in the fairness of these methods in future.44 The Department has previously acknowledged the 39 Qq 41–43 40 Q 41; DWP ARA 2021–22, pages 70, 230 41 Committee of Public Accounts, Department for Work and Pensions Accounts 2019–20, Twenty-Sixth Report of Session 2019–21, HC 681, 18 November 2020 42 HM Treasury, Treasury Minutes – Government responses to the Committee of Public Accounts on the Twenty- Fifth to the Twenty-Ninth reports from Session 2019–21, CP 376, February 2021 43 Q 68; DWP ARA 2021–22, page 229 44 Q 60 The Department for Work and Pensions’ Accounts 2021–22 – Fraud and error in the benefits system 17 potential for data analytics and machine learning to generate biased outcomes that could have an unfair impact on certain groups of claimants. For instance, if a machine learning 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.45