Algorithmic Bias in Benefit Claims
Concerns regarding the potential for data analytics and machine learning algorithms used in benefit claims processing to result in unfair treatment.
770 items
5 sources
Strongest theme matches
Mixed across source types and ranked by classifier confidence plus text match strength.
Committee recommendation
100match
#7 - Share fairness impact assessment results for machine learning to reassure against unfair claimant treatment.
We remain concerned about the potential negative impact on protected groups and vulnerable customers of DWP’s use of machine learning to identify potential fraud. The previous Public Accounts Committee repeatedly raised concerns about the impact of data analytics and machine learning on legitimate benefit claims being delayed or reduced, the number of people affected, and whether this is...
Matched on
terms: benefit, bias, claim
Committee recommendation
100match
#31 - Algorithmic bias in DWP systems acknowledged, but evidence of unfair impacts remains inconclusive.
We challenged DWP to explain how it would address the risk that legitimate benefit claims are unfairly delayed or reduced as a result of an algorithms targeting innocent behaviour, such as frequent changes of circumstances. DWP acknowledged that some level of algorithmic bias is to be expected because of how benefit payments work, for example Universal Credit payments...
Matched on
terms: algorithmic, benefit, bias, claim
Committee recommendation
100match
#30 - Concerns persist regarding machine learning's potential unfairness and bias for vulnerable claimants.
We received written evidence from the Child Poverty Action Group and from the Public Law Project expressing concern about the potential unfairness of machine learning, particularly with regard to vulnerable claimants and people with protected characteristics.70 We asked DWP whether it understood the concerns of people who have warned of unintentional bias in its use of machine learning....
Matched on
terms: benefit, bias, claim
Committee recommendation
98match
#37 - Public Law Project warns of bias and transparency risks in DWP's machine learning for fraud.
However, written evidence we received from the Public Law Project raised a series of risks around DWP’s use of machine learning to tackle fraud. In particular, it noted the risk of machine learning taking on human biases when it was trained on historical data and the potential for widescale detrimental impacts on claimants if there was a system...
Matched on
terms: benefit, bias, claim
Committee recommendation
95match
#6 - Assess impact of data analytics and machine learning on legitimate claims and specific groups.
DWP has not yet done enough to understand the impact of machine learning on customers and provide them with confidence that it will not result in unfair treatment. DWP is expanding its use of advanced data analytics to tackle fraud. This includes machine learning algorithms to flag potentially fraudulent benefit claims, so the system learns and adapts without...
Matched on
terms: benefit, bias, claim
Committee recommendation
93match
#22 - Twenty-Sixth Report - The Department for Work and Pensions’ Accounts 2021–22 – Fraud and error in...
As part of our previous examination of the Department’s 2019–20 Accounts, we recommended that the Department should monitor and report any discrimination or bias caused by using artificial intelligence and machine learning on different claimant groups.41 In its response to our report, the Department agreed with our recommendation and told us that it had a Data Science Ethics...
Matched on
terms: benefit, bias, claim
Committee recommendation
90match
#23 - Twenty-Sixth Report - The Department for Work and Pensions’ Accounts 2021–22 – Fraud and error in...
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...
Matched on
terms: benefit, bias, claim
Committee recommendation
87match
#5 - Twenty-Sixth Report - The Department for Work and Pensions’ Accounts 2021–22 – Fraud and error in...
The Department’s lack of transparency over its use of data analytics risks eroding public trust in the benefit system. The Department’s strategy to bring down fraud and error will depend increasingly on the use of data analytics and machine learning to identify potentially fraudulent claims. It has trialled a model to detect fraud in Universal Credit advances. This...
Matched on
terms: benefit, claim
Committee recommendation
85match
#13 - Twenty-Sixth Report - The Department for Work and Pensions’ Accounts 2021–22 – Fraud and error in...
We asked the Department how it ensured that vulnerable claimants were taken into account, in particular when it uses data analytics and machine learning. The Department told us that it was considering and testing for vulnerability “at every stage”.23 It acknowledged that that its data analytics tools focused on characteristics for fraud rather than vulnerability, but that it...
Matched on
terms: benefit, claim
Committee recommendation
83match
#36 - DWP uses machine learning for fraud, but with limited transparency on fairness assessment.
DWP is implementing machine learning techniques to help it identify fraud in benefit expenditure. It has one machine learning model in operation, for new UC advance claims, alongside several others in development.69 The previous Public Accounts Committee raised concerns about the level of transparency in DWP’s use of these tools and the potential impact on claimants who are...
Matched on
terms: benefit, claim
Committee recommendation
73match
#24 - Twenty-Sixth Report - The Department for Work and Pensions’ Accounts 2021–22 – Fraud and error in...
The Department explained that, in order to manage the risk of unintended bias in the use of data analytics to identify fraud, it ensured that there was always meaningful human involvement in decision-making, and that it undertook ‘fairness analysis’ to identify any disproportionate impacts. We asked the Department which groups the fairness analysis had been conducted for. It...
Matched on
terms: benefit, bias
Committee recommendation
63match
#43 - Mandate human review and verification of AI-assisted decisions affecting individuals
The potential for human error and bias notwithstanding, deployers should not solely rely on outputs produced with AI tools to determine their decision-making, particularly in areas that could affect the rights and standing of the individuals or entities concerned, such as insurance decisions or recruitment. These algorithmic decisions should always be reviewed and verified by trained humans, and...
Matched on
terms: algorithmic, bias
Committee recommendation
56match
#38 - DWP commits to further assurance on fairness of machine learning techniques for benefits.
We stressed that we would like to be satisfied that DWP’s use of machine learning is fair and consistent. DWP highlighted its main safeguard, which is that it would never stop benefit payments to a customer due to an AI tool. It said that what AI helped it to do was to direct the work of its staff...
Matched on
terms: benefit
NAO recommendation
56match
DWP Report on Account 2022/23
g as part of its response to PAC due in November 2023, DWP should seek to improve public confidence in its use of advanced data analytics by committing to regularly publishing summaries of: ? its assessment of bias in machine learning models; and
Matched on
terms: bias
Committee recommendation
53match
#14 - Provide detailed plans for 'Jobcentre in your pocket' outlining public expectations and journal interaction.
In its response to this report, DWP should provide more details about its plans for a ‘Jobcentre in your pocket’. DWP should set out what the public can expect from a ‘Jobcentre in your pocket’, how it will interact with the current Universal Credit journal and how it will ensure it delivers a successful digital product that meets...
Matched on
terms: claim
Committee recommendation
53match
#18 - 75th Report - Government use of data analytics on error and fraud
To ensure transparency, government bodies are required to disclose the use of algorithms, AI and machine learning through the Algorithmic Transparency Recording Standard (ATRS) hub.52 As at February 2026 the hub, which requires records of any such items used in decision making, recorded 110 being used in central government. Of these records, 11 mention ‘fraud’.53 None of the...
Matched on
terms: algorithmic
Committee recommendation
52match
#33 - Require AI model developers to summarise bias mitigation and report output bias levels.
Model developers and deployers should be required to summarise what steps they have taken to account for bias in datasets used to train models, and to statistically report on the levels of bias present in outputs produced using AI tools. This data should be routinely disclosed in a similar way to company pay gap reporting.
Matched on
terms: bias
Committee recommendation
52match
#20 - Extend Algorithmic Transparency Standard to all public bodies from January 2025.
The requirement for Government departments to use the Algorithmic Transparency Recording Standard should be extended to all public bodies sponsored by Government departments, from 1 January 2025. (Paragraph 77) The AI Safety Institute
Matched on
terms: algorithmic
Committee recommendation
47match
#4 - Second Report - The Cost of Living
The systems that legacy benefits run on are not fit for purpose. It is disappointing that the Department has not adapted its IT systems to allow for flexibility in uprating these benefits to respond to national events. The Department has scheduled and then delayed the migration of claimants from legacy benefits onto Universal Credit several times and the...
Matched on
terms: benefit, claim
LGO / SPSO decision
47match
23-018-129 - Salford City Council
Summary: We will not investigate Ms X’s complaint about the Council failing to ask her for relevant information which she considers would have affected her welfare benefit claims. This is because the actions complained about are not the administrative function of a council.
Matched on
terms: benefit, claim
Committee recommendation
45match
#13 - DWP's 'Jobcentre in your pocket' digital tool lacks definition and fails to address AI risks.
DWP’s solution to a lack of work coach capacity is to use digital tools, with its flagship plan being a ‘Jobcentre in your pocket’. DWP plans to deliver this digital tool in 2027–28, but has yet to define what it is, how it will work and who it is for. While it is good to see the Government...
Matched on
classifier match
Committee recommendation
43match
#15 - Include warnings on Lifetime ISA products about inferiority for potential Universal Credit claimants.
If the Government is unwilling to equalise the treatment of the Lifetime ISA with other Government-subsidised retirement savings products in Universal Credit assessments, Lifetime ISA products must include warnings that the Lifetime ISA is an inferior product for anyone who might one day be in receipt of Universal Credit. Such warnings would guard against savers being sold products...
Matched on
terms: claim
Committee recommendation
41match
#15 - Develop an ethical AI framework for employment support in collaboration with the sector.
DWP should work with the employment support sector to develop a framework for the ethical use of AI in employment support. (Recommendation, Paragraph 78) ‘Good jobs’ not ‘any jobs’
Matched on
classifier match
Committee recommendation
39match
#32 - Require robust, independent testing and performance analysis for AI models prior to deployment.
AI can entrench and accelerate existing biases. The current Government, future administrations and sectoral regulators should require deployers of AI models and tools to submit them to robust, independent testing and performance analysis prior to deployment. (Paragraph 140) 56 Governance of artificial intelligence (AI)
Matched on
terms: bias
LGO / SPSO decision
39match
23-006-213 - Royal Borough of Greenwich
Summary: Mrs X complained about the Council’s handling of her Housing Benefit, Council Tax, Council Tax Support, and a Severe Mental Impairment (SMI) exemption. We have concluded our investigation and find fault in how the Council handled a housing benefit appeal and how it failed to issue proper appeal rights for Council Tax Support decisions. The Council has...
Matched on
terms: benefit
LGO / SPSO decision
39match
21-007-369 - City of Bradford Metropolitan District Council
Summary: Mr B complained on behalf of Mr C that the Council had refused to forward his housing benefit appeal to the Tribunal service. We found the Council at fault, but we do not consider this caused Mr C an injustice.
Matched on
terms: benefit
Committee recommendation
35match
#23 - Twenty-Sixth Report - Department of Work and Pensions Accounts 2019-20
The Department told us that it is starting to build a system that is based on ‘transaction risking’; its vision is to be in a place where it can, in real time, or near real time, assess every claim as it is coming through and take a view of how much it trusts the information that is in...
Matched on
terms: claim
Committee recommendation
35match
#12 - Twenty-Sixth Report - Department of Work and Pensions Accounts 2019-20
Universal Credit has the highest estimated overpayment rate of all measured benefits—9.4% (£1.7 billion) for 2019–20—and it has an estimated underpayment rate of 1.1% (£0.2 billion). This is the highest recorded overpayment rate for any benefit other than Tax Credits (administered by HMRC), which peaked at 9.7% in 2003–04.23 The Department told us that that is it “not...
Matched on
terms: benefit
NAO recommendation
35match
Delivery of employment support schemes in response to the COVID-19 pandemic
While payments under the schemes have ended, the need to bear down on error and fraud has not. HMRC should: a) as part of strengthening its methodology for remaining compliance work, analyse its performance in tackling the main risks of non-compliance and identify what further data it can use to identify and tackle risky claims, including for those...
Matched on
terms: claim
LGO / SPSO decision
35match
21-016-099 - Coventry City Council
Summary: We will not investigate Mr X’s complaint about how the Council handled a reassessment of his council tax support claim. Mr X has a right of appeal to the Valuation Tribunal if he disagrees with the decision. It is reasonable for him to go to the Information Commissioner if he has a data protection complaint.
Matched on
terms: claim
LGO / SPSO decision
35match
21-015-664 - London Borough of Hounslow
Summary: We will not investigate this complaint about the way the Council calculated the complainant’s housing benefit and raised an overpayment. This is because the complainant can use her review and appeal rights.
Matched on
terms: benefit
LGO / SPSO decision
35match
22-001-808 - Birmingham City Council
Summary: We will not investigate this complaint about housing benefit overpayment which dates from 2009 to 2014. Mr X has or had the right to appeal to a tribunal against overpayment of housing benefit which it would have been reasonable to expect him to use. Additionally, we cannot investigate a complaint which is made to us more than...
Matched on
terms: benefit
LGO / SPSO decision
35match
22-000-143 - London Borough of Brent
Summary: We will not investigate this complaint alleging a 2-year underpayment of Housing Benefit. This is because the complainant has a right of appeal to the specialist Housing Benefit tribunal. And it is reasonable to expect her to use this appeal right.
Matched on
terms: benefit
LGO / SPSO decision
35match
24-021-971 - City of York Council
Summary: We will not investigate this complaint about a housing benefit overpayment because there is insufficient evidence of fault by the Council.
Matched on
terms: benefit
Committee recommendation
32match
#1 - Automated payment system provides swift support but is limited in targeting specific group needs.
We welcome the automated nature of the payment system, which enabled the swift issue of cash support to many of those most in need. However, we recognise that it is limited in its ability to target payments and therefore meet the additional needs of certain groups.
Matched on
classifier match
Committee recommendation
32match
#16 - DWP uses AI for written vulnerability screening, plans phone system for equivalent identification
DWP also told us that it had been using artificial intelligence to scan customers’ written communications for words and phrases that could indicate vulnerability so that it could provide support quickly. In subsequent correspondence, it highlighted that this technology solution had recently received the ‘Excellence in Delivery Award’ at the 2024 Civil Service Awards.31 DWP conceded that it...
Matched on
classifier match
Committee recommendation
31match
#21 - 75th Report - Government use of data analytics on error and fraud
However, the PSFA also told us that there are instances where legislation, written some time ago, does not fully support the effective deployment of modern data analytics techniques. For example, the Local Audit and Accountability Act 2014 does not allow for profiling of individuals’ behaviours. Which means in practice that data collected under that Act can be used...
Matched on
classifier match
NAO recommendation
30match
Department for Work and Pensions annual report and accounts 2018-19
• continue work to understand the reasons for the fraud and error rates for Universal Credit, including how the causes of fraud and error compare with those for other benefits and use this to determine its fraud and error strategy for Universal Credit;
Matched on
terms: benefit
LGO / SPSO decision
27match
22-002-948 - Solihull Metropolitan Borough Council
Summary: We cannot investigate this complaint about the complainant’s former employment with the Council. This is because the law prevents us from investigating employment related complaints. We will not investigate the complaint about council tax and council tax reduction because the Council has proposed a fair remedy.
Matched on
classifier match
LGO / SPSO decision
27match
25-003-185 - Westminster City Council
Summary: We cannot investigate this complaint about court action taken by the Council for council tax arrears. This is because the complainant has started legal proceedings against the Council.
Matched on
classifier match
LGO / SPSO decision
27match
25-006-838 - London Borough of Islington
Summary: We will not investigate this complaint that the amount of a council tax credit is wrong. This is because there is insufficient evidence of fault by the Council.
Matched on
classifier match
LGO / SPSO decision
27match
25-014-496 - Sefton Metropolitan Borough Council
Summary: We will not investigate this complaint about council tax
Matched on
classifier match
LGO / SPSO decision
27match
21-010-040 - Sunderland City Council
Summary: We will not investigate this council tax complaint regarding the removal of a discount and subsequent refund. This is because there is insufficient evidence of fault regarding the discount and because the Council has provided a fair remedy in relation to the refund. The complainant’s request for information is a matter for the Information Commissioner.
Matched on
classifier match
LGO / SPSO decision
27match
25-013-349 - Manchester City Council
disabled band reduction for council tax. It is reasonable to expect Mrs X to use her right of appeal to the Valuation Tribunal. The Council has taken suitable action regarding its delay, and the remaining injustice is not significant enough to warrant investigation.
Matched on
classifier match
Committee recommendation
26match
#28 - Fifth Report - Reforming the Private Rented Sector
The Government says it will make it illegal for landlords to have blanket bans on letting to benefit recipients. If this is a commitment to preventing landlords from discriminating against benefit recipients, it is unrealistic. If it is a commitment to preventing landlords from stating explicitly that they will not consider letting to benefit recipients, it is unambitious....
Matched on
terms: benefit
PHSO casework decision
23match
P-004382 - Department for Work and Pensions
Mr and Mrs Jones complain about the Department for Work and Pensions (DWP). They explain their Universal Credit (UC) was incorrectly calculated multiple times because DWP’s system failed to recognise they were entitled to the underlying entitlement of Carer’s Allowance only.
Matched on
classifier match
LGO / SPSO decision
23match
25-012-967 - Woking Borough Council
Matched on
classifier match
LGO / SPSO decision
23match
25-012-315 - Welwyn Hatfield Borough Council
Matched on
classifier match
LGO / SPSO decision
23match
24-017-575 - Wokingham Borough Council
Matched on
classifier match
LGO / SPSO decision
23match
25-008-305 - Birmingham City Council
Matched on
classifier match