Using machine learning technology and techniques such as graph theory to identify businesses fraudulently try to exploit Government funding support schemes

About 7.5% of the £4.9 billion distributed to businesses in pandemic ‘Bounce-Back Loans’ was obtained fraudulently, according to BEIS figures. Training Machine Learning systems with data on known fraudsters, and honing them with ‘graph theory’ techniques, we could trawl through through Companies House and HMRC datasets to spot behaviours and relationships that may suggest fraudulent activity. This would much improve the targeting of audits and investigations, strengthening both enforcement and deterrence work.

Gathering data to present a more up-to-date picture of emergency accommodation availability across the UK for use in crisis situations

Through the Crown Commercial Service’s buying framework, government departments have purchased more than 4.7 million room-nights for groups including refugees, asylum seekers and those quarantining for COVID-19; an unknown number have been bought outside CCS agreements. A single spreadsheet is used to track requests and available properties, but there is much duplication and waste in the system. Meanwhile, there is no centralised system recording the accommodation available within government properties. A platform presenting real-time data on available rooms – across the government estate, and among private providers – would both cut administrative costs and help civil servants to find suitable accommodation much more quickly.

Developing a data dashboard to help identify and tackle modern slavery

Incidents of modern slavery in the UK are typically discovered by public servants working in a wide range of fields, who then make referrals to the police. We have the data to support a much more targeted approach to this crime: a dashboard combining a wide range of datasets would reveal locations and organisations with an elevated risk of modern slavery. Detection and investigation work would then focus both on areas where several risk factors combine, as well as those where gaps in the data may indicate that activity is being hidden from public authorities.

Using online gaming technology to conduct policy experiments in virtual worlds

The fast-growing availability of data has much improved our understanding of the impacts of individual policies and services – but much of this information can’t help us to improve macroeconomic policy, where interventions have complex effects reaching across society. We do, however, have a set of ready-made test beds: thousands of people participate in online games, which could be used to test out economic policies. Following experiments to understand how players’ responses may differ from their behaviour in the real world, we would work with game developers and operators to explore people’s responses to changes in areas such as pricing, inflation and subsidies – providing a unique and valuable set of data to inform macroeconomic policymaking.

Adapting government services for those with mental health issues using patient-authorised healthcare data

When individuals have mental health problems, enforcement action by public bodies can potentially worsen their condition without generating any benefits: if depression is preventing someone from filing their company accounts, for example, a fine or prosecution is likely to damage the business while deepening their depression – making it still harder for them to file their accounts. Similar issues exist around services such as Universal Credit, self-employment tax returns and Vehicle Excise Duty. Averting unhelpful enforcement action when individuals are suffering a mental health episode, an optional mental health vulnerability service would help both to minimise the severity of mental conditions, and to improve the targeting of enforcement at those in a position to respond.

Using data to connect-up cross-government security clearance processes

Every government body requires contractors to clear a baseline security check. These can take months, cost an average £115, and are valid for 15 years – but they are not usually portable, so a new one is required whenever a contractor moves between civil service organisations. Issuing a standardised, portable credential would save public money by reducing delays, administrative costs and check fees. And a similar system would produce big savings for employers and volunteers in other checking and clearance systems.

Implementing software and artificial intelligence to refine how Digital Mail Service items are categorised

HMRC receives 15 million items of customer correspondence annually. Often, their journey to the required team is not a direct one: items may sit in the wrong queue for some time before being forwarded on and be redirected many times before arriving in the right hands. However, we now have both a vast dataset of scanned items, and data on their ultimate destination – making this problem an ideal candidate for the application of Optical Character Recognition and Machine Learning technologies. Trained using historical data on the final destination of each item of correspondence, an ML algorithm would vastly improve the distribution of mail across the organisation: getting correspondence directly and rapidly to the correct team would save civil servants’ time, speed up casework, and provide a better service to the public.

Connecting datasets across government to improve levels of compliance for child maintenance payments

The Child Maintenance Service is responsible for tracing parents who try to evade their responsibilities and securing maintenance payments. But while its Searchlight system includes data on benefits recipients and the employed, it does not cover the self-employed: the CMS currently maintains a long list of untraced parents, regularly conducting searches for each of them, even while these people complete annual tax returns and report their income to HMRC. Routinely sharing information between HMRC and the CMS would reduce delays, cut administrative costs, bring down the benefits bill, and help prevent parents from evading their duty to contribute to their children’s upbringing.