Transparency, welfare and digital public services
As this report has demonstrated, understanding how Universal Credit functions is no simple task. It requires large amounts of personal data to operate, there are regular changes to the design and functioning of the system, the digital account is personalised, and there is a growing dependence on automation. Data about how the policy is performing is published in an ad hoc way and is not reliably machine readable.
By publishing more information about how Universal Credit works, opening up the development process and rethinking how it publishes data and statistics, DWP could make Universal Credit into an exemplar of transparency for a digital public service. This could provide leadership across the public sector, but would also highlight the very real efforts DWP has made to learn from the challenges of digitising a complex policy.
Understanding public services that change over time
The transparency of a digital service may seem like a peripheral concern relative to many of the issues thrown up by Universal Credit, but it is not. The digitisation of public services means that the choices of service designers and software developers have become critical in how the public access services they have a legal right to. As immigration solicitor, Jonathan Kingham, notes in this article on the digitisation of Brexit era immigration systems in the UK:
Unlike with Rules and legislation changes, there was little opportunity to scrutinise the detail of what are in fact significant changes to the immigration system prior to their coming into force (bar selective ‘user testing’, which is rarely transparent to all). And, as with so many tech developments, as the process or ‘app’ itself increasingly takes centre stage over the content (in this case the law) that underpins it, there are risks.1
Just as the personalisation of commercial social media platforms has led to a dilution of the shared public experience of current affairs, the opacity and mutability of digital services means there is a risk of dilution of the shared public understanding of how public services function. There is also a risk of information asymmetry, where the government has a fine-grained, real-time view of how the public are using a service, but those wishing to hold the government to account for its actions operate largely in the dark.
These risks should also be seen against growing public concern about how data is used and growing concern across the technology sector about the implications of automated decision making and design. The 2018 Digital Attitudes Report, by the think-tank Doteveryone, for example, found that people feel disempowered by the lack of transparency in digital services and an appetite to understand better how data about them is used.2 If there is going to be long-term public support and effective scrutiny of Universal Credit, DWP needs to become much more transparent about how its systems works and how they change, and government and civil society organisations need to think differently about how they audit digital public services. What follows are some approaches that DWP should consider.
Publishing data about how the system changes
One simple change the department could make would be to maintain a set of public registers that describe the functioning of the Universal Credit ecosystem. This would go some way to answer the criticism by the United Nations Special Rapporteur that there is no list of the constituent parts of the system, and help researchers, auditors and the public understand the different components that make up Universal Credit and how they change over time. These public registers should include (but not be limited to):
- a register of the different systems used to deliver and administer Universal Credit
- a unique identifier for each deployment along with a date-time stamp and short release notes explaining any changes
- the data models used in each system along with a description of each data point
In addition to the registers listed above, DWP should maintain a design archive to record significant changes to the design of the user interface of Universal Credit. This would enable debate and understanding of how the system is developing over time. Discussions on websites like AdvisorNet about changes to the user interface of Universal Credit, and DWPs consultations on service design changes, suggest there would be an audience for such an archive.
Rather than publishing these artefacts in an ad hoc and inconsistent way, the aim should be for this data to published in a structured format, and to do so automatically as a byproduct of the development process. This would ensure currency while keeping costs low. Given such information does not fit within one of the existing GOV.UK publishing formats, DWP should work with the GDS to find an appropriate way to publish it. Organisations with an auditing or review function, such as the NAO and the Institute for Government should also start to review and comment on the implications of user interface choices made in digital public services.
Opening up the development process
There are also changes that DWP could make to its development process to make it more transparent. The NAO reports documenting the Universal Credit development process clearly show that DWP has adopted an agile methodology that allows them to develop policy changes, prioritise them, test options with end-users, and then verify the impact once live. DWP should publish as much of this information as possible.
The department should also open source as much of the source code as possible. While there may be legitimate reasons for some of the code not being open (for example fraud detection systems), today, even the code used for the core Universal Credit calculation remains closed. The GOV.UK website points people looking to understand how it works to external ‘benefit calculators’ run by third sector organisations.3 4. As well as having an impact for those wanting to claim, the fact the calculation is closed makes it harder for campaigners, think-tanks and political parties to model changes to the welfare system.
Data and statistics
In addition to data from the development process, DWP needs to change its approach to publishing data about the performance of the policy. There is dissonance between this agile “test and learn” approach and the openness of DWP in relation to Universal Credit. Despite clearly having a modern and capable analytics function within the team, the approach to publishing data remains in the age of quarterly “Management Information” reporting, and discovering what data is published remains difficult. DWP should move to a model of detailed, consistent and real-time publication of data using the GOV.UK performance platform.5 It should also work with GDS to make those datasets indexed and discoverable on GOV.UK.6
The datasets that DWP publishes should, in part, be led by the needs of third sector organisations who have an interest in improving how Universal Credit works. However, a few suggestions for real-time data DWP should explore are:
- the types of to-do items set for claimants, along with their performance
- amount of work coach time and its effectiveness
- sanctions and their impact
- the time taken to complete an application, along with the number of failed applications
- analysis of the payment cycles of employed claimants (e.g. how many of those claiming and in employment are paid monthly)
If DWP does not start publishing this information, support organisations and campaign groups should start to collate these datasets independently.
The working-age welfare system is critical to British society, and its digitisation raises questions that transcend the political debate around welfare reform. DWP should aim to make accountability, understandability and trust core design principles of the Universal Credit service, and work to understand the needs of organisations who might hold them to account. It has shown the effectiveness of developing an in-house digital capability in responding to the delivery of a complex policy. It should follow this by showing how openness and transparency can improve the quality of a public service and increase trust in those who manage it.
Jonathan Kingham, “Computer says no: facing up to the full implications of a digitised immigration system”, Free Movement, 8th January 2019, https://www.freemovement.org.uk/computer-says-no-digitised-immigration-system/ ↩
Doteveryone, “People, Power and Technology - The 2018 Digital Attitudes Report”, 22nd February 2018, https://doteveryone.org.uk/report/digital-attitudes/ ↩
These organisations have had to reverse engineer their calculators based on policy documents and legislation ↩
“Performance - GOV.UK”, GOV.UK, https://www.gov.uk/performance ↩
This should at the very least be done in the GOV.UK data model, but GDS should explore linking to this information from the “start page” of services ↩