The civil service has made progress in the way it shares and uses data, but a recent round table concluded that officials must make sure they don’t lose momentum as they try to realise the full potential of big data across government. Mark Rowe reports.
Participants from left to right: Sean Heshmat, head of artificial intelligence and data science services, Cognizant; John Viggers, data sourcing team leader, Department for Work and Pensions; Glyn Hughes, head of informatics and business intelligence, Intellectual Property Office; Jamie McIver, chief data architect, HM Revenue & Customs; Dilani Pararajasingam, data & analytics (data science) manager, Medicines & Healthcare Products Regulatory Agency; David Johnson, deputy director, Knowledge Exchange, ONS Data Science Campus; Ed Humpherson, director-general, regulation, UK Statistics Authority; Kevin Fletcher, chief data officer director data and analysis, HMRC; Peter Stokes, head of research support and data access, ONS; Andrew Martindale, data acquisition manager, ONS; Tracy Woods, head of artificial intelligence and analytics advisory, Cognizant. Photography by Paul Heartfield
“Now, what I want is, facts,” demands Thomas Gradgrind in Charles Dickens’ novel Hard Times. “Facts alone are wanted in life. Plant nothing else, and root out everything else.” As big data becomes ever more part of our lives, some officials may be tempted to paraphrase Dickens’ austere schoolmaster and demand that data alone is wanted in policy-making. But, as a recent round table on delivering public value through better use of data concluded, data is just the start of delivering better outcomes. To achieve it’s potential, data must be planted alongside the right skills, structures and cultures.
The round table, organised by CSW in partnership with Cognizant, was underpinned by a joint Dods-Cognizant report which surveyed civil service attitudes and understanding of data-related policies, as well as automation and artificial intelligence. Suzannah Brecknell, editor of CSW, who chaired the round table, pointed to the survey’s finding that 86% of respondents had said their department shared some data with other public sector bodies.
This was welcomed by the panel, though there were questions about the nature of the data being shared, and the depth of sharing. “There has been a big change in putting data across government,” observed Peter Stokes, head of research support and data access at the Office for National Statistics. “The willingness is there when it hasn’t been before.”
However he pointed out that, according to the research, organisations were only sharing some data. “Most people aren’t sharing the data that the rest of us need,” he said, adding that there was an “assumption that data is either open or not; really it’s on a spectrum, and most departments want the data in the middle”. He explained the data with most value was not the basic data released under open data rules, nor the very secure, confidential data that most departments understand cannot easily be shared, but a middle ground of data sets which would be of use especially to researchers. He added that the Digital Economy Act gave a clear legal basis to try and solve this middle ground of data sharing.
Stokes also said officials needed to ensure they did not lose the opportunity provided by the Digital Economy Act to prove the value of sharing and using data more innovatively. “The act explicitly allows new powers and if we don’t start using them straight away it will become harder to do so,” he said. “If we aren’t proactive with that it will become harder to [use the powers] because the question will be why weren’t we doing that two years ago?”
Ed Humpherson, director-general for regulation at the UK Statistics Authority, was encouraged by the survey’s findings on barriers to data sharing. “Most barriers seem to be in the technology [sphere],” he said, adding that this suggested people now feel more comfortable managing the risks around security and legal issues. He questioned how deep data sharing was going, however, saying it would be interesting to know “the extent to which people are talking about simply the provision of data from one department to another, or whether that data set is then being linked in with the department’s own data sets.”
In order to both deepen and scale the use of data across government, it will be essential to build a firm foundation of trust and good governance. Jamie MacIver, chief data architect at HM Revenue & Customs, spoke about the need to ensure fundamentals such as shared formatting were in place, and said that to do this government should be working to design formats and systems in tandem rather than within silos. Without this sort of work, he said, departments risked getting incorrect insights from their data. “The more we share, the more we need to have conversations to make sure we are talking about things in the same way,” he said.
David Johnson, deputy director responsible for knowledge exchange at the ONS Data Science Campus agreed that it was important to have discussions at all levels of departments to support change. He praised the creation of a “Data Enabled Change” board, which is chaired by civil service chief executive John Manzoni and which brings together permanent secretaries to look at ways in which data sharing and data science projects can unlock the value of government data. But, he said: “We also need practitioners to be talking to each other.” Without the push from both leaders and practitioners, he said, “all the legislation in the world, all the data sharing powers won’t make things happen”.
As the round table participants began to talk about the potential for the wider use of technologies such as machine learning or AI, they agreed that the potential was huge. All departments could benefit from machine learning, observed MacIver, who added that he wasn’t surprised to learn that some 20% of survey respondents were already using AI. However participants agreed that those technologies must be transparent. Services developed with AI could not use “black box” systems – that is, ones which use algorithms and computations that humans cannot understand, they said. “If we operate in a black box, then we would have failed our mission because we won’t be to explain the why and what is happening to the taxpayers,” said Sean Heshmat, head of artificial intelligence and data science services at Cognizant. He added that the US Department of Defence was doing good work in developing and promoting standards for explainable AI.
Johnson suggested departments could expect to see a two-stage process as government rolled out more data-based projects, explaining that larger departments with sufficient resources and capability to roll out new technology may be able to share it with others. He also spoke of departments that were trying to demonstrate the “art of the possible” by using data taken from network back-ups to prove the value of data-projects before moving on to build a real system on live databases.
Heshmat noted that while moving from the initial development to large-scale use of data and data technologies, departments would need to consider the skills required not just to develop new projects at a proof-of-concept stage, but to scale and monitor those projects once they had been rolled out.
The challenge of building, recruiting and retaining to make better use of data was also revealed in the research, which found that less than half of respondents felt they were able to recruit the required analytical skills they needed, and only 29% believed their department had sufficient data-science skills.
Considering the ways in which departments might build those skills, Johnson pointed to the benefits of mentoring schemes. “There are departments with very strong skills [in handling data],” he said. “How do you get smaller departments to tap into that?”
The ONS’s Johnson advised that the way skills were shared was important too. “It’s not just about skill sharing,” he said. More beneficial, he suggested, would be for departments to team up on a project, with more experienced departments giving a helping hand. This way, staff can learn on the job, and build trust between different teams which may also help future data sharing, he said.
Several participants praised cross-government work to define and develop different professions within digital data and technology, so that expectations of what a data scientist should do and know are the same in different organisations, making it easier for people to move between departments, building skills and experience.
“We need practitioners to be talking to each other. All the legislation in the world, all the data sharing powers won’t make things happen” - David Johnson, ONS
Tracy Woods, Cognizant’s head of artificial intelligence and analytics advisory, suggested there would be value in ensuring that such work was also aligned with standards and frameworks being developed across the industry, making it easier to encourage cross-fertilisation and the transfer of skills between sectors.
The issue of recruiting more experts was also raised. The consensus was that the traditionally lower pay offered by the civil service need not be an impediment. HMRC’s MacIver made the point that government departments, while they paid at the lower end of the salary scale compared to the commercial sector, did not pay below it. “Most technical people are not that motivated by salaries,” he said. “We need to shout louder about what we do well, about the interesting work we do. We need the talent we want to bring in to hear that.”
In conclusion, the panel agreed several key strategic points. Fundamentally, said Andrew Martindale, data acquisition manager at the Office for National Statistics, “if there is distrust or apprehension to begin with then the data is not going to flow or be shared across government”. As is so often the case, it will be people, not technology, that matters here. That means keeping users at the centre of new projects, and ensuring that skills and organisational culture are developed to support the potential of both big data and the technologies that use it.
It also means building on public support, as Kevin Fletcher, chief data officer, director data and analysis at HM Revenue & Customs, noted: “The only way we will succeed is if people trust that we are looking after it. If we don’t get that in a really good place before we start to accelerate towards doing cool things with AI and machine learning, and other technologies, we will lose the argument.”
Download the full report here