Machine Learning report says the government needs to recognise the value of public sector data and properly balance this against concerns about privacy and access

Machine learning relies on having access to large datasets – Photo credit: Flickr, Robert Scoble, CC BY 2.0

The government has a crucial role to play in making sure the data it holds is accessible and easy to use, but needs to be ready to manage significant datasets strategically, the Royal Society has said.

In its report, Machine Learning, the society sets out the potential offered by the emerging field and how the government should support research and adoption of the technology.

Machine learning – a strand of artificial intelligence that trains computer systems to find patterns in data in a way that eventually allows them to make their own decisions without human input – relies on having access to large datasets, such as the ones that government holds.

Related content

Google’s DeepMind and NHS restart data-sharing deal with greater transparency
DH trials algorithm tool to analyse consultation responses in effort to handle ‘click democracy’
Civil service chief John Manzoni urges change as data revolution ‘comes for government’

The Royal Society said that the UK had a “strong history of leadership” in machine learning, and that the wealth of accessible public sector data – such as the government’s set of authoritative lists, or registers – had positioned it as a leader in the field.

An increase in the use of the technology could have a positive economic effect, the society said, and urged the government to develop and support a machine learning sector by developing open standards for data use and enhancing the availability and usability of public sector data.

“Access to public sector data could catalyse a range of economic activity: the direct value of public sector information to the UK economy has been estimated at £1.8bn, with wider social and economic benefits from this totalling up to £6.8bn,” the report stated.

The society also said that the government needed to “recognise the value of some public sector data” and consider how it manages the data it holds to make the most of this potential.

“As machine learning becomes a more significant force, the ability to access data becomes more important, and those with access can attain a ‘first mover feedback’ advantage that can be significant,” the report said.

“When there is such value at stake, it will be increasingly necessary to manage significant datasets or data sources strategically.”

“When there is such value at stake, it will be increasingly necessary to manage significant datasets or data sources strategically.”

The society emphasised the importance of healthcare data, saying that it was one of the UK’s “key data assets”.

“NHS data is a unique information resource for the UK, and making effective use of it will be vital for ‘UK plc’,” the report said.

However, the widespread use of the UK’s healthcare data is controversial, and previous attempts to increase its use across the NHS – such as the Care.data programme that the government cancelled last year – have failed.

The Royal Society acknowledged such potential issues, saying that, given the nature of healthcare data, “there are natural and legitimate concerns about how the value of its use can be balanced against concerns about privacy, and questions about who should be able to access it”.

The report said that access to NHS data “must clearly be under very carefully regulated conditions to ensure protection of individual privacy”.

It added: “If this balancing act is resolved, and if appropriately controlled access mechanisms can be developed, then there is huge potential for NHS data to be used in ways that will both improve the functioning of the NHS and improve healthcare delivery.”

The report recommended the creation of a separate policy framework that specifies how – and by whom – such sensitive information can be accessed or used, and said that the form and function of such a framework is being developed as part of a separate piece of work with the British Academy.

However, the society said that it was “not appropriate to set up governance structures for machine learning per se”, and that specific questions about machine learning should be handled in a sector-specific way, rather than by creating an overarching framework for all uses of machine learning.

‘Public awareness of machine learning is low’ 

The society also assessed the government’s role in increasing the public’s understanding of machine learning as the technology becomes more ubiquitous, and commissioned a public opinion survey on the topic from Ipsos Mori.

According to the survey, most people haven’t heard of the term but many are aware of its applications, with attitudes varying depending on how machine learning was being used.

The survey also found that little awareness of the potential uses of large datasets, including how it could be used to improve public services, for example.

Support for the use of personal data tended to be conditional, with people being more concerned about their data being used for commercial purposes and preferring it if it was used for something that results in tangible public benefits, for instance in healthcare, transport or crime prevention.

The survey also showed doubts about whether a computer would be able to make a better decision than humans and ambiguity about the level of control and automation that can, or should, be given to a computer.

There were also concerns about techniques that cluster individuals or the use of correlations between datasets that initially appear unrelated, the survey said.

The society said that there should be “continued engagement” between researchers working on machine learning and the public, with funding made available for large scale programmes of public engagement activities.