Royal Society report points to possible uses in targeting interventions, incident response, legal processes and healthcare – and highlights importance of open data
Machine learning could have application in most areas of the public sector, according to the authors of a new report on the technology published by the Royal Society.
Their testing of public attitudes shows there are no ‘no go areas’, and that there are potential uses in areas as diverse as identifying people at risk and legal processes.
Titled Machine learning: the power and promise of computers that learn by example, the report is the latest example of the rising interest in how the technology could revolutionise many operations in the public and private sectors.
Speaking at the launch event the leader of the working group that prepared the report, Professor Peter Donnelly, said it has been given an increase in momentum over the past couple of years by an improvement in the algorithms used, the increasing processing power of computers, and an increase in the availability of open data.
He also said it is distinct from artificial intelligence in being focused on specific tasks in its use of algorithms, and does not go as far as AI in matching some human skills such as the ability to make judgements and contextualise information.
The report points to the potential of machine learning in several areas of public services, including some that have so far have not been widely associated with the technology.
Preventing NEETs
One is targeting interventions of ‘at risk’ groups, such as in identifying young people who may be close to dropping out of education or failing to find employment. Machine learning could be used to analyse data from their school records and related sources and create models that predict the likelihood of them becoming NEETs (not in education, employment or learning).
This could prompt schools to provide them with extra support: the report says it has already been trialled by Essex County Council.
It could also be used to improve responses to incidents such as flooding, using data from first responders, earth observations and social media. There is often limited time available to analyse the large quantities of data involved, and machine learning could make it possible to develop models for anticipating how the incidents could develop and how to direct resources.
It could also be used in optimising energy infrastructure – designing systems that respond more effectively to peaks in demand – and predicting compliance with the law. For example, there have been efforts by tax authorities to learn from patterns of transactions which companies or individuals are trying to evade paying their tax.
The report also follows others in highlighting the potential of machine learning in healthcare, where it could support doctors in providing more accurate diagnoses.
Compilation of the report involved a survey of public attitudes, which the authors said revealed overall support for the use of the technology in public services. Despite widespread unfamiliarity with the term and reservations around specific applications, people did not indicate that there were any areas in which it should not be used.
Difficult tasks
Professor Nick Jennings, vice provost (research) at Imperial College London, said: “The issue is how we can do difficult tasks more effectively, and to just rely on humans is not the most sensible approach. For example, when we want to help target resources better.
“The relationship between humans and a computer programme is really important. Some tasks can be automated entirely, but we see it very much as a partnership in which machine learning can help to make suggestions and work alongside people as an automated assistant.
“We’re not talking about a wholesale removal of human judgement from these things.”
The report recommends that, while the UK has made good progress in the field, the Government should sustain its efforts to support it by making re-usable data available in new wave of “open data for machine learning”. Where datasets are unsuitable for general release, there is a need for progress with policy frameworks and agreements to make them available for specific users under clear legal constraints.
It also says the Government has a key role to play in the creation of new open standards, for example for metadata, and in embedding the importance of machine learning in education. One thing it could do is to ensure the key concepts are taught to everybody, taking this into account in the next review of the national curriculum.
Skills priority
One of the priorities is to equip students with the skills to work with the relevant systems across professional disciplines, and universities will have to ensure it is included in a wide range of courses. Professional bodies should also contribute to this effort.
“We are at an early stage of the development of this technology,” Donnelly said. “Now is the time to think about how we as a society want it to develop, and how the relationships will unfold.”
He added: “Machine learning will have an increasing impact on our lives and lifestyles over the next 5-10 years. There is much work to be done so that we take advantage of the potential and ensure the benefits are shared, especially as this could be a key area of opportunity for the UK.”
Image from detail of report cover, from the Royal Society
Are you interested in learning more about the potential for machine learning and artificial intelligence in public services? UKAuthority is staging a conference, Rise of the Bots, in London on Tuesday 20 June looking at possible applications in the short and long term, and the surrounding organisational and ethical issues. More details and registration here.
