A report from The Behavioural Insights Team (BIT) has counselled the government that data has “great potential to be used in public policy”.
BIT – a former Downing Street policy unit which, in 2014, was spun out into a limited company jointly owned by its employees, the Cabinet Office, and innovation charity Nesta – has published the first report created by its data-science team, which was established in January. Since then, it has run eight “exemplar” projects, covering four ambitions for improving public services: better-targeted inspections; improved quality of randomised controlled trials (RCT); better decision-making; and the use of predicting the impact of traffic accidents.
For its work on targeted inspections, BIT used data-science tools and techniques to identify the schools it considered to be most at risk of underperforming. Of the schools rated by Ofsted as “requires improvement” or “inadequate”, 65% fall within the riskiest 10% as identified by BIT. The riskiest 20% included 85% of schools in Ofsted’s two lower categories.
BIT also claimed that, “using publicly available data published by the Care Quality Commission and other sources,” it has also developed a model that would enable 95% of failing GP practices to be identified by inspectors by visiting just 20%.
In the case of RCT improvement, BIT crunched data from previous trials to divide people into two broad groups, based on observed character traits. It then ran a study in which machine learning was used to predict which of two different messages would be more likely to prompt a positive response from people in the respective groups. One group of study subjects was sent the message the algorithm had predicted would be more successful, while the other group‘s choice of message was made at random. Although the results did not show a “statistically significant” difference in the responses of either group, BIT believes it could improve the design of its model for this test.
In its bid to show how data science could occasion better decision making, BIT used “natural language processing” to assess a selection of child social work cases that had been “flagged for no further action”. The goal was to assess the text of each case and predict which of these, in fact, would result in the implementation of a child-protection plan or a child being taken into care. The cases picked as the most likely to unexpectedly require further action amounted to just 6% of the overall number of cases examined – but encompassed 45.6% of those cases that, ultimately, had resulted in additional measures.
“We are working with social workers to build a digital tool that can be used to help inform their decisions,” BIT said.
In respect of predicting serious traffic collisions, BIT worked with data from East Sussex, where traffic accidents “have bucked a national trend for fewer incidents of killed and seriously injured (KSI) casualties”. The data scientists’ analysis revealed that “behavioural factors, and not road conditions” were the biggest predictor of death or serious injury.
“We have been able to bust some myths – for example, about older drivers, and goods vehicles,” BIT added. “Motorcyclists, the young, and people in early middle age are disproportionately more likely to be involved in KSI incidents in East Sussex.”
In the BIT report’s foreword, civil service chief executive John Manzoni said: “There is great potential for government to improve the performance and productivity of services through the smarter use of data. This data includes outcomes, use patterns, costs, and citizen experiences. With this wealth of data, we have an obligation to make government services the best they can be. This means learning from where services are working well, and improving where they are not.”
He added: “Across government we are already transforming the way in which citizens and the state engage, through the expansion of digital technology. The next step is to ensure that the data gained is constantly driving improvement. For that, the application of data science will be key.”
BIT employs around 100 people across offices in London, Manchester, New York, Singapore, and Sydney. During its time within Whitehall, it was popularly known as “the nudge unit”, as it promoted the implementation of the principles of the nudge theory of political and behavioural science.