Algorithms & accountabilities: an interview with Jonathan Slater, former UK education chief

Last year, the Department for Education’s use of an algorithm to decide exam results generated huge public anger, resulting in the departure of permanent secretary Jonathan Slater. He tells Matt Ross how events unfolded – and highlights the gaps in accountabilities and infrastructure exposed during the pandemic
In 2001, Jonathan Slater left his job as director of education at the London Borough of Islington to join the civil service. At the time, he recalls, he “thought the move from local to central government would be essentially a matter of scale.” He would still be an “impartial official working for politicians,” he believed – just “operating on a bigger stage.”
On arriving at the Cabinet Office, though, he quickly realised that “it was a much bigger change than that.” In essence, he says, “when I was a council employee, I worked for the council. But when I was a civil servant, I worked for the minister. And that’s an obvious point, but it’s absolutely huge.”
As an example, he recalls a time when falling pupil numbers required Islington to shut a primary school – always a controversial decision. Slater gave his advice on which school should close at a public council meeting; and when it was accepted, he and the council’s chair of education visited the school to set out their reasoning. “I explained to the teachers, parents and pupils why I’d recommended the school’s closure, and the chair of education explained why he’d made that decision,” he says. “Our roles were clear, and they were different.”
Providing and explaining his advice in public, Slater felt intensely accountable at Islington. By contrast, within central government the “model of ministerial accountability, in which the civil servant’s job in public is to explain the minister’s policy – not to explain their own advice – is a first-order principle in the unwritten constitution.” Under Whitehall’s model, civil service advice is private and secretaries of state are – at least in theory – fully responsible for all of their department’s activities. But as Slater would discover, it doesn’t always work quite like that.
Coming home
After joining the Cabinet Office, Slater rose through a series of performance management and organisational reform roles in the ministries of justice and defence. In 2016, he was appointed permanent secretary of the Department for Education (DfE) – the job he enjoyed most in government. “There’s something about education which is inspiring,” he says. “My father was a teacher; my mother was a children’s social worker. I don’t think it’s a coincidence that I ended up as the head of the department that oversees both of those professions.”

The department seemed to find him equally inspiring. Staff engagement rose strongly over the following years, and he worked hard to improve diversity. “When I arrived there were almost no black or Asian senior civil servants,” he says. “When I left, we’d hit our interim target of 10%.” Meanwhile, the introduction of new collaboration and communications tools altered DfE’s culture and working methods.
“Going online provided all sorts of opportunities for improving the connectedness of the organisation, in ways that I’d never have believed,” says Slater. The workforce was “becoming less hierarchical and more empowered, and the traditional divide between the policymaker and the operational delivery person was breaking down.” Reforms such as moving the weekly senior leaders’ meeting online rendered people’s physical locations irrelevant, fostering “more democratic and equal” decision-making – and would leave the department well prepared for a year of remote working.
The need for networks
The DfE did, however, have a problem. Ever since 2010, the focus had been on turning schools into academies: taking them out of local authority control, and giving them autonomy under the DfE’s supervision. But this task was far from complete, and the department lacked effective ways to provide support, coordination and services to all of the country’s schools. “The school system is in the process of being nationalised, but it hasn’t been completely,” he explains. “What I was struck by when I arrived in the department was that a lot of the infrastructure that you would see in some other systems – whether it was a national or a local system – was not really in place.” The department had been focusing on creating academies, he adds, rather than on “putting in place the infrastructure to manage such a system.”
The government was aware of the challenge, and planned to ‘academise’ all schools by 2020 – creating a single, nationwide system to ease central oversight and support. But in the face of backbench opposition, those plans were abandoned shortly after Slater’s arrival. And he had to tread carefully in his attempts to strengthen DfE’s nationwide management systems: ministers were, “perfectly legitimately, keen to make sure that the school system was still a bottom-up one, and were nervous about infrastructure being put in place above the level of the individual school.”
Nonetheless, Slater started to put those systems in place. He created two new commercial director posts: one for the department, and one for school procurement. And he began reshaping the network of regional commissioners – established to manage the creation of new academies – into “the regional outlets for the Department for Education.” Building a national and regional infrastructure was “a big, big task – and it was a task that I set myself to work on,” he adds. “We made a lot of progress before the pandemic arose.”
Ill-equipped into the pandemic
When lockdown came in March 2020, however, DfE lacked the national systems and networks to rapidly organise key services such as computer equipment, free school meals and support for online teaching. Nor did it have a pandemic response plan: “There’s the question as to whether such a thing should have been there,” says Slater. “But it was not.”
As a result, fundamental questions – such as which department should manage the provision of free school meals – were being made at great haste, and then delivered through an incomplete national infrastructure. Given that task, DfE commissioned a current supplier to deliver meal vouchers, using an existing buying framework: the choice at that point, Slater comments, was “either to introduce the thing three months late – in which case you’ve missed your chance – or bring it in and there are teething troubles.”

It’s clear that Slater regrets that DfE – like departments across the western world – lacked a pandemic response plan. “Ideally, you’d have made those choices earlier,” he says. “That would be the point of planning for a pandemic.” But with neither a prepared strategy nor a mature infrastructure, officials did what they could: “If you rely upon a few civil servants from the DfE buying computers as quickly as they can from China, they do as well as any group of civil servants can in the circumstances,” he comments. “And those civil servants worked incredibly hard. But that’s not enough: you need a proper infrastructure in a situation like this.”
So the department struggled in the pandemic’s first months. Then came disaster: an algorithm, designed to replace cancelled exams by deciding pupils’ A-level grades, produced results that sparked huge controversy – prompting both the policy’s collapse, and Slater’s exit from the department.
Enter the algorithm
It wasn’t simply the use of an algorithm that caused the problem: algorithms had been shaping pupils’ grades for a decade. “Back in 2010, the government had been very concerned about ‘grade inflation’, and wanted to make sure that essentially the same proportion of children got each grade each year,” Slater explains. An algorithm was introduced to ensure that, even if raw exam scores rose substantially, the overall proportion of children awarded high grades could only grow fractionally each year.
“Ministers found it hard to believe that the increasing proportion of kids doing well in GCSE and A-levels could actually represent what was going on,” he comments. “They thought that this must be grade inflation, and they wanted to put a stop to it.” So for a decade England has used a system of “comparative outcomes”, in which a particular exam score is no guarantee of a certain grade – though that was, as Slater notes, “invisible to the public at large”.
When the pandemic led to exams’ cancellation, ministers faced a difficult decision. “You had two choices,” Slater says. “You could either use an algorithm, or you could go with teacher assessment unmodified by an algorithm.” Given strong evidence that teachers tend to be over-optimistic about their pupils’ likely attainment, it was clear that using unmodified assessments would lead to “a big increase in exam results: an enormous amount of grade inflation.”
“Ministers debated the pros and cons of those two options, but were concerned to avoid grade inflation – as they have been since 2010,” he continues. “It’s been a big part of their reform agenda, and they didn’t want to throw it away. And I thought that was a legitimate choice.”
How did it work?
“It’s more complicated than this,” Slater cautions – but in essence, teachers were asked to rank the children in their classes, then the algorithm allocated them grades to fit that school, class and teachers’ previous results. Grades were also adjusted to reflect the year group’s performance in schools’ regular ‘SAT’ tests. “Ministers were entitled to say that they want to avoid grade inflation. And as a mathematician, it seemed to me that it should be possible to design an algorithm which would achieve that objective,” he comments. “Whether you would be able to persuade the public of it was the question.”
They could not. When the results came out, the media found dozens of high-flying students who’d been marked down to keep grades in line with their classes’ previous results – a problem that hit bright kids in poorly-performing schools the worst. And the algorithm could only be used where class sizes were large enough to generate sufficient data, so pupils in smaller classes were awarded the more generous teacher assessments – but this favoured private school students. As a result private schools had a bumper year for results, while sixth-form colleges flatlined amidst a welter of hard luck stories.
There was no easy answer here: had DfE relied solely on teacher assessments, notes Slater, “there was no question but that the middle class kids would have benefited disproportionately. There’s plenty of evidence that that’s what happens.” But the algorithm was also generating “surprising numbers of what were thought of as anomalies, where the difference was not between an A and a B, but an A and a D. And that seemed very much more concerning.” The decisions of leaders in Wales and Scotland to abandon their own algorithms “put more and more pressure on the DfE to do likewise,” he recalls – and, just four days after results day, education secretary Gavin Williamson announced that the government would instead set grades using unadjusted teacher assessments.
Allocating accountability
Within days, both Slater and Sally Collier – the chief executive of exams regulator Ofqual, which had developed the algorithm – had left their roles, with PM Boris Johnson saying that there was “a need for fresh official leadership”. But Williamson – who, under the principle of individual ministerial responsibility, should have carried the can – kept his job, causing outrage among constitutional experts and staff unions.
FDA union general secretary Dave Penman accused Johnson of throwing civil service leaders “under a bus without a moment’s hesitation to shield ministers from any kind of accountability”. At think tank the Institute for Government, director Bronwen Maddox called for Williamson to quit: “Unless there are consequences for ministers of the decisions that are their responsibility, the UK’s principles of democratic accountability will become meaningless,” she wrote.
Slater won’t comment directly: “I’m one of the interested parties, aren’t I? So I’m not objective.” He’d been nearing the end of his five-year term, he adds, and had planned to “go non-executive this year anyway”. He’s already taken a visiting professor at Kings College London and started pro bono work with Lambeth’s BAME mental health charity Black Thrive, and expects to work with local authorities on developing their children’s services.
“Nobody forces you to apply to be a permanent secretary; and the nature of working in the civil service is that you’re working for politicians,” he adds. “In the end, the relationship with the secretary of state is one that can determine whether you stay or go; I chose to do the job, and you take a bit of a risk in those circumstances. And for myself, the rewards of being the permanent secretary for four and a half years were worth it.”
He does, however, note that “you could contemplate a different system, in which civil servants were more accountable for their advice – as exists more in local government.” Altering the system of accountabilities in central government would involve a “big-P political choice”, he acknowledges, and change the power relationship between ministers and Parliament – but it’s clear that he’d like to see changes. “I’ve seen at first hand the benefits of a system in which advice is given more transparently: it improves the accountability of the person giving it,” he says carefully. “You can tell that I think there are some merits in it.”

He also raises questions over the system of accountabilities around Ofqual, a non-ministerial department. Although the exams regulator reports directly to Parliament rather than the education secretary, it produced the algorithm on Williamson’s instruction. “The idea that Ofqual and its board could be solely accountable for something which is essentially a policy choice,” he says, illustrates the confused accountabilities within the system. “That’s the constitutional arrangement, and doesn’t seem to be sustainable,” he comments. “Ministers, in the end, do need to be accountable for a system like that.”
No way back
Ultimately, opting for teacher assessments did indeed result in steep grade inflation – by some 52%, in the case of those awarded A or A* at A-level. “And ministers have said that they don’t want this year’s intake to be disadvantaged, so there’ll be grade inflation again this year,” Slater comments. “Are they going to take it away the following year? It’s going to be a brave secretary of state that’s going to take that down again.”
In fact, Slater has big questions over the future of England’s exam system. “Does it make sense to have exams for people at the age of 16 when they carry on at school [to age 18]?” he asks. “Does it make sense to have A-levels? Why not an assessment of their readiness to do a degree?”
For him, there are two clear priorities over the coming years: addressing the ‘attainment gap’ between rich and poor students, and improving young people’s wellbeing and mental health. “The so-called attainment gap had been declining over recent years, but had stopped declining before the pandemic and was flat-lining,” he comments. “Now it’s going right up again” – with evidence showing that “all of the gains made in the previous few years have been wiped out by the pandemic.”
On mental health, the international PISA survey “showed that the wellbeing of teenagers in England was among the lowest in the whole of the OECD,” he says. “It was right at the bottom – and this was before the pandemic.” Both these crucial indicators were already going in the wrong direction, he adds – and during 2020 “both have got a lot worse. So whatever comes out of this pandemic needs to address those two challenges head-on.”
And Slater’s former department should, he believes, have one further priority. “The Coalition government deliberately abolished the [DfE’s] national infrastructure, because they wanted to create more bottom-up innovation. And that’s a perfectly legitimate choice – but it doesn’t work in a pandemic,” he argues. For the good times as well as the bad, he says, DfE needs a stronger national infrastructure: “Even in ‘steady state’, if you’ve got a good way of teaching maths, why wouldn’t you want to make sure that everybody’s got access to it?”
“The things that we were starting to put in place weren’t capable of withstanding the pandemic,” he comments; and while “you need these systems more in a crisis than you do in steady state, you always need something.” That infrastructure could be established at local, regional or national level: “You probably need a combination of all three – but you have to have it, and that’s crucial. It has to be designed,” he concludes. “And given what we’ve learned over the last year, I expect it will be.”
Global Government Forum: five thoughts for better government
To help our readers get the best out of Global Government Forum, we ask interviewees five standard questions – four seeking practical advice and opinions, and one to reveal something a little more personal. This is an edited version of Jonathan Slater’s answers.
Can you name one lesson or idea from abroad that’s helped you or your colleagues?
“I spoke to my German, Danish, French and Republic of Ireland counterparts to get their perspectives before recommending to ministers that primary school children should come back to school in half class sizes after the May half-term. Essentially, we followed the examples of Germany and Denmark, adjusted for the English school system.”
Are there any projects or innovations from the UK that might be valuable to your peers overseas?
“I introduced the world’s first social impact bond when I was at the Ministry of Justice – using social finance to fund work by local charities to rehabilitate offenders leaving Peterborough prison. It worked: reoffending rates fell more at the prison than elsewhere. I took a call from the White House, asking me how we did it.”
How can we improve the ways that senior public officials work with and learn from their peers overseas?
“I would have loved to have had the opportunity to work in a team of people from different nations on a social issue of international concern.”
What are the biggest challenges in your field in the next few years?
“Enabling people to learn the skills required for a world of automation and ‘net zero’, in ways which tackle inbuilt social inequalities. Giving much greater priority to promoting wellbeing (happiness).”
What’s your favourite book?
“As a child: Kes, by Barry Hines. In recent years, The Nix by Nathan Hill.”