Scream If You Want to Go Faster! Why Government Technology Needs (Much) Better Governance
Dominic Cummings’ techno-enthusiasm is infectious — and, this year, it’s been spreading all over government.
There doesn’t appear to be a written plan — at least not in the public domain — but there are certainly recurring themes. This is a dream of a low-friction, innovation paradise in which numbers tell the truth while bureaucrats (and ethicists) get out of the way. It is less a vision for society, more an obsession with process and power.
The risks of this approach can be seen in the handling of the A level grading and the NHSX track and trace app. Both of these projects needed better governance processes — including open and transparent ways of working, forward planning, and the acceptance of external expert guidance – but instead, both have been allowed to experience significant failure, due in part to a culture of secrecy and a lack of oversight.
The New Technocracy
The emergence of a patchwork of UK innovation initiatives over the last few months is notable. Rather than fiddling with increments of investment, there is a commitment to large-scale, world-leading innovation and enthusiasm for the potential of data.
But there is also a culture of opacity and bluster, a repeated lack of effectiveness, and a tendency to do secret deals with preferred suppliers. Taken together with the lack of a public strategy, this has led to a lot of speculation, a fair few conspiracy theories, and a great deal of concern about the social impact of collecting, keeping, and centralising data.
But it seems very possible that there is actually no big plan — conspiratorial or otherwise. In going through speeches and policy documents, I have found no vision for society —save the occasional murmur of “Levelling Up” — and plenty of evidence of a fixation with the mechanics of government.
This is a technocractic revolution, not a political one, driven by a desire to obliterate bureaucracy, centralise power, and increase improvisation.
And this obsession with process has led to a complete disregard for outcomes.
Chips with Everything
Although there is no public strategy, there is a lot of disparate news coverage of the UK government’s technical present and future. The following is not a complete survey — just an indication of a few things in flight; doubtless there are many more.
There are plans for increased data sharing across the civil service, while the creation of a British DARPA to “boost transformative research” has been approved, and the UK is now — somewhat surprisingly — co-owner of “struggling mega-constellation start-up OneWeb”.
10ds is on its way — apparently a “pseudo start-up” based in No 10, driving a “quantitative revolution … to ensure all decision making is made on the best available evidence”, and NHSX have been working with Palantir to create a “single source of truth” for the pandemic.
Many of these initiatives are underpinned by a fascination with quantitative data — which is also telegraphed across many posts on Cummings’ personal blog.
Drag-and-Drop Data for Instant Government
Cummings’ (surprisingly recent) paper on ARPA and PARC sets out a commitment to save the world with “dramatic breakthroughs that advance knowledge and civilisation” — but there is no detail on what these breakthroughs might be, or what they will achieve. Instead, there is a focus on process, and on the right kind of genius, who is — luckily — able to conjure the right kind of number.
In many of his posts, Cummings comes across as being in awe of data and those who create it. The idea that physicists swung the Vote Leave victory indicates that his head might be turned by a nice lab coat rather than a really good formula in an Excel table, and his dream of it being as easy to “insert facts, data, and models in political discussion as it is to insert emoji” 😉 speaks to a sort of consumerist, on-demand thirst for snippets, rather than a deep understanding of complexity. It’s app-informed, drag-and-drop data for instant government.
Appropriately, Cummings is also beguiled by both speed and dismantling bureaucracy — breaking out the bold letters to celebrate some off-the-cuff commissioning in the early days of ARPA. “No forms,” he writes, celebrating the great art of making it up as you go along. “No budget. No ‘roadmap’. No committee. No lawyers.” Elsewhere in the same document he is positively giddy:
“There are huge nonlinear returns from speed. If you can speed things up by a factor that seems ‘totally unrealistic’, even ‘mad’ according to those in charge of a normal bureaucracy, then you can get huge qualitative performance changes that more than make up for some bureaucratic untidiness (and you also disorientate competitors).”
I can find no positive mention of scrutiny or governance, or any recognition of the field of data ethics in his vision of data-driven government.
And the same mood is apparent in a number of recent government speeches and policy announcements.
Moonshots
On 1 July Michael Gove gave a speech about the glory of statistics that bemoaned the twin straightjackets of regulation and bureaucracy, saying,
“The whole culture of Government, and the wider world of political commentary, is hostile to risk, adventure, experimentation and novelty”
Gove’s speech devotes more words to berating his colleagues than to setting out the kind of society that a public servant should be proud to serve. It is undoubtedly a process document, dressed up with a little history — doubly dismissive of the National Audit Office and enthusiastic about “big risks, and radical experiments”.
Earlier the same day, business secretary Alok Sharma announced that:
We have a once-in-a-generation opportunity to strengthen our global position in research, unleash a new wave of innovation, enhance our national security and revitalise our international ties. We will use this opportunity to pursue ambitious new goals — the ‘moonshots’ that will define the next decade and beyond. By stretching our ambitions and engaging with and learning from people and communities all over the UK, we will create long-lasting economic and societal benefits for our country.
This slightly delirious jumble of metaphors calls on multiple mental models of heroic innovation, without saying what will actually happen. We will climb very high, but we don’t know where we’ll go.
A third document published on the same day, by the PM’s Council for Science and Technology, was more circumspect. “Principles for Science and Technology Moonshots” stresses the need for “processes, and governance structures” and asserts the importance of choosing problems well and wisely, “understanding human perceptions, trust, and behaviour” and “looking beyond the traditional ‘hard sciences’ for insights”.
The disparity of these three documents makes the extent of the hyperbole clear. There is little political regard for either complexity or specificity; goals are there to be measured against, but who sets them and the impact of achieving them is not considered — and the most common criteria for success appears to be “world-leading”, “innovative”, “novel” — or simply “fast”.
In the BEIS R&D Roadmap, the big social challenges are vaguely enumerated as “achieving net zero carbon emissions and supporting healthy ageing” and a slightly haphazard collection of projects are celebrated, but there is no view of their aggregate or long-term consequences, or of the desired mix and match of scale.
The impact of synthetic phonics (a technique for teaching children to read), for instance, is given as “between £3,300 and £8,800 per student in present discounted value based on how their future earnings will increase”, while the benefit of “next generation training infrastructure” for Defence staff is unquantified — although it is presumably much easier to measure.
In his Ditchley speech, Michael Gove asked, “What are the metrics against which improvement will be judged?”, but neither he nor any of his colleagues appear to have come up with an answer.
Project Speed
Similar themes are all discernible in the last speech I’m going to turn to: Chief Secretary to the Treasury Steve Barclay’s “New Radicals”, given at centre-right think tank Onward.
This was ostensibly about the forthcoming spending review, but it is jazzed up with some technical language to prove this isn’t any old spending-review — it’s a data-driven, iterative one.
Altogether it’s an odd speech, combining a Keanu Reeves does Data mood with the carpenter’s cautious adage to “measure twice, cut once”. Barclay is torn between towing the innovative party line and playing it safe: he sets out a vision for pacey, agile national infrastructure, inspired by how fast web pages load, and renames the Infrastructure Delivery Taskforce “Project Speed”.
Elsewhere he softens on Gove’s dismissal of the NAO and acknowledges (but does not name) the risks of depending too much on data — but before long, the hotchpotch of agile and iterating and “disparate and novel data sources” is too much. In the long middle passage, Barclay manages to advise both against changing plans and against not changing plans, before saying, “The real failure would be to only ever play it safe, and to never try anything new.” This is neither measurable nor a goal, and is a bizarre definition of a thriving society. “We’re in recession, 40,000 people have died, and thousands of teenagers have lost their place at university, but at least we tried something new!”
This speech shares the fixation with process: there is much talk of “speeding up projects, measuring outcomes, and using data”, but no mention of who might set the outcomes or what they are, and it fixates once again on ways of working, not what might be achieved — ultimately calling for:“a focus on the quality of the decision, not the efficacy of the outcome:
“it’s important we draw the distinction between the decisions we make and outcomes those decisions generate.
The quality of a decision won’t necessarily match the quality of the out-come.
In a world of imperfect information we can never perfectly predict outcomes upfront, but what we can do is improve the decisions we make.”
Some Projects Should Succeed
“It’s important we draw the distinction between the decisions we make and the outcomes those decisions generate” shows a complete disregard for consequences. Barclay seems to be dismissing whole disciplines here — including risk management, project management and plain old commonsense. And Gove, in fact, says something similar:
“We need, as a Government, to create the space for the experimental and to acknowledge we won’t always achieve perfection on Day One… some projects will misfire, some will seem promising but fall at the final hurdle, but along the way we will end up with unexpected gains”
This would be fine if Gove and Barclay had teamed up to run a Young Enterprise Initiative after school on Wednesday afternoons, but it doesn’t quite work for governing a country in a national emergency. Some things should be safe bets; some things should just work; some things should be tried and tested, and there should be forms to fill out and budgets and roadmaps and committees and lawyers.
As the A level results fiasco and Track and Trace have shown, this government is not as good with either data or technology as Cummings might have hoped. No one wants bad outcomes for thousands of teenagers or to be continually stuck at home, waiting for an app. Everyone wants the expected gains, not some random surprises.
The insistence on secrecy and lack of oversight has blighted both projects — as has the inability to define what good looks like for the public.
Success for grading this year’s A levels could have been defined at the start as “ensuring fair treatment for outliers”, and every subsequent decision mapped against that.
Success for track and trace might have focussed on effectiveness rather than process and exceptionalism; instead of pursuing a go-it-alone, centralised strategy that aimed to be “world-beating”, England could have aimed to “collaborate as broadly as possible to save lives”.
In both cases, the strategy was confused with the method of delivery.
Openness and transparency are also essential, not just for cultivating public trust, but because they make it possible to work with others and benefit from good advice.
Despite repeated requests from civil society, the Data Protection Impact Assessment for the NHSX Track and Trace app was not published for several months, and the external ethics board took a long time to materialise and was then unceremoniously dumped.
The Royal Statistical Society alerted Ofqual to statistical issues in May, and fellows were only permitted to further assist if they signed an NDA — yet Tom Haines’ excellent blog post outlines the necessity of bringing in external help to check and verify the model. The fact that the Chair of Ofqual also chairs the Centre for Data Ethics and Innovation, yet doesn’t appear to have called on his board of ethical experts for help or advice, or recognised the social and political implications of the model is also concerning.
Neither of these projects called for particularly novel or innovative forms of governance. No one needed to spin up any block chain or make a smart contract. To go back to Cummings’ earlier point about ARPA, both could have been immeasurably improved with forms, budgets, roadmaps, committees and lawyers. Being fast and secretive with no paper trail is fine for a spy in a paperback novel, but it’s not good enough for government.
Some projects deserve to succeed, but they need to be given a chance — and so do we. There’s no more time for this “scream if you want to go faster” approach to getting things done.
Just because a project has a whiff of “digital” about it, the rulebook doesn’t need to be torn up in an explosion of innovation. Goals can be set, risks managed, consequences mapped. It’s definitely not rocket science.
