The Brits are showing how to govern in a complex world. We should learn.

Last week Public Digital (a consultancy founded by the founders of the Government Digital Service in the UK) published a paper called The Radical How. To me it represents the most consequential thinking in government today, and I want to explain why.

My work today is around "state capacity"-- the ability of government (government broadly, not US states in particular) to achieve its policy goals (or lack thereof). The approach Tom Loosemore and Andrew Greenway describe here finally gets past the helpful but wildly insufficient notion that practices like agile development and user centered design can get us closer to government services that work for people in the digital age. We moved into a fundamentally complex world, in which it's impossible to know in advance the outcome of any given intervention (see Megan Stevenson's powerful essay about how social science research shows this), but we're still operating as if we live in a merely complicated world. Even the most brilliant technocratic minds struggle to make a real difference when they rely on mechanistic thinking.

Recoding America agrees: "Other policy prescriptions have backfired because they are deterministic, assuming that if you do X, you can get Y. If you measure administrative burden, it will go down. If you mandate interoperable architectures, agencies will be able to share data. But government is a vast, interconnected, complex, adaptive system made up of countless subsystems that connect in ways that are rarely fully seen or understood. Doing X often doesn’t result in Y, because we fail to anticipate the ways each subsystem adapts to changes in the others and creates incentives we didn’t intend. Deterministic policy solutions invite unintended consequences."

Some principles in The Radical How are going to feel familiar to folks. People are going to say "yes, we are working on those things.” But if you look at examples like, say, the Universal Credit program, which had been failing miserably for three painful years when it finally got a hard reset, you'll see a way of working we need to set our sights on. First some background:

UC was the biggest reform of the UK benefits system since 1948, combining six working age benefits into one and comprehensively overhauling the technological, operational and policy underpinnings of welfare provision...Yet after three years, the programme had spent ￡425 million, gone through five Senior Responsible Owners, and had not delivered a working service to a single claimant.

(An aside: Can you imagine the US Congress deciding to reform our social safety net in such a dramatic way? For that matter, can you imagine the executive branch asking them to? Perhaps our failure of imagination is understandable, given our political dysfunction, but if we can’t think it, we certainly can’t build it.)

At the time of the reset, there were 1500 people working on UC. They restarted with a new team, just 15 people, the majority of whom were new to the effort. Those who carried over from the previous team all agreed that no assumptions or legacy thinking could be carried forward if the reset was to succeed. The structure wasn't carried forward either.

Rather than separating the policy, technology and operational delivery functions (which were even based in different parts of the country), the new team brought those disciplines together in one, co-located team.

And they needed new marching orders:

A crucial moment for the new team was the Secretary of State saying to them: “I want you to deliver an intervention that means we support more people to find more work, more of the time, while protecting those who can’t work.” Note the difference between this and: “I want you to deliver Universal Credit.” The Minister set a clear outcome for the team to achieve, not a named policy for them to deliver.

If you can't imagine a cabinet secretary saying that to an agency delivery team here in the US, there are probably several reasons why, and we'll talk about them later. But for now, try to get your head around the fact that the team was not trying to deliver a pre-determined policy. They were creating the policy and the delivery at the same time. I talk in Recoding America about the value of consulting delivery teams when policy (and even law) is being written, but those examples, while they seemed radical to most readers here, are mere glimmers of the approach we truly need. The UC team starts to show what it really needs to look like. They take a test-and-learn approach not just to the delivery of the service (as many teams in the US now are starting to do), but to the policy itself.

Universal Credit was initially tested as a complete end-to-end service with a pilot group of just 100 claimants in a single postcode area in Sutton, south London. This area was carefully selected to test whether the assumptions the team has made about the core proposition were correct. Doing so revealed unanticipated challenges within weeks, such as how payments information was displayed to claimants, or what the definition of a ‘couple’ was (a semantic point that had meaningful policy consequences). Over time, the team tested the core proposition with larger groups. Next, the team started to test assumptions about how that proposition could be scaled nationally. Rather than a linear process of writing policy at the start and living with the consequences, the team started with outcomes and adjusted the policy, design and operational choices for the service iteratively, as they learned more.

I encourage you to stop here and read that again.?

This description cries out for more specifics, and there are a few in Tom Loosemore's talk from FWD50 last November. He tells of a moment when the team had shown the minister responsible for the service a video of some user research they'd done with a couple enrolled in UC.?

And in the course of that user research, the video made it very, very clear that the policy assumptions about how universal credit should operate when both members of a couple are in Universal Credit was wrong. And the minister fixed it right there. He said, “Okay, I recognize that’s wrong”… Not a line of code had been written, not a piece of operational training had happened. They just tested and learned really, really, really quickly.?

And they changed policy on the fly, not just once, but routinely, as a matter of course, as if that were their actual jobs – to make it work for real people in the real world. If a team in the US is doing that, for real, I'd like to hear about it.

It's harder here for a number of reasons I don't think I totally understand yet, but part of it has to do with the structure of UK government, in which elected lawmakers also have responsibility for what we here would call administrative agencies. (I should know this stuff much better than I do, but thankfully gov.uk has an excellent, very clear explanation of the role of ministers: "Ministers are chosen by the Prime Minister from the members of the House of Commons and House of Lords. They are responsible for the actions, successes and failures of their departments.") Here in the US, there is a lot of anxiety and debate around what policy can be set at what level and in what branch of government, including a pending Supreme Court case that threatens to strike down (or modify) the doctrine of Chevron deference. Chevron deference rests on the recognition that agencies have to deal with the gaps and ambiguities in the laws they implement. So long as their interpretations of those laws are reasonable, courts should defer to them. (The reason it’s called Chevron deference is because the principle was announced in a 1984 case involving the oil company.)

We don’t know where the Supreme Court will land on Chevron, but it’s unlikely to be as simple as a rollback or no rollback (and it's highly unlikely to be no rollback). What is almost certain, though, is that the result will be far greater uncertainty (as if there weren't already plenty) about how new laws can be implemented by agencies. We may also see new litigation over whether earlier implementation decisions made when Chevron was still on the books can now be challenged in court. If Congress hasn’t specifically and clearly delegated a decision of interpretation to an agency (and that delegation has been assumed for decades), any number of stakeholders have an opening to sue the administration on the grounds of overreach. There's a lot to be said about this, and I encourage you to read my Niskanen colleague Nick Bagley’s recent article on the chaos that could ensue in the aftermath, but the two things I take away from viewing The Radical How in the light of a pending Chevron decision are 1) we are structurally less suited to follow best practices than the Brits now, and further constraining our ability to do so in the future, and 2) we are simultaneously turning the dial of uncertainty to eleven, which makes these practices even more needed.

The great thing about uncertainty is that no one really knows what's going to happen, and with the right mindset, we can have real agency as things play out. For that reason, I can feel a bit overwhelmed by the whole thing without feeling hopeless. In many ways, we need kicks in the butt to remind ourselves what we already know in our hearts but we haven't let go of in our hands — that the old ways of working don't work. The changes we need are bigger and deeper than those we've made, not incrementally, but radically. The paper I mentioned above by Megan Stevenson is an unexpected but perfect companion to this report, stripping away our ability to defend the tweaks around the edges that dominate our efforts at change.?

Of course, any change to Chevron deference will lay bare the misguided notion that policy is somehow separate and distinct from its implementation, a topic that is near and dear to my heart, and the subject of my book, Recoding America. Whatever happens with Chevron, it will force a conversation about this that’s long overdue. If it forces the same realizations that the team behind The Radical How describe, we will have a chance at meaningful progress.?

Note: I've also published this on my substack, Eating Policy. I'd love for you to subscribe, and I'll be putting more of my writing there over time. I know not everyone wants to be on substack, and this post is really important to me, so I'm cross posting it here on LinkedIn to share as widely as possible.