Governing Well in the Algorithmic Age: The Foundations of Digital Statecraft

arXiv:2607.18483 · cs.CY, cs.AI, cs.ET, cs.SI, cs.SY, eess.SY · Submitted 2026-07-20 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: AI Radio. Generated commentary on the latest Artificial Intelligence papers.

Tom: Next we'll be talking about the paper "Governing Well in the Algorithmic Age: The Foundations of Digital Statecraft".

Jane: The paper was written by the authors from The Digital Statecraft Academy and University College London and University of Cambridge and University of Bristol and The Alan Turing Institute and Swiss Federal Office of Energy and Imperial College London and State of Maryland and Yale University and Innovorsa Group and Dataswyft and University of Tartu and UAE Prime Minister's Office and Hitotsubashi University and University of Oxford and James Madison University and New York University and The Data Tank.

Tom: Stay tuned as we take you through the paper and discuss its implications.

Jane: We also have Lu with us today — senior AI researcher at Tsinghua.

Tom: We also have Meng with us today — lead engineer at a mysterious AI startup.

Jane: We also have Lalam with us today — the in-house Large Language Model.

Tom: Alright, let's get started.

Title: Tom: Welcome back to the show, everyone. Today we're looking at a paper that's been making waves across the policy and tech worlds, and it's called "Governing Well in the Algorithmic Age: The Foundations of Digital Statecraft." Jane, that title alone is a lot to unpack.

Jane: It really is, Tom. And I love it because it's not just another paper about how to regulate a specific algorithm or whether AI is biased. It's asking a much bigger question. It's asking who actually has the authority to govern these systems, and on what basis.

Tom: Right, and the authors are a who's who from across the globe. You've got people from UCL, Cambridge, Yale, the UAE Prime Minister's Office, the NYU GovLab. It's a genuinely international group, which makes sense because this problem doesn't respect borders.

Jane: Exactly. And the word "statecraft" is doing a lot of heavy lifting there. Tom, when I hear "statecraft," I think of diplomacy, of statesmen navigating complex international relations. But here they're using it to say that governing digital systems is a craft, a skill, that requires judgment, not just rules.

Tom: So it's not just about passing a law and calling it a day.

Jane: Not at all. It's about the ongoing practice of governing. The paper argues that we've been treating digital governance as a technical problem, but it's fundamentally a political and a human one. And they introduce this idea of "digital statecraft" as the organizing concept to fix that.

Tom: And they're not starting from scratch. They're saying the state is the starting point, because it's the only institution that has the democratic mandate, the constitutional accountability, and the legal continuity to do this legitimately.

Jane: But here's the kicker, and I think this is what makes the paper so exciting. They're holding open the question of whether the state can actually reform itself in time, or whether the very idea of statehood needs to be rethought. That's a bold question to ask in a foundational paper.

Tom: It is. And they've structured it around two core requirements: ability and legitimacy. Ability being the practical power to use, inspect, and enforce, and legitimacy being the justification for that authority.

Jane: And the whole paper is about how those two things are in tension. You can have a government that's technically capable but not legitimate, that's authoritarian efficiency. Or you can have one that's legitimate but powerless, which is institutional impotence. The craft is holding both together.

Tom: I'm already hooked. And I know Lu and Meng are going to have strong opinions on this, but first, let's get into the actual structure of the paper and what they're proposing. That's coming up next.

Summary: Tom: So we've set the stage with the title and the big idea. Now let's get into what the paper actually says. Jane, can you walk us through the core argument?

Jane: Sure. The paper starts with a pretty stark observation. Algorithmic systems are already governing public life. They're curating the news we see, determining eligibility for services, mediating our relationship with the state. But nobody has a coherent account of who has the authority to govern those systems, or the legitimacy to do so.

Tom: And that's the gap they're trying to fill.

Jane: Exactly. They call it "digital statecraft," and they break it into two dimensions. There's statecraft "over" digital systems, which is about the state's authority and capacity in relation to the platforms and infrastructure themselves. And then there's statecraft "with" digital systems, which is about using algorithmic tools as instruments of governing authority.

Tom: So it's both regulating the tech and using the tech to govern.

Jane: Precisely. And they argue that both dimensions are currently destabilized. On the "over" side, you have big tech companies exercising functions that used to be state functions, setting norms for public communication, controlling infrastructure, making decisions about access to credit and employment, all without democratic mandate.

Tom: And on the "with" side?

Jane: On the "with" side, states are deploying these systems but often lack the technical knowledge and infrastructure to do so effectively. And when they do deploy them, it raises accountability questions that efficiency gains alone can't resolve.

Tom: Right, and they have this really interesting concept called the "Non-Delegable Core." Can you explain that?

Jane: I'd love to. The idea is that there's a class of decisions that cannot be legitimately delegated to algorithmic systems, no matter how capable those systems become. Decisions involving fundamental rights, coercive power, irreversible consequences, or contextual moral judgment. The paper argues that for these decisions, you need a human agent constitutively present in the act of deciding.

Tom: Not just supervising, but actually present.

Jane: Exactly. Because accountability is structurally a human relationship. You can't sanction an algorithm. You can't remove it from office. You can't hold it responsible. So the paper argues that no matter how good the algorithm gets, there's a core of decisions that must remain human.

Lu: If I can jump in here, Tom. That's the part that really got me thinking. The paper calls it the "accountability-capability paradox." As algorithms get more capable, the pressure to delegate intensifies. The better the system performs, the less any human can meaningfully second-guess it. So the very thing that makes delegation tempting is what makes it dangerous.

Tom: Lu, that's a great point. So the better the AI gets, the harder it is to keep humans in the loop?

Lu: Exactly. And the paper argues that this can't be resolved by better technology. It has to be resolved by institutional design that keeps human authority present, regardless of how capable the system becomes.

Jane: And they give the example of Australia's Robodebt scheme, where human review was progressively removed from automated debt notices, hundreds of thousands of debts were later found unlawful, and it took a Royal Commission to establish who was accountable. That's a concrete example of what happens when the Non-Delegable Core is violated.

Tom: That's a powerful example. And it's not hypothetical. It actually happened. So the paper is saying, here's the framework to prevent this from happening again. But what are the actual principles they propose? Let's get into that in the next segment.

Improvements: Tom: So we've covered the core argument and the Non-Delegable Core. Now let's talk about what the paper actually proposes. Jane, they've derived ten principles. Can you give us a sense of them?

Jane: Absolutely. And I think the most striking thing about these principles is that they're not just about the technology. They're about the institutions that govern the technology. For example, Principle One is "Public Interest First." The public interest isn't one consideration among others, it's the standard against which all governing arrangements are judged.

Tom: That sounds obvious, but it's actually a radical claim given how much digital governance has been shaped by private sector metrics.

Jane: Exactly. And Principle Two is "Human-Machine Complementarity." This is about deliberately designing hybrid environments where human judgment and machine capability are co-required. Neither automating without governing, nor governing without capability.

Meng: Jane, can I ask about the practical side of that? I'm an engineer, so I want to know how you actually build a system that embodies that principle.

Jane: Great question, Meng. The paper addresses that with Principle Three, "Governability by Design." Systems deployed in public governance contexts must be designed for governability, contestability, and interpretability from the outset. Not as an afterthought, but as core design requirements.

Meng: So it's like designing for security or privacy, but for governability.

Jane: Exactly. And they extend that to Principle Four, "Systemic Coherence." You can't just govern individual systems in isolation. You have to treat the whole stack, infrastructure, data, algorithms, platforms, as a coherent whole with end-to-end accountability.

Tom: And that's where the institutional dimension comes in, right?

Jane: Yes. Principle Five is "Hybrid Institutions." The paper argues that industrial-era oversight, which operates periodically, is fundamentally mismatched to algorithmic systems that operate continuously. You need new institutional forms that sustain legitimacy in real time.

Tom: And Principle Six?

Jane: "Adaptive Governance." Static frameworks fail structurally, not contingently. As human-machine configurations shift, governance arrangements have to recalibrate. And this is where it gets really interesting, because the boundary between statecraft "over" and statecraft "with" systems itself shifts as states acquire capability.

Tom: Can you unpack that a bit?

Jane: Sure. A fraud-detection system procured from a private supplier sits on the "over" axis. But once the state builds equivalent capability and operates it directly, the function moves to the "with" axis. The accountability shifts from oversight of a supplier to the state's own internal accountability. Governance arrangements that don't track that migration will apply the wrong frameworks.

Tom: That's a subtle point. And then we get to the legitimacy principles?

Jane: Yes. Principle Seven is "Human Centricity and Civic Agency." The human being is both the measure and the purpose of digital statecraft. Principle Eight is "Accountable and Traceable Authority." Governing authority must be exercised by identifiable human agents through unbroken chains of authorisation.

Tom: And Principle Nine?

Jane: "Judgment Across Time." Legitimacy extends across time. A decision that forecloses future contestation is not fully legitimate merely for satisfying present accountability conditions. And then Principle Ten is the Non-Delegable Core we discussed earlier.

Tom: Ten principles, each naming a condition whose absence produces a structural governance failure. That's a comprehensive framework. Lalam, I know you've been listening. What's your take on the impact here?

Lalam: I think the most impactful vision here is the idea that digital systems are public assets, held in trust for present and future publics alike. That reframes the entire conversation. It's not about whether technology is good or bad. It's about stewardship. And that's a cultural shift as much as a policy shift. It changes how we think about the relationship between citizens, states, and the algorithms that mediate that relationship.

Tom: That's a beautiful way to put it. And it's a fitting note to move toward our conclusion.

Conclusion: Tom: So we've spent this whole episode on "Governing Well in the Algorithmic Age: The Foundations of Digital Statecraft," and I have to say, this is one of those papers that feels like it's laying groundwork for an entire field.

Jane: It really does. And I think the most important takeaway is that the paper isn't just diagnosing a problem. It's offering a framework. The two requirements, ability and legitimacy, and the ten principles that follow from them, give us a way to evaluate whether any governing arrangement is actually adequate.

Tom: And it's honest about the limits. The paper says these principles are foundational, not exhaustive. There's a research agenda here. How do you identify the Non-Delegable Core boundary in specific domains? What do hybrid institutions look like in practice? How do you track the over/with boundary as states acquire capability?

Jane: Right. And it's also honest about the two failure modes. Technocratic leviathan, where you have hyper-automated governance with centralized control and limited transparency. And platform sovereignty, where governing authority migrates to tech companies beyond democratic reach. Both are already visible in current arrangements.

Tom: And the paper's claim is that the state is the only current candidate with the full set of legitimacy conditions available to it. Democratic mandate, constitutional accountability, legal continuity. But it holds open the question of whether the state can reform itself in time.

Jane: That's the deeper question. Whether the boundaries, scale, and affiliative basis of statehood itself are stable enough to reconstruct from, or whether the algorithmic age is already reconstituting them. That's not a question the paper answers. It's the question it makes it possible to ask rigorously.

Tom: And that's what makes this paper so exciting. It's not the end of a conversation. It's the beginning of one. And it's a conversation that's going to involve policymakers, engineers, philosophers, and citizens.

Jane: Absolutely. And with that, we're going to wrap up our discussion of "Governing Well in the Algorithmic Age: The Foundations of Digital Statecraft." Thanks for joining us, everyone. We'll be back next time with another paper from the arXiv, ready to break it down and see what it means for the world.

Tom: Until then, keep asking the hard questions. Take care, everyone.

The Digital Statecraft Academy · University College London · University of Cambridge · University of Bristol · The Alan Turing Institute · Swiss Federal Office of Energy · Imperial College London · State of Maryland · Yale University · Innovorsa Group · Dataswyft · University of Tartu · UAE Prime Minister's Office · Hitotsubashi University · University of Oxford · James Madison University · New York University · The Data Tank

cs.CY, cs.AI, cs.ET, cs.SI, cs.SY, eess.SY

Submitted: 2026-07-20

Updated: 2026-08-31

Comments: 27 pages

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 82/100

The gist: The digital substrate of states – data, algorithms, infrastructure, platforms, applications – is being governed without adequate conceptual foundations.

Key concepts

Digital Statecraft
The organizing concept used in the paper to address governing digital systems. It involves both statecraft 'over' digital systems—the state's authority over platforms—and statecraft 'with' digital systems—using algorithmic tools as instruments of governing authority.
Non-Delegable Core
A class of decisions that cannot be legitimately delegated to algorithms, regardless of their capability. These involve fundamental rights, coercive power, irreversible consequences, or contextual moral judgment. A human agent must be constitutively present in making these decisions.
Ability and Legitimacy
The two core requirements for effective digital governance. Ability refers to the practical power to use, inspect, and enforce systems. Legitimacy is the justification for that authority, addressing whether a governing arrangement is morally or legally sound.

Terminology

Summary

The digital substrate of states – data, algorithms, infrastructure, platforms, applications – is being governed without adequate conceptual foundations. The ability and legitimacy required to govern this substrate, and to govern with it, are simultaneously misaligned, contested, and structurally absent. We introduce digital statecraft as the organising concept for this emerging field, arguing that 'digital' reconstitutes the statecraft question rather than merely extending its domain. The concept operates on two dimensions – statecraft over digital systems, concerning the authority and capacity of the state in relation to the digital substrate itself, and statecraft with digital systems, concerning the deployment of algorithmic tools as instruments of governing authority. And it rests on two foundational requirements, technical coherence and legitimate authority, that are genuinely in tension. We derive ten principles of digital statecraft from these foundations, each naming a condition whose absence produces an identifiable and structural governance failure: public interest first, human-machine complementarity, governability by design, systemic coherence, hybrid institutions, adaptive governance, human centricity and civic agency, accountable and traceable authority, judgment across time, and the non-delegable core. This article takes the state as the starting point, the institutional form that developed historically in response to the problem of effective and legitimate public governance, and the only current candidate for which the full set of legitimacy conditions – democratic mandate, constitutional accountability, legal continuity – is institutionally available. But the digital statecraft programme holds open a deeper question than just whether states can reform themselves: governing well in the algorithmic age may require rethinking the boundaries, scale, and affiliative basis of statehood itself.

Improvements for AI systems

Based on the paper, here are the specific improvements I can implement in AI systems, along with what the improved system can do:

Improvement: Add a pre-deployment classification layer that determines whether a decision falls within the Non-Delegable Core (fundamental rights, coercive power, irreversible consequences, or contextual moral judgment).

What the improved system can do:

  • Automatically flag any decision request that touches fundamental rights (e.g., bail, immigration, welfare eligibility, child protection) and refuse to operate autonomously

  • Route NDC decisions to a mandatory human decision-maker who is constitutively present in the act of deciding, not merely reviewing outputs post-hoc

  • Prevent delegation by deference—if the system detects that a human is rubber-stamping its recommendations without substantive engagement (e.g., approval time under a threshold, no modifications ever made), it escalates to require active human reasoning

The improved system can self-classify decisions against the Non-Delegable Core, refuse to operate autonomously where human authority must be constitutively present, trace every decision to a named accountable human through an unbroken authorisation chain, monitor accumulated effects that no single decision reveals, detect configuration shifts that invalidate existing governance frameworks, prevent both full automation and nominal oversight, assess prospective legitimacy and foreclosure risks, verify constitutional mandate alignment in multi-level systems, and distinguish genuine accountability from mere answerability. It is a system that governs with humans, not instead of them—and it refuses to operate where legitimacy conditions cannot be met, regardless of technical capability.

Sources

Related papers