{"id":"ff60f691-043e-4707-a5a8-c09a4ff95d1c","arxiv_id":"2607.18483","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Digital statecraft: a framework claiming legitimate governance in the algorithmic age requires states to hold technical ability and legitimate authority together, bounded by a Non-Delegable Core of decisions no algorithm may make.","lead":"This paper proposes a new conceptual framework, 'digital statecraft', for how governments should govern and use digital systems, built on two requirements (technical ability and legitimate authority) and ten derived principles. A smart generalist would read it because it tries to unify the scattered debate on AI regulation, platform power, and digital sovereignty into one coherent way of thinking about governance.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The Non-Delegable Core is stipulated, not derived: Section 3.3 never shows why accountability requires a human 'constitutively present in the act of deciding' rather than a human-authorised, human-reviewable chain. Principle 10 therefore rests on an extra premise.","rationale":"The reader identified the weakest assumption as the claim that authorisation is structurally a human relationship and that a chain terminating in an algorithmic system has not been traced to anyone. I agree this is where the NDC is most fragile, but I would sharpen it: even granting that authorisation must terminate in humans, the paper has not shown that those humans must be present in the act of deciding rather than in the authorising and reviewing chain. The four Section 3.1 conditions are compatible with an algorithm making a decision under human institutional authorisation, provided there is meaningful accountability, traceability, consent, and revisability. Section 3.3 introduces 'constitutive presence' as a fifth requirement without deriving it. This is a significant gap because the NDC is the paper's most distinctive contribution; if it is stipulated rather than derived, Principle 10 is a normative preference about institutional design rather than a structural limit. That said, the paper is an explicitly foundational working preprint, it anticipates and answers three objections, and it honestly defers contextual specification of the NDC boundary. The concern does not destroy the framework; it means the central claim is currently supported by an under-argued premise, which warrants the existing CONDITIONAL verdict rather than ACCEPT or REJECT. I therefore recommend no change to the reader's verdict.","tokens_in":24021,"tokens_out":4550,"duration_ms":50399,"concrete_test":"Analytical test: formalize Section 3.1's four conditions as predicates over decision procedures and check whether an algorithmic decision-with-human-appeal procedure satisfies them. Concretely, take a high-stakes case (e.g., welfare eligibility denial) where a rule-based algorithm produces the initial decision, with mandatory random audit, automatic suspension of adverse determinations, and an administrative tribunal with power to reverse. Define accountability as an identifiable human agent subject to sanction/removal, traceability as a valid authorisation chain ending in an elected body, revisability as effective human appeal, and consent as genuine citizen capacity to contest. If all four predicates are satisfiable, Section 3.3's 'constitutive presence' requirement is not derivable. A complementary empirical test: run a quasi-experiment comparing citizen/official accountability rating","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim is the Non-Delegable Core: a class of decisions whose legitimacy requires human authority constitutively present in the act of deciding, not merely supervising after the fact. The load-bearing step is Section 3.3's move from the four Section 3.1 conditions—accountability, traceable authority, consent, revisability—to this stronger requirement. That move is not derived; it is asserted. Accountability requires an identifiable agent who can be sanctioned or removed; such an agent exists in a chain that runs legislature → minister → algorithm, provided the minister is genuinely answerable. Traceable authority requires an unbroken chain to a democratic mandate; a chain that includes algorithmic classification but terminates in an elected or appointed official is unbroken. Revisability can be satisfied by effective administrative appeal to a human tribunal. The paper's rebuttal to the 'framework authorship' objection shows only that ex ante rule-making cannot be the whole story for under-specified decisions; it does not show why ex post human review cannot bear accountability. The accountability-capability paradox (capable systems invite deference) is an empirical pressure, not a structural impossibility. Robodebt is evidence that removing human review is catastrophic, but not that constitutive presence is logically required. Thus the NDC's hard boundary collapses into a demanding but contingent design requirement unless the extra premise is defended.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This conceptual paper introduces 'digital statecraft' as an organising concept for the governance of and through digital systems. It proposes two foundational requirements — technical coherence (ability) and legitimate authority — and argues that these are in structural tension. From these foundations it derives ten principles, culminating in the Non-Delegable Core (NDC): a class of governing decisions that cannot be legitimately delegated to algorithmic systems regardless of technical capability. The paper also introduces an over/with distinction (statecraft over the digital substrate versus statecraft with algorithmic instruments), develops systems-level and institutional-level ability conditions, and defends a state-centric starting point while holding open the question of whether statehood itself is being reconstituted. The article is explicitly positioned as foundational and agenda-setting, with six research lines left for the field.","tokens_in":24236,"tokens_out":3478,"duration_ms":36326,"significance":"If the NDC claim is correct, the paper supplies a principled boundary for which government decisions cannot legitimately be automated, and the over/with framework plus the vertical/horizontal traceability distinction provide useful diagnostic tools for accountability failures. The manuscript is unusually careful in its hedging: footnotes 4 and 6 concede real limits, and Section 4 candidly states that the framework does not answer whether states can rebuild the required conditions. Its systems-level ability table (Table 2) and the distinction between decision-level and aggregate NDC are genuine contributions. The central limitation is that the paper's most distinctive claim — the NDC as a structural, not contingent, limit — rests on an extra philosophical premise that is asserted rather than defended. The 'derivation' of ten principles is also better described as analytic articulation from chosen definitions than as logical derivation. These are fixable within the manuscript's scope, but they are load-bearing.","major_comments":[{"comment":"The NDC's hard limit rests on an unargued premise. The Section 3.1 conditions — accountability, traceable authority, consent, revisability — do not by themselves entail that a human agent must be 'constitutively present in the act of deciding.' A chain that runs legislature → minister → algorithm satisfies traceable authority if the minister is genuinely answerable, and revisability can be met by effective ex post review before a human tribunal. The rebuttal to the 'framework authorship' objection shows only that ex ante rule-making is insufficient for underspecified decisions; it does not show why ex post human review cannot bear the required accountability. Robodebt demonstrates the catastrophic consequences of removing human review, but it does not demonstrate a structural impossibility of post-hoc accountability. Consequently, Principle 10 reduces to a demanding but contingent design","section":"Section 3.3 and Principle 10"},{"comment":"The paper repeatedly says the ten principles are 'derived' from the two foundational requirements, but the derivation step is never shown. Most principles (e.g., Public Interest First, Human Centricity and Civic Agency) are normative assertions introduced with arguments from public-value and democratic theory; they do not follow from technical coherence and legitimate authority alone. The stated test — 'each naming a condition whose absence produces an identifiable and structural governance failure' — is not operationalised, so it cannot distinguish derivation from taxonomy. The authors should either make the derivation explicit or reframe the contribution as an analytic framework with normative commitments, rather than a derivation.","section":"Sections 1 and 4, Table 4"},{"comment":"The state-centric premise is load-bearing but asserted: 'The state is the only candidate for which the full set of legitimacy conditions is institutionally available.' No argument is given for why democratic mandate, constitutional accountability, and legal continuity are jointly available only to states, nor for why the state's deficits are all 'contingent — buildable, not inherent.' The paper's own concession that Global South states face 'structural' constraints on infrastructure dependency and regulatory capacity complicates this distinction. If the claim is empirical, evidence is needed; if it is normative, the paper should say so and justify the priority. As written, the conclusion that digital statecraft must begin from the state is stronger than the supporting text warrants.","section":"Section 4"},{"comment":"The NDC is introduced as a claim about 'a class of decisions' but is then split into decision-level and aggregate levels. The aggregate NDC is a different kind of claim: it concerns patterns across decisions, not the act of deciding. This is a useful extension, but the relationship between the two levels is not fully specified. In particular, if aggregate NDC is the same 'underlying test' applied at population level, the paper should explain why an absent monitor violates the Section 3.1 accountability condition even when no individual decision does. As written, the aggregate NDC appears to be an additional normative requirement rather than a consequence of the same derivation.","section":"Section 3.3 vs. Section 3.1"}],"minor_comments":[{"comment":"The phrase 'authorisation is structurally a human relationship (Hart, 1961)' is a strong interpretive claim. Hart's concept of law concerns rule-governed legal systems, not the claim that authorisation must terminate in a human actor. Either provide a direct quotation or qualify the attribution.","section":"Note 6 / Section 3.3"},{"comment":"The example of a fraud-detection system migrating from the over-axis (procured) to the with-axis (state-operated) is initially clear, but the with-axis example conflates internal accountability structures with the state's relationship to its own systems. The over/with boundary deserves a crisper definitional statement to avoid the appearance of post-hoc classification.","section":"Section 2.3"},{"comment":"The NDC is cited to Engin & Hand (2025b, Zenodo) and the ten principles to the DSA Founding Manifesto (Engin et al., 2025a). Since the NDC is the article's most distinctive claim, the argument should be self-contained; relying on a non-peer-reviewed preprint and a manifesto is risky for a journal submission.","section":"Section 3.3 and Principle 10"},{"comment":"The text says the principles are 'derived from this tension,' while Section 4 says 'the derivation offered here is foundational rather than exhaustive.' These two framings should be harmonised, especially if the authors decide to move away from derivation language.","section":"Section 4 / Table 4"},{"comment":"Minor copy issues: 'fro m' appears in the running header text; 'for the derivation' in Section 4 should be 'the derivation'; the phrase 'diagnostic and severally correctable' in Section 4 is unclear. These are presentation issues only.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is an ambitious, well-written agenda-setting piece, but its central NDC claim is imported from a Zenodo preprint and a DSA manifesto, and the key step — from accountability conditions to constitutive human presence — is asserted rather than demonstrated. The derivation language oversells what is actually a normative framework with analytical structure. For the journal, I would consider whether this is best framed as a research article or as a programmatic essay; if the former, the derivation needs to be made explicit or softened. The unusually heavy self-citation to non-peer-reviewed outputs should be addressed in revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague — quick take on arXiv:2607.18483. This is a conceptual working paper that does something genuinely useful: it gives the scattered AI-governance and digital-sovereignty literatures a common vocabulary. The over/with distinction (governing the substrate vs governing with it), the ability/legitimacy pairing, aggregate vs decision-level NDC, and vertical vs horizontal traceability are all real additions. The paper is also unusually honest — it flags in footnotes and Section 4 that the NDC boundary, hybrid institutions, and the deeper statehood question are left open. That candor is earned.\n\nThe main soft spot is the one you'd expect: the Non-Delegable Core is stipulated, not derived. Section 3.3 simply asserts that accountability requires a human 'constitutively present in the act of deciding,' rather than a chain that terminates in an accountable, removable official who genuinely reviews and can overrule the algorithm. The stress-test note is right that ex post review, if substantive, can satisfy accountability. The paper's responses — framework authorship is insufficient, recursive AI oversight just extends the chain — don't close the gap. The accountability-capability paradox is a real empirical pressure, but it isn't a structural impossibility. So Principle 10 reduces to a strong design recommendation unless the extra premise is defended. That matters, because the NDC is advertised as the paper's most distinctive claim.\n\nTwo smaller issues. The abstract says the principles are 'derived'; what the paper actually does is articulate them from chosen definitions, which is fine but should be labeled as such. And the empirical generalizations — e.g., that the state is the only actor with the full legitimacy set — are asserted, not argued. The paper itself concedes some of this.\n\nNone of this kills the framework. The over/with architecture and the ability/legitimacy tension survive the NDC critique, and the paper is worth engaging seriously. I'd be happy to referee it — it would benefit from a careful revision that either defends the NDC premise or explicitly downgrades it to a contingent requirement. It's the kind of paper a reading group could chew on for a while.","headline":"A useful organizing framework for AI governance, with a central hard limit that is asserted rather than proven.","tokens_in":24957,"tokens_out":2212,"would_cite":true,"duration_ms":21813,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Legitimate government in the algorithmic age has a structural limit: a class of decisions that cannot be delegated to AI no matter how capable it becomes.","keywords":["digital statecraft","legitimacy","accountable authority","non-delegable core","algorithmic governance","human-machine complementarity","systemic coherence","public interest"],"falsifier":"A functioning jurisdiction in which a fully automated decision affecting fundamental rights (e.g., welfare eligibility or parole) has operated for years with unbroken, demonstrable accountability—traceable to a named human authority who accepts responsibility, is subject to sanction, and maintains public confidence—would refute the Non-Delegable Core. Under the paper's account, such a case should not exist.","tokens_in":23743,"feed_emoji":"⚖️","tokens_out":4783,"duration_ms":40981,"temperature":0.7,"pith_summary":"This paper introduces digital statecraft as the missing conceptual foundation for governing the digital substrate of states—data, algorithms, infrastructure, and platforms—and for governing with those systems. Its central claim is that legitimate governing authority requires two conditions at once: the technical ability to act, and a legitimacy that can justify and account for that action. Those requirements, the paper argues, are in genuine tension, and that tension has a structural limit: the Non-Delegable Core, a class of decisions involving fundamental rights, coercive power, irreversible consequences, or contextual moral judgment that cannot be delegated to algorithmic systems regardless of their capability. From this foundation the paper derives ten principles of digital statecraft, each naming a condition whose absence produces an identifiable governance failure.","feed_headline":"A hard limit on AI in government: the Non-Delegable Core","feed_subtitle":"Decisions touching rights, coercion, or irreversible harm require a human in the act of deciding, not just in the loop.","key_machinery":"The central object is the Non-Delegable Core (NDC), a class of governing decisions—those involving fundamental rights, coercive power, irreversible consequences, or contextual moral judgment—whose legitimacy depends on human authority being constitutively present in the act of deciding. The NDC is paired with the accountability-capability paradox: the more capable an algorithmic system becomes, the harder it is for humans to second-guess it, which makes the structural requirement of human presence more, not less, necessary. The machinery also includes the over/with distinction (governing over digital systems versus governing with them) and the two-level structure of the NDC (decision-level a","core_discovery":"The paper claims that 'digital' does not merely extend the domain of statecraft but reconstitutes it: the machinery of government is being rebuilt around algorithmic systems, non-state actors, and distributed authority. On its account, the state must be able to act on the digital substrate (statecraft over systems) and through it (statecraft with systems), and both dimensions require technical coherence and legitimate authority. The most distinctive claim is that the tension between these requirements has a limit—the Non-Delegable Core—where accountability, traceable authorisation, and judgment across time converge on a requirement that technical capability cannot satisfy: a human agent must","pith_inferences":["The NDC argument, if sound, generalizes beyond states: any entity exercising quasi-public governing authority—including platform companies that shape public discourse—would be bound by the aggregate NDC's requirement of an accountable monitor for systemic effects.","A testable implication of the accountability-capability paradox is that as AI models improve, formal human-in-the-loop arrangements will increasingly become nominal rather than substantive; this can be measured by tracking whether human reviewers override system recommendations less over time and whether their reviews are ever acted upon.","The framework implies that 'governability by design' should be treated as a procurement requirement, with contractual penalties for ungovernable systems—an operational extension the paper gestures at but does not develop.","A pilot comparison could test the theory directly: jurisdictions that adopt constitutionally-present human authority for NDC decisions versus those relying on post-hoc review, measuring accountability outcomes such as successful appeals, sanctions of responsible officials, and public trust."],"forward_implications":["If the Non-Delegable Core holds, jurisdictions that automate welfare, bail, immigration, or child protection decisions without a human agent constitutively present in the act of deciding are structurally illegitimate, regardless of how accurate the underlying systems become.","The over/with framework supplies a diagnostic: governance failures can be located as failures of ability or failures of legitimacy, and the traceability condition demands both horizontal (cross-border) and vertical (devolved-tier) accountability.","Institutions must become hybrid—operating at algorithmic tempo with machine monitoring and human judgment—rather than periodic oversight bodies, and adaptive to shifting human-machine configurations.","Digital sovereignty is reframed as a legitimacy condition: enforceable accountability over infrastructure, not ownership or territorial control, since traceable authority breaks when states depend on infrastructure they cannot hold to account.","The ten principles together name two concrete failure modes: technocratic leviathan (ability without legitimacy) and platform sovereignty (governing authority migrating to companies without democratic mandate)."],"fun_headline_variants":["Statecraft's hard limit: the Non-Delegable Core","AI in government hits the Non-Delegable Core","The Non-Delegable Core: why AI can't decide alone","Governing with AI: the Non-Delegable Core constraint","Digital statecraft's non-negotiable human judgment"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The entire Non-Delegable Core rests on the philosophical premise that authorisation and accountability are structurally human relationships—that a chain of authority ending in a machine cannot be traced to anyone—and on the empirical claim that only the state currently holds the full set of legitimacy conditions.","fun_headline_variants_meta":{"raw":{"variants":["Statecraft's hard limit: the Non-Delegable Core","AI in government hits the Non-Delegable Core","The Non-Delegable Core: why AI can't decide alone","Governing with AI: the Non-Delegable Core constraint","Digital statecraft's non-negotiable human judgment"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000188,"raw_usage":{"total_tokens":1201,"prompt_tokens":809,"completion_tokens":392,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":553,"completion_tokens_details":{"reasoning_tokens":323}},"tokens_in":553,"tokens_out":392,"duration_ms":3467,"temperature":1.0,"reasoning_tokens":323,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T15:16:49.885217+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A functioning jurisdiction in which a fully automated decision affecting fundamental rights (e.g., welfare eligibility or parole) has operated for years with unbroken, demonstrable accountability—traceable to a named human authority who accepts responsibility, is subject to sanction, and maintains public confidence—would refute the Non-Delegable Core. Under the paper's account, such a case should not exist.","supporting_citations":[],"review_version":1}