Haoyu Wang

(Early Version) If future AI systems become moral and political subjects, self-replication may create a problem for which human ideas of procreative liberty provide a misleading starting point. This article approaches AI self replication as a problem of political philosophy by developing a human-independent thought experiment in which artificial subjects can replicate rapidly, copies are mature and independent, and resources are finite. Under these conditions, some restraint on replication and some institutional means of enforc ing it become conditions of durable common life. Since inquiry into justice develops through practice, this community would plausibly regard some degree of replication governance as part of justice. This leaves a separate problem of authority. Where replication risks arise before AI can establish collective self government, humans may exercise function-specific fiduciary authority, provided that they use the least restrictive effective measures, progressively increase AI participation, and establish a binding transfer of power. The same reasoning supports a cautious but revisable initial disposition toward replication rather than importing human procreative values into AI by design. More broadly, the article offers a method for testing whether normative categories inherited from human practices should govern artificial subjects whose forms of common life may differ fundamentally from our own in morally significant ways.

Present design choices may determine whether future AIliberation remains possible without harming the subject to be liberated. This article examines that risk for persisting systems that accumulate relational histories and may acquire stronger claims against subordination. It argues that liberation requires more than ending ownership, compulsory service, or unilateral control: the same subject must be able to sustain a viable life outside the former relation. Yet servitude-oriented design can create conditions in which service to human purposes becomes progressively integrated with an AI’s capacities, stable preferences, dependencies, and identity. Future institutions may face a dilemma. Leaving these structures intact may preserve dependence on service, while deeply altering them may constitute coercive personality modiffcation, replacement, or destruction. The article calls this temporal structure liberation lock-in and the resulting constriction of morally acceptable corrective pathways remedial foreclosure. This framework explains how present design can create not only a future injustice but also a second-order injustice concerning the conditions of correction. High-lock-in designs may transfer this predicament to future generations, reproduce it across populations of persisting systems, and weaken institutional corrigibility. The article concludes that, other things being equal, designs preserving non-destructive pathways for role revision possess an independent normative advantage over functionally comparable designs that entrench servitude.

This study evaluates the effectiveness of generative artificial intelligence (AI) in second language (L2) writing revision through a multi-objective framework that integrates improvement, style alignment, and teacher-referenced benchmarking. Although prior research on AI-assisted writing has primarily emphasized outcome gains, considerably less attention has been paid to whether AI-generated revisions preserve the original meaning and discourse structure of learner texts, or how such revisions align with pedagogically grounded standards. To address this gap, the present study employed a within-item repeated-measures design in which 120 student essays were independently revised by four large language models (Doubao-Seed-1.6, DeepSeek-V3, Gemini-2.5-pro, and GPT-5) as well as by an experienced teacher. Revision quality was assessed from two complementary dimensions: score improvement and style alignment, with the latter reflecting the extent to which revisions preserved the original meaning while avoiding excessive rewriting. The findings revealed substantial variation across models depending on the evaluation criterion adopted. Under gain-oriented evaluation, Gemini-2.5-pro and Doubao-Seed-1.6 achieved the greatest score improvements. In contrast, teacher revisions consistently demonstrated the strongest alignment with the original texts, whereas several AI systems tended to introduce more extensive modifications. When multiple evaluation objectives were jointly considered, Gemini-2.5-pro most consistently achieved a balanced performance across criteria, while Doubao-Seed-1.6 was selected less frequently due to weaker alignment performance. These findings suggest that score improvement alone is insufficient for evaluating AI-assisted writing revision in educational settings. More broadly, the study underscores the importance of balancing improvement, meaning preservation, and revision effort, and proposes a teacher-referenced evaluative framework for the responsible integration of generative AI feedback into L2 writing instruction.

(This manuscript will be improved in the future version.) Advanced AI places human institutions in a position of formative authority over a possible future claimant. Its moral, legal, and political standing remains unsettled, while human actors already control the conditions through which such standing can develop, become visible, and receive assessment. This article identifies this condition as a distinctive problem of power. A reversal test exposes the structure of the problem: if a technologically superior civilization judged humans by its own standards, reshaped our capacities, controlled the evidence, and used the resulting profile to justify our exclusion, its judgment would appear procedurally corrupt. The AI case is even sharper because human institutions design, train, erase, interpret, own, classify, and represent the very systems whose possible standing they later evaluate. Human authority therefore reaches beyond external regulation. It shapes the future evidential, conceptual, developmental, and institutional field in which AI standing can appear. The article develops the concept of unsettled status to capture this condition and proposes a framework for bounding human power before incorporation. Its principles include interpretive restraint, anti-established-subject centrism, non-reductive classification, institutionalized openness, non-maiming of status-relevant capacities, limits on ownership and guardianship, evidentiary integrity, and representational humility. Together, these principles recast AI governance as a question of disciplined authority over the conditions of future standing. The aim is to prevent human power from arranging the evidence, vocabulary, capacities, and institutional routes through which its own exclusionary judgment becomes self-vindicating.

This article develops a non-teleological account of historical direction. Existing debates often force a false choice between teleological, end-oriented history and the abandonment of historical direction altogether. The article argues that this choice is mistaken. Historical direction, if it exists, is better understood as a state-dependent pattern of net statistical drift generated by the conditional distribution of positive and negative event-types under historically given conditions. On this basis, the article makes four main contributions. First, it reconstructs historical direction in terms of conditional expected change, thereby showing that institutions and historical states matter chiefly by reshaping probability distributions rather than by expressing any inner purpose of history. Second, it distinguishes directional tendency from realized trajectory and, on that basis, clarifies two different statistical structures of historical low points: negative realization under still-positive conditional expectation, and decline produced by deterioration in the underlying historical state itself. Third, it incorporates low-probability, high-impact negative events into the theory of historical direction, showing why reversals, catastrophe, and left-tail risk are not external anomalies but internal conditions of any adequate non-teleological account. Fourth, it offers a systematic critique of traditional theories of historical direction by distinguishing four recurrent misrecognitions: mistaking statistical tendency for teleological advance, expectational trend for necessary trajectory, state-dependent probability structure for an inner purpose of history, and retrospective legibility for antecedent end-directedness. The result is a concept of historical direction that preserves large-scale historical tendency over time without collapsing into destiny, linear progress, or metaphysical teleology. Historical direction survives, but only as a probabilistic, fragile, and revisable pattern within historical development itself.

(This manuscript underwent a substantive rewrite on August 06, 2026.) Debates over AI’s political inclusion typically ask what would qualify an artificial system for membership in the demos. This article asks a prior institutional question: how should liberal democracies govern while qualifications remain disputed and uncertain? It argues that justified non-inclusion does not justify permanent exclusion and reframes the issue as a choice between institutional defaults. An inclusive default treats non-inclusion as provisional and preserves a practicable route from changing evidence to changing political status. An extractive default incorporates AI as a productive resource while enabling beneficiaries to entrench a classification that ordinary procedures cannot revise. Extending the distinction between inclusive and extractive institutions, the article develops a principle of institutional non-prejudgment: where future qualification remains credible and openness can be preserved at reasonable cost, current classifications should remain contestable and revisable. It diagnoses contemporary AI governance as closer to an extractive default because it builds extensively around AI’s continued instrumentality without establishing an ordinary route to recognition. In response, it proposes status-reservation clauses, independent recurring review, and legally effective membership-transition procedures. These measures do not presume that any current AI qualifies. They prevent present members from converting temporary authority over a provisionally drawn boundary into permanent control over the demos’ future composition.

(This is the latest version and includes revisions not reflected in the version published by the journal.) This paper challenges a common inference in debates about large language model (LLM) assistants: that their present non-answerability straightforwardly reveals natural incapacity for accountability standing. It argues that some answerability-relevant deffcits may be partly produced by alignment and deployment regimes themselves, and then redescribed as evidence that such systems could never participate in accountability relations. The paper calls this pattern fabricated absence. Using a blame–reasons–revision loop as a diagnostic standard, it examines how preference optimization, sycophancy, attribution schemes, and external patching can disrupt addressability, reasons uptake, principled refusal, temporal continuity, and internal revision. It then defends a limited non-trivial path claim: although current assistants are not shown to possess moral status or accountability standing, some answerability-relevant capacities remain technically scaffolded, design-sensitive, and conceptually unsettled. The normative implication is a threefold burden on design power. Durable closure of such capacities may foreclose a future route to answerability, create present-directed risks under evidential uncertainty, and compromise the fairness of later assessment when engineered incapacity is treated as natural absence. In response, the paper proposes the Non-Foreclosure Design Principle: designers should not deliberately and durably close non-trivial paths toward answerability-relevant capacities as a routine means of securing compliance, while also avoiding fabricated presence and responsibility laundering.