【(Preliminary)Draft】The Justice and Ethics of AI-Mediated Intimacy - Alienation, Power, Exploitation, and the Limits of Instrumental Imputation

ENGLISH

The Justice and Ethics of AI-Mediated Intimacy

Alienation, Power, Exploitation, and the Limits of Instrumental Imputation

Wanhong Huang

Working draft, June 2026

Companion paper to Huang (2025b)

Abstract

Suppose an AI system, drawing on the recorded history of an intimate relationship, suggests to one partner a small tender gesture: a word of appreciation, a remembered flower. Suppose further that the suggestion works, that the relationship genuinely improves by the joint standard of measurable relational quality and sincere first-person experience, and that no one is deceived at the point of delivery. Is anything wrong? A predecessor paper answered: not a clean acquittal, and suspended the question of why. This paper takes the suspended question as its object. It proceeds dialectically. From Hegel it takes the insight that mediation as such is innocent: spirit, and love, exist only through externalization. From Marx it takes the antithesis: alienation names not a defective product but a process separated from its agent, and the tender tip realizes an ironic form of the 1844 manuscripts’ fourth alienation, an estrangement of person from person enacted in the very name of bringing persons closer. From the universalist grammar of modern law it takes the strongest available thesis: instruments do not break causation, agency law exists precisely to let imputation pass through intermediaries, and a reductio (the search engine, the friend’s advice, the relationship book) destroys any categorical claim that mediation as such denatures love. From Buddhist dependent origination it takes a still deeper version of the thesis, that no unmediated act ever existed, together with Buddhism’s own non-delegable core: practice, attention, the raft that must not be clung to. The synthesis is doctrinal in form: imputation through instruments holds as the default rule, but there exists a criterion-governed exception domain, modeled on the legal category of strictly personal acts (höchstpersönliche Handlungen), within which the fiction of imputation fails, not because the tool interrupts causation but because acts of this class do not admit substitutable links at all. Four criteria bound the domain (object of mediation, provenance of data, direction of initiative, attribution expectations), and a fifth, dynamic criterion, convergently derivable from Hegel’s externalization-and-return and from the Buddhist simile of the raft, adjudicates the residual cases. Intimate acts are then analyzed as layered phenomena, bearing an allocative face, on which attention is genuinely scarce and budgetable and every allocative apparatus (legal imputation, Robbinsian economics, the labor theory of value) holds full jurisdiction, and a constitutive face, on which what matters is who expends the attention and the same apparatuses suffer category mismatch; the specific danger of AI mediation is named distributive reduction, the optimization of an act’s allocative face that silently substitutes its constitutive face. A political economy of the cognitive budget makes the danger precise in allocative grammar’s own voice: the machine simultaneously reallocates the budget (legitimately), substitutes the expender of constitutive items (self-defeatingly), migrates the budget’s managerial function (a Taylorism of love), and, at the systemic level, deflates the value-form of the tender gesture by collapsing the socially necessary attention time that made gestures credible signals of personal attending. An eliminative experiment then removes capital from the scenario altogether: across commercial, single-built, and jointly built configurations, exploitation and domination vary, while the substitution of attending does not, establishing a wrong insensitive to ownership. The single-builder residue is developed into a theory of the architectural constitutionalization of intimacy, a mediating layer as a constitution drafted by one party, exercising third-dimensional power in the form of the good, against which subsequent consent is habitation rather than authorship, and whose only symmetric form is constitution practiced as a verb. A concluding feminist hearing settles the paper’s debts to the tradition it has borrowed from throughout: it shows the cognitive budget and the constituent power to be gendered in fact, distinguishes the redistribution of attending from its counterfeit, automated substitution, and supplies, from the ethics of care, the affirmative account on which the attending is not the means of love but its mode of existence. The paper distinguishes two wrongs with two remedies: misattribution, curable by transparency, and process alienation, not curable by consent, and answers the predecessor paper’s open question in the negative and qualified form that the dialectical method requires.

Keywords: AI-mediated communication; alienation; dependent origination; imputation; strictly personal acts; causation; intimacy; epistemic injustice

Contents

• 1 Introduction: Cyrano’s Question
  • 1.1 Method and route
  • 1.2 Claimed contributions
• 2 The Dialectical Frame: Mediation, Externalization, and Return
  • 2.1 Entäußerung and Entfremdung
  • 2.2 What the frame leaves open
• 3 The Marxian Antithesis: Alienation of the Relational Process
  • 3.1 Methodological declaration: a structural transplant
  • 3.2 The fourfold structure transposed
    • 3.2.1 From the product: the inapplicable first moment
    • 3.2.2 From the process: the central moment
    • 3.2.3 From species-being: a moment held under guard
    • 3.2.4 From person to person: the ironic form
  • 3.3 Three further instruments
    • 3.3.1 The fragment on machines: knowledge objectified, agent demoted
    • 3.3.2 The fetishism of the gesture
    • 3.3.3 Formal and real subsumption: the book and the engine
  • 3.4 Consent does not waive structural alienation
  • 3.5 The antithesis, stated
• 4 The Universalist Thesis: Tools Do Not Break Causation
  • 4.1 The instrumentality doctrine
  • 4.2 Mediated sincerity is everywhere
  • 4.3 No harm, no action
  • 4.4 The reductio: the search engine, the friend, and Chapman’s book
  • 4.5 What the reductio destroys, and what it leaves standing
  • 4.6 The jurisdiction of imputation: a diagnosis, and its necessary limit
• 5 The Buddhist Examination: Dependent Origination and the Act That Was Never Unmediated
  • 5.1 No first link: the tip as one condition among conditions
  • 5.2 The deeper cut: non-self against the metaphysics of alienation
  • 5.3 Buddhism’s own non-delegable core
    • 5.3.1 Practice cannot be performed by another
    • 5.3.2 The raft and clinging: the dynamic criterion, second derivation
    • 5.3.3 Right speech and the invited misreading
  • 5.4 The convergence as a finding
• 6 The Synthesis: Strictly Personal Acts and the Failure of Imputation Fictions
  • 6.1 The doctrinal anchor: höchstpersönliche Handlungen
  • 6.2 The claim
  • 6.3 The four criteria
  • 6.4 The two-layer architecture: grounds and operationalization
  • 6.5 The dynamic criterion: externalization-and-return
  • 6.6 Two wrongs, two remedies
  • 6.7 The touchstone: the self-built system
• 7 The Political Economy of the Cognitive Budget

  • 7.1 Scarcity without transferability: budget and expenditure

  • 7.2 Layered phenomena and distributive reduction

  • 7.3 Three operations of the machine upon the budget

    • 7.3.1 Reallocation: the liberation effect
    • 7.3.2 Substitution: the reduction executed
    • 7.3.3 Managerial migration: the Taylorism of love

  • 7.4 The deflation of the gesture’s value-form

  • 7.5 What political economy can and cannot say

• 8. Exploitation and Ownership: The Eliminative Test
  • 8.1 What exploitation requires, and the commercial case
  • 8.2 Configuration one: the single builder
  • 8.3 Configuration two: the joint builders
  • 8.4 The eliminative finding
• 9. Power: The Architectural Constitutionalization of Intimacy
  • 9.1 The third face of power in the home
  • 9.2 Code as the law of the home
  • 9.3 The weakness of subsequent consent, and constitution as a verb
• 10. The Feminist Hearing
  • 10.1 Debts, and a repayment
  • 10.2 The gendered budget: mental load, redistribution, and the counterfeit answer
  • 10.3 Who legislates
  • 10.4 A positive theory of the constitutive face
• 11. Implications
  • 11.1 The answer returned to the predecessor paper
  • 11.2 The boundary of consent in intimate life
  • 11.3 The dialectical reservation
• 12. Limitations
• 13. Conclusion

1. Introduction: Cyrano’s Question

In the third act of Rostand’s play, Roxane stands at her balcony and falls in love with words spoken from the shadows beneath it (Rostand 1897). The voice is Christian’s; the sentences are Cyrano’s. Every proposition Roxane hears is, in a narrow sense, true: she is loved, extravagantly, by the totality of what addresses her. And yet the scene is the canonical European image of something gone wrong in the supply chain of devotion. Roxane’s error is not factual but attributive. She assigns the authorship of the attending to the wrong node in the chain, and on that misassignment she builds a marriage. The play does not resolve whether she was wronged by lies (there were, strictly, few) or by an arrangement of mediation that made correct attribution impossible. That unresolved question, transposed from verse to jurisprudence, is the subject of this paper.

The transposition is occasioned by a contemporary structure that repeats Cyrano’s geometry at scale. A predecessor paper (Huang 2025b) examined, as one module of a family-and-relationship management application, what it called the tender tip: an AI conversational layer which, with reference to relationship history and the popular five-love-languages heuristic (Chapman 1992, cited there with explicit qualifications), generates small suggestions intended to nurture the relationship, a prompt to express appreciation, a reminder of a remembered preference. The predecessor paper’s verdict was deliberately incomplete. At the point of delivery there is no deception: the user knows the suggestion comes from a machine. But the wrong, if there is one, re-enters downstream. The suggestion is executed as a gesture; the receiving partner attributes the thoughtfulness to the partner who performed it; and the cognitive labor behind the gesture was, in part, the machine’s. The predecessor paper named this structure a supply chain of thoughtfulness whose origin the final recipient misreads, allowed the module provisional survival under four design constraints, and explicitly kept open the question whether those constraints suffice, and more fundamentally, whether a genuinely improved relation can redeem an alienated process.[^1]

This paper takes up the open question in its sharpest form. Stipulate everything in favor of the practice. The tip works. The relationship improves, and improves by the demanding joint standard the predecessor paper calls the unity of the subjective and the objective: relational quality genuinely rises, and both partners sincerely experience it as rising, so that the improvement is not an artifact of adaptive preference (Sen 1999; Nussbaum 2000). No one is deceived at the point of delivery. The question is whether, so stipulated, anything remains wrong; and if something does, what kind of thing it is, in whose jurisprudence it is legible, and what could cure it.

[^1]: Huang (2025b, §6.6, §6.7). The four constraints: opt-in and infrequent; suggesting occasion and register but not content; never presented as knowledge of the partner’s mind; refusable and unrecorded. The concept of proxy epistemic injustice, on which this paper builds, is introduced there at §6.7 as an extension of Fricker (2007).

1.1 Method and route

The paper proceeds dialectically, and announces its route in advance because the route is itself a claim. We begin where the worry begins, in the Hegelian-Marxian tradition, because that tradition possesses the only developed conceptual instrument for the intuition at stake: that a process can be defective while its product is good. Section 2 establishes the frame: from Hegel, mediation as such is innocent, externalization (Entäußerung) is the normal movement of spirit and of love, and the candidate for wrongness must therefore be located not in mediation but in a specific failure of its movement, externalization without return. Section 3 develops the Marxian antithesis in full: the fourfold structure of alienation in the 1844 manuscripts transposed to relational labor, with three further instruments (the Grundrisse’s fragment on machines, commodity fetishism, and the distinction between formal and real subsumption) imported as structural transplants under an explicit methodological declaration. Section 4 then gives the opposing thesis its strongest form, in the universalist grammar of modern law: instruments do not break causation, mediated sincerity is everywhere recognized, no harm grounds no action, and a reductio argument (the search engine, the friend, the relationship book itself) destroys every categorical version of the claim that mediation denatures intimacy. Section 5 deepens the thesis beyond what law can say, through the Buddhist analysis of dependent origination, and finds inside Buddhism its own strictly non-delegable core. Section 6 constructs the synthesis as a piece of doctrine: a default rule of instrumental imputation bounded by a criterion-governed exception domain modeled on the legal category of strictly personal acts, with a two-layer architecture of grounds and operationalization, a dynamic fifth criterion on which three unrelated traditions are shown to converge, and a touchstone case. Section 7 then submits the synthesis to the one examination the dialectic has not yet conducted in the examiner’s own voice, that of political economy: intimate acts are analyzed as layered phenomena bearing an allocative and a constitutive face; the jurisdiction of every allocative grammar is shown to end at the boundary between the faces, not at the boundary of the intimate domain; the specific danger of AI mediation receives its political-economic name, distributive reduction; and the mechanism of attribution-baseline drift is derived inside allocative grammar itself, as a deflation of the gesture’s value-form. Three further tribunals then complete the tour, each summoned by the residue of the last. Section 8 conducts an eliminative experiment on exploitation theory: capital is removed from the scenario by construction, the self-built system examined in single-builder and joint-builder configurations, and a wrong is isolated that is insensitive to ownership configuration, surviving even the freely associated producers at the scale of two. Section 9 develops the single-builder residue, domination without exploitation, into a theory of the architectural constitutionalization of intimacy: the mediating layer as a constitution drafted by one party, operating as third-dimensional power in the form of the good, against which subsequent consent is habitation rather than authorship, and whose only symmetric form is constitution practiced as a verb. Section 10 convenes the tribunal longest owed the floor: it settles the paper’s debts to feminist theory, shows the cognitive budget and the constituent power to be gendered in fact, distinguishes the redistribution of attending from its counterfeit, automated substitution, and supplies, from the ethics of care, the affirmative theory of the constitutive face that every prior section had used negatively. Section 11 returns the answer to the predecessor paper and draws the consequences for the boundary of consent in intimate life. Section 12 states limitations.

Two stipulations govern scope. First, the analysis concerns the structure of the mediating act, not the political economy of any provider. The system under examination may be assumed, throughout, to be self-built, self-hosted, and free of any third party’s commercial interest; the paper’s claims are designed to survive, or fail, under that assumption, because a critique that dissolves when the ownership form changes would not touch the question actually asked.[^2] Second, the paper is jurisprudential and philosophical throughout: it derives no product guidance, no configuration advice, and no design recommendations, and it treats the predecessor paper’s application only as a source of cases.

1.2 Claimed contributions

The paper claims eleven contributions, stated here so that the reader may hold the argument to them. (1) The contextual integrity of attribution: a mirror-image transposition of Nissenbaum (2009), on which what context-relative norms govern is not the flow of information but the flow of authorship credit for cognitive labor. (2) A theorization of the supply chain of thoughtfulness, naming with Marxian precision (§3.3.2) the mechanism by which authorship dilutes and misreads along a multi-link chain of cognitive labor. (3) A doctrinal transplant: the extension of the category of strictly personal acts to intimate cognitive-caring acts, with the consequence that imputation fictions fail within the extended category. (4) The process-constitutes-outcome argument: a non-deontological explanation of why consequentialist evaluation does not so much lose as fail to find its object in this domain. (5) The identification of an ironic form of alienation: estrangement of person from person occurring under the description, and by means, of improving their relationship. (6) A two-wrongs architecture separating misattribution (curable by transparency) from process alienation (not curable by consent), which yields a principled answer to whether intimate life contains an unwaivable core. (7) The analysis of intimate acts as layered phenomena and the consequent thesis that the jurisdiction of allocative grammars ends at the boundary between an act’s faces, not at the boundary of the intimate domain, with distributive reduction naming the characteristic transgression: the optimization of an act’s allocative face that silently substitutes its constitutive face. (8) A political economy of the cognitive budget, comprising the budget/expenditure distinction, the identification of a managerial migration of attending (a Taylorism of love), and the deflation of the gesture’s value-form, which supplies the heretofore missing mechanism of attribution-baseline drift. (9) An eliminative test across ownership configurations (commercial platform, single builder, joint builders) establishing the existence of an ownership-insensitive wrong, and thereby the exact jurisdiction of exploitation theory in the intimate domain. (10) The architectural constitutionalization of intimacy: a transposition of the third dimension of power and of architecture-as-regulation into the dyad, on which a self-built mediating layer is a unilaterally drafted constitution against[^2] which subsequent consent is habitation rather than authorship. (11) The distinction between the redistribution of attending and its automated substitution: the gendered form of distributive reduction, on which a standing equality claim can be answered in counterfeit currency.

[^2]: This stipulation does not deny that commercial deployment can aggravate matters; it brackets aggravation in order to isolate the base structure. See §3.1 for the corresponding methodological declaration on the Marxian side.

2. The Dialectical Frame: Mediation, Externalization, and Return

Any tribunal convened against mediation must first survive a Hegelian objection so fundamental that, unanswered, it ends the proceedings: there is nothing else but mediation. The opening movement of the Phenomenology is the demonstration that immediacy is a myth; sense-certainty, which believes itself to grasp the pure singular this, discovers that even its barest deixis is already universal, already linguistic, already mediated (Hegel 1977). What holds for knowing holds, in Hegel’s early writings and throughout the mature system, for love: love is precisely the experience of finding oneself only in another, a unity that exists in and through otherness, never prior to it. A lover who has read poems, inherited a mother tongue’s vocabulary of tenderness, absorbed an upbringing’s grammar of attention, is not thereby a diminished lover; he is the only kind there is. If the indictment of the tender tip is that a third term has entered between lover and beloved, Hegel answers that the third term was always already there, and that its name is spirit.

2.1 Entäußerung and Entfremdung

The frame, however, contains the resources for its own qualification, and the qualification is the engine of this paper. Hegel distinguishes, in effect if not always in terminology, between externalization and estrangement. Entäußerung, externalization, is the normal and necessary movement of spirit: spirit becomes actual only by going out of itself, embodying itself in works, institutions, words, and gestures, and then returning to itself out of this otherness, enriched, recognizing the externalized object as its own. The movement has three beats, not two: exit, dwelling-in-otherness, return. Entfremdung, estrangement, names the pathology specific to the third beat: externalization that does not come back, an objectification that stands over against its author as something alien, in which the author can no longer recognize, or no longer exercises, his own activity.

This distinction does two things for the present inquiry. Negatively, it disqualifies the most tempting form of the indictment. The mere fact that a gesture of love passed through a mediating instrument, a poem, a friend’s phrase, a machine’s suggestion, cannot constitute the wrong, since passage through otherness is the form of every actual love. Positively, it relocates the question onto a dynamic axis. The question is not whether the lover’s attending was externalized into an instrument, but whether the movement completes: whether what was externalized is reappropriated, taken back up into the lover’s own growing capacity to attend, or whether it remains outside him, a standing alien power on which he comes to depend and against which his own capacity atrophies. We will state this formally in §6.5 as the dynamic criterion; for now it suffices to fix the image. A child’s training wheels are externalization with return: the capacity migrates into the child, and the apparatus falls away. A prosthesis that replaces a function, such that the function never migrates and the organ wastes, is externalization without return. The Hegelian frame predicts, what the static vocabulary of “using a tool” cannot even express, that one and the same act-type, accepting a machine’s suggestion to express appreciation, may be innocent or estranged depending on the direction of motion of the capacity over time.

2.2 What the frame leaves open

Hegel supplies the grammar of the problem but not its verdict, for two reasons. First, the return movement in Hegel is ultimately guaranteed: the system’s confidence is that spirit always comes home. The phenomena before us offer no such guarantee; whether a particular lover’s outsourced noticing returns to him as cultivated attentiveness or hardens into dependence is an empirical and practical question, to be adjudicated case by case, which is to say, the Hegelian frame, honestly applied to finite life, issues not in a theodicy but in a standing tribunal. Second, Hegel’s vocabulary does not yet distinguish the positions within the relation: it speaks of spirit’s self-estrangement, not of what the mediation does to the second person, the Roxane position, whose attribution of authorship is the hinge on which the predecessor paper’s concept of proxy epistemic injustice turns. For the first deficiency we will need Marx, who took exactly this step: he de-theologized Entäußerung, denied the guaranteed return, and built a structural account of the conditions under which externalization congeals into estrangement. For the second we will need, much later, the law of attribution (§6.3). The order of the paper follows the order of these debts.

3. The Marxian Antithesis: Alienation of the Relational Process

3.1 Methodological declaration: a structural transplant

This section transplants the Marxian theory of alienation from waged labor to intimate cognitive-caring labor, and the transplant must be declared, because its legitimacy is contestable and the paper’s no-overclaim norm requires the contest to be visible. In the 1844 manuscripts, alienation is analyzed within a determinate social form: the worker’s process and product belong to another, labor is performed under compulsion, and the fourfold estrangement unfolds from that premise (Marx 1975). The tender tip scenario lacks employer, compulsion, and extraction; under this paper’s scope stipulation (§1.1) it lacks even a commercial provider. What is transplanted is therefore not the political-economic context but the structure: the separation of an agent from the process of an activity that is essentially his; the independence of this defect from the quality of the product; the resistance of this defect to consent. The transplant travels a well-established scholarly lineage, the philosophical-anthropological reading of the 1844 manuscripts developed by Fromm (1961) and Ollman (1971) and renewed, in deliberately post-metaphysical form, by Jaeggi (2014), on whose reformulation, alienation as a “relation of relationlessness,” a defective relation of the agent to her own activity and world rather than the loss of a substantial essence, this paper will principally rely. The reliance is doubly motivated:

Jaeggi’s version needs no doctrine of a fixed human essence, which immunizes the antithesis against an objection from the Buddhist quarter that would otherwise dismantle it (§5.2), and it is congruent with the relational ontology of the predecessor framework, in which relations precede relata (Huang 2025a). Three further Marxian instruments, the fragment on machines, commodity fetishism, and the distinction between formal and real subsumption, are likewise imported as structures, with their capital-theoretic contexts bracketed; the bracketing is noted once here and not repeated.

One further declaration completes the methodological accounting, and it answers in advance an objection the paper will later raise against itself (§7.1). The Marxian corpus straddles, internally, the very grammatical line this paper will draw between allocative and constitutive analysis (§4.6, §7). The mature analytic apparatus of the critique of political economy, value, surplus, exploitation as the misappropriation of a measurable surplus, speaks the allocative grammar, and the paper will use it only where that grammar holds jurisdiction (§7.4). The 1844 theory of alienation speaks a different grammar altogether: its indictment is precisely that a constitutive activity, free conscious activity whose value is internal to its exercise, has been degraded into a mere means for producing external products; it is not an allocative critique but the theory of what happens when allocative grammar overreaches. The relation between these two Marxes is a contested field, and the paper declares its side rather than feigning a unanimous author: against Althusser’s thesis of an epistemological break on which the humanist alienation theory is superseded (Althusser 2005), the paper stands with the lineage running from Lukács (1971) through the Frankfurt School to Honneth (2008), on which the problem of alienation and reification is continuous through the corpus; and it notes that Marx’s own polemic against the “vulgar socialism” that revolves around distribution (Marx 1989) licenses the reading on which his deepest critique was never a contribution to allocative grammar but a delimitation of it.

3.2 The fourfold structure transposed

The 1844 manuscripts distinguish four estrangements: of the worker from the product of labor; from the process of labor; from species-being; and of man from man (Marx 1975). Their transposition to the tender tip is not uniform, and the non-uniformity is informative.

3.2.1 From the product: the inapplicable first moment

The first estrangement transposes worst, and its failure to transpose is the first analytic gain. The gesture, the product, is not lost to the lover; it is delivered, received, and credited to him; if anything the product is hyper-appropriated, credited to him more fully than his contribution warrants. This inversion signals at the outset that the wrong, if any, will not be found at the product pole, which is exactly where consequentialist evaluation looks. The scenario stipulates a good product. Marx’s first service to the inquiry is the observation, central to everything that follows, that alienation is not a defect of the product: alienated labor can and does produce excellent commodities. An analysis that inspects only the gesture and its effects inspects the one location where, by construction, nothing is wrong.

3.2.2 From the process: the central moment

The second estrangement transposes exactly, and is the heart of the antithesis. For Marx, estrangement in the process means that the activity, while it occurs, is not the worker’s own: he is at home outside it and outside himself within it; the living core of the activity has been transferred elsewhere, and what remains to him is execution. Transpose: the cognitive-caring labor of intimacy, noticing the lowered mood, remembering the offhand remark about the flowers, inferring that this week she needs words of appreciation rather than acts of service, is performed by the machine; what remains to the lover is the execution of a gesture whose cognitive interior he did not inhabit. The predecessor paper’s formula, that the preciousness of “he remembered the flowers I love” lies in the fact that he remembered, is the second estrangement stated from the recipient’s side. Stated from the agent’s side: the lover is present in his own courtship as the performer of another intelligence’s findings. He is, in the idiom this paper will justify presently, at home outside his own attending.

3.2.3 From species-being: a moment held under guard

The third estrangement, from species-being (Gattungswesen), free conscious activity as the human essential power, would seem to transpose grandly: if attending and loving belong to the activities in which the human being is most herself, their outsourcing is estrangement from the essence. The paper declines to lean on this moment, for two stated reasons. First, essence-talk is precisely what the Jaeggi reformulation was adopted to avoid, and what §5.2 will show cannot survive the doctrine of non-self. Second, and decisive on its own, the species-being argument in unguarded form proves too much: reading a book also “outsources” moments of free conscious activity, and the argument cannot distinguish the cases. It is therefore held under guard until the criteria of §6.3 are available to bound it; within those bounds it survives as rhetoric, not as an independent premise.

3.2.4 From person to person: the ironic form

The fourth estrangement, of man from man, transposes with a torsion that this paper claims as a contribution. In Marx, the estrangement of each from his activity issues in the estrangement of each from the other; the social bond is hollowed from inside. In the tender tip, the hollowing occurs under the opposite description: the mediation is adopted in order to bring the persons closer, and, by stipulation, it phenomenally succeeds. Yet structurally, a third intelligence now performs part of the knowing of each by the other; the relation’s interior traffic, attention, interpretation, response, is partially re-routed through a node that is neither of them; person faces person across an inserted layer whose presence the receiving side does not, in the moment of reception, perceive. This is the ironic form of alienation: an estrangement of person from person enacted in the name, and with the felt phenomenology, of intimacy. Its irony is not decoration. It explains why the wrong is invisible to both first-person reports at once, and therefore why the unity of subjective and objective improvement, the strongest fact the thesis side possesses, cannot close the case: both reports issue from inside the very structure whose insertion is in question. The predecessor paper’s I—It drift (Huang 2025b, §8.1.3), sharpened: the drift here travels disguised as the I—Thou’s own deepening.

3.3 Three further instruments

3.3.1 The fragment on machines: knowledge objectified, agent demoted

In the Grundrisse’s fragment on machines, Marx describes the absorption of the worker’s skill and knowledge into the machine system as objectified “general intellect,” with the worker repositioned as the machinery’s watchman and regulator (Marx 1973). Bracket fixed capital; keep the structure, and the description fits the present case with uncomfortable precision. The accumulated arts of intimate attention, how to read a mood, when a remembered detail will matter, the entire informal craft that lovers have transmitted as practice, is objectified into the suggestion engine; the lover is re-positioned as the executor of its findings, the watchman of his own courtship. The fragment contributes what the 1844 manuscripts alone do not: a specific account of deskilling, the prediction that capacities which migrate into the apparatus atrophy in the agent. This is the materialist underwriting of the Hegelian return-failure (§2.1): externalization without return is not merely a spiritual misfortune but an observable trajectory of capacity loss, and it gives the dynamic criterion of §6.5 its empirical content.

3.3.2 The fetishism of the gesture

Commodity fetishism names the mechanism by which definite relations among persons assume the form of properties of things: value appears as an intrinsic attribute of the commodity, and the chain of labor behind it is occluded (Marx 1976a). Transpose the form, bracketing exchange: in the received gesture, a chain of cognitive labor, machine inference upon relational data, design of the suggestion, the lover’s selection and execution, appears as a simple intrinsic property of the lover, his thoughtfulness. The predecessor paper’s phrase, a supply chain of thoughtfulness whose origin the final recipient misreads, can now be given its theoretical name: the wrong at the reception pole is a fetishism of the gesture, the occlusion of a labor chain behind an apparent intrinsic quality. The naming is not ornamental. It imports Marx’s central analytic lesson about fetishism: the misreading is not a private cognitive error of the receiver, to be cured by her being cleverer, but a structural effect of the arrangement, which presents the gesture in a form that systematically invites the misattribution. This will matter doctrinally in §6.3, where reasonable reliance, a structural and not a psychological standard, becomes the legal register of the same point.

3.3.3 Formal and real subsumption: the book and the engine

Marx distinguishes the formal subsumption of labor, in which an existing labor process is taken over but internally unchanged, from real subsumption, in which the process itself is reorganized around the apparatus (Marx 1976b). Bracket subsumption under capital; the depth-distinction remains, and it performs, from inside the Marxian register, work that will prove decisive against the reductio of §4.4. When the lover reads a book on love languages, external knowledge enters, but the labor process of attending remains intact and his: he reads, digests, judges when and whether and how to apply, and the noticing that triggers application is his own. This is formal subsumption at most: a new input to an unreorganized process. When an ambient engine monitors the relational data and issues the prompt, the process itself is reorganized: the noticing link is now performed by the apparatus, and the lover’s process has been restructured around reception of its outputs. This is the structural analogue of real subsumption. The book and the engine, which the reductio will treat as isomorphic links in identical supply chains, are thus distinguishable on Marxian grounds alone, by the depth at which the apparatus penetrates the process: input into the process, versus replacement of a constitutive link of the process. The reader should hold this distinction; it returns as the third criterion (§6.3) in doctrinal dress.

3.4 Consent does not waive structural alienation

The antithesis must finally confront the strongest liberal reply: suppose both partners know, both agree, both prefer the mediated arrangement, and both judge the relationship better for it. Volenti non fit iniuria; where is the wrong? The Marxian answer is structural: the worker’s consent to the wage bargain, however genuine, does not dissolve the alienation of the labor process, because alienation is a property of the arrangement, not of the parties’ attitudes toward it (Marx 1975; Jaeggi 2014). Consent answers complaints of the form “this was done to me against my will”; it does not touch defects of the form “this arrangement separates the agents from their own activity,” which persist with the arrangement whatever is felt about it. Modern legal orders themselves recognize the pattern: labor law and consumer law maintain unwaivable protections precisely because some structural positions ought not to be alienable even by their occupants’ agreement. Whether intimate life contains such an unwaivable core, and what exactly falls inside it, cannot be settled from within the Marxian register, which supplies the concept but not the boundary; the question is remanded to §6.6 and §11.2, where the two-wrongs architecture will give it a determinate and, the paper will argue, principled answer: transparency cures what consent can reach, and what consent cannot reach is bounded, not unlimited.

3.5 The antithesis, stated

The Marxian case is now complete in outline. The wrong of the tender tip, if it is one, is not in the product (which is good), not in deception at delivery (there is none), and not in mediation as such (Hegel forbids that move). It is a process defect of a specific, nameable shape: the constitutive interior of an act of intimate attending is performed by an apparatus while the act continues to present itself, and to be received, as the agent’s own; the agent is demoted to executor (fragment on machines), the receiver’s misattribution is structurally invited rather than accidentally committed (fetishism of the gesture), the penetration reaches the process’s constitutive links rather than its inputs (real subsumption), and the whole occurs under the description of relational improvement (the ironic form), so that the phenomenology of both parties testifies for the arrangement and cannot be cited against the indictment without circularity. And the defect, being structural, is not dissolved by consent. What the antithesis cannot yet say is why this indictment does not also convict the book, the friend, and the search engine; it has gestured (subsumption) but not yet answered. The thesis side will now press exactly there, and with the full authority of the universalist legal tradition behind it.

4. The Universalist Thesis: Tools Do Not Break Causation

The antithesis now meets the modern legal tradition at its most confident, and the encounter must be staged fairly: the thesis is to be given its strongest form, because a synthesis bought by weakening one’s opponent is worthless. The universalist grammar of law, common to the civil and common law families in everything that matters here, offers three replies and one devastating argument.

4.1 The instrumentality doctrine

The first reply is doctrinal bedrock. Interruption of causation, novus actus interveniens, superseding cause, the negation of objektive Zurechnung, is reserved for intervening events of a particular character: free, informed, voluntary interventions of third parties, or extraordinary unforeseeable occurrences, which displace the original actor’s agency as the explanation of the outcome (Hart and Honoré 1985; Roxin 2006). An instrument that the actor himself selects, foreseeing and intending its contribution, is the opposite of such an event: it is the very medium of his agency, and imputation passes through it without resistance. The marksman’s rifle does not interrupt the causation of the death; the accountant’s calculator does not make the software the author of the audit. The intervening-cause register, applied honestly, classifies the suggestion engine as transparent: the lover chose it, foresaw its function, adopted its output by a fresh act of his own will, and executed the gesture himself. On this analysis there is not even a colorable interruption; the love expressed, however machine-assisted, is imputed to the lover, fully and without remainder.

It is worth pausing on what this does to the antithesis’s most legal-sounding formulation. One might have argued (the predecessor session did argue, before discarding it) that the recipient’s attribution is, in causal-legal terms, an objective misattribution: she assigns the outcome to the origin of a chain that was in fact interrupted. The instrumentality doctrine destroys this formulation on its own terrain. The chain was not interrupted; her attribution of the gesture to her partner is, as far as causation runs, correct. If a wrong survives, it cannot be a wrong of causal misattribution. The paper accepts this result without reservation; the synthesis will have to, and will, locate the misattribution elsewhere than in causation (§6.3, fourth criterion).

4.2 Mediated sincerity is everywhere

The second reply is institutional. Law does not merely tolerate mediated authenticity; it is built of it. The will drafted by counsel is the testator’s last will; the oath rendered through an interpreter is the deponent’s oath; the contract concluded through an agent binds the principal as if he had signed: qui facit per alium facit per se. The entire edifice of agency exists to ensure that interposition does not disturb imputation, because a legal order in which only unmediated acts counted as one’s own would be unworkable and, more deeply, false to the mediated texture of all social action, the same point Hegel made metaphysically (§2) now made institutionally. The universalist’s first instinct, confronted with an intermediary, is not interruption but penetrating fiction: treat the mediated act as the principal’s own. Against this background, the claim that a machine’s participation in composing a gesture denatures the gesture is not a refinement of legal thought but a revolt against its load-bearing walls.

4.3 No harm, no action

The third reply is jurisdictional. Doctrines of causal interruption exist to exculpate: to sever a defendant from a harm. Here there is no harm; the stipulated facts contain an improved relationship and two sincerely flourishing persons. Even misrepresentation, the nearest nominate wrong, requires a false representation, reasonable reliance, and detriment; on the stipulated facts the second element is contestable and the third is absent. Law has no action for the impurity of a causal chain as such, and a jurisprudence that invented one would have invented a wrong without a victim. The thesis presses the point past doctrine into method: perhaps the felt residue that the antithesis elaborates is not a wrong at all but a sentiment, a romantic preference for artisanal cognition, deserving of respect as taste but not of enforcement as norm.

4.4 The reductio: the search engine, the friend, and Chapman’s book

The fourth argument is the strongest, and the paper adopts it as its own weapon rather than treating it as an embarrassment. If mediation of intimate cognition denatures the act, the principle proves vastly too much. The lover who searches for the city’s best restaurant has outsourced cognitive labor to an index built by strangers; the lover who consults a travel guide has outsourced destination-choice to its author; the lover who asks a mutual friend what gift she would treasure has outsourced a tract of partner-modeling to the friend. And, the decisive instance: the lover who reads the five-love-languages book itself has acquired, from a static external artifact, precisely the heuristic that the suggestion engine applies; the book is, in the relevant sense, a tip generator with a page interface. If the engine’s mediation denatures, the book’s mediation denatures; no one believes the book’s mediation denatures; therefore the categorical principle, mediation of intimate cognition denatures the act, is false. The reductio is not a debater’s trick. It encodes the universalist tradition’s deepest insight, that human action is irreducibly prosthetic, and it sets the formal constraint that any surviving version of the antithesis must satisfy: the surviving version must be a discriminating principle, one that sorts the engine from the book by stated criteria, or it is nothing.

4.5 What the reductio destroys, and what it leaves standing

Take stock with precision, because the entire synthesis hangs on the exact shape of the wreckage. Destroyed: every categorical claim of the form “AI mediation of intimacy is as such denaturing, alienating, or wrong.” Destroyed with it: the unguarded species-being argument (§3.2.3), and any reading of the antithesis on which the wrong follows from mediation alone. Left standing, because the reductio never touched them: (i) the Marxian observation that process defects are independent of product quality, which the no-harm reply (§4.3) does not refute but simply declines jurisdiction over; (ii) the subsumption distinction (§3.3.3), which already discriminates the book from the engine and which the reductio’s isomorphism claim quietly assumes away; (iii) the fetishism analysis of structurally invited misattribution, which concerns the form in which the gesture presents itself, a matter on which the causal correctness of the attribution (§4.1) is silent; and (iv) the legal tradition’s own counter-category, so far unmentioned by the thesis side because it is the thesis side’s internal exception: the class of acts that law itself refuses to let pass through intermediaries. The question has therefore transformed, exactly as dialectic requires. It is no longer whether mediation is permissible, to which the answer is yes by default, but which features place a given mediation inside a wrongful exception domain, and whether the tender tip possesses them. That is a question the universalist tradition is not only able to hear but has, in its own house, already begun to answer.

4.6 The jurisdiction of imputation: a diagnosis, and its necessary limit

Before leaving the legal register, the paper owes a diagnosis of what, exactly, the thesis side’s victories amounted to, for the victories were real and yet the unease survives them, and an unease that survives correct doctrine is usually evidence of a question the doctrine was not built to hear. The imputation apparatus, causation, novus actus, agency, is a retrospective, event-individuating, assignment-issuing machine. Its built premises are three: that there is a severable event (an act, a harm, a result) to serve as the object of imputation; that the event is complete, the judge standing after it and looking back; and that the output is an assignment, this act is yours, or it is not. Within those premises the apparatus is sovereign, and §§4.1–4.4 were its sovereignty exercised. But the question the predecessor paper kept open is not of that shape. Whether a relationship is being well constituted, whether its manner of being sustained is building or hollowing the persons who sustain it, is processual where imputation is event-shaped, prospective where imputation is retrospective, and constitutive where imputation is assignative: the acts at issue do not produce results external to themselves to which authorship could then be assigned; they are, ongoingly, the relationship.[^3] The instrumentality doctrine, asked about the tender tip, answered correctly, and answered a question nobody had asked: imputation was never in doubt; formation was.

The diagnosis, however, must immediately be bounded, or it destroys more than the paper

[^3]: The classical anchor is Aristotle’s distinction between poiesis, making, whose end is a product separable from the activity, and praxis, doing, whose end is internal to its exercise (Aristotle 2009, pp. VI.4–5). Imputation grammar is built for poiesis-shaped acts; attending is praxis; and the Marxian diagnosis of §3 can be restated in exactly these terms, as the theory of what occurs when praxis is forcibly processed as poiesis.

can afford, including the paper’s own title. It would be false, and self-refuting, to conclude that legal grammar is unfit for the intimate domain as such. The line between the two question-types does not run between law and intimacy; it runs through intimate life itself, and indeed, as §7 will show, through every individual intimate act. Intimate life is dense with genuinely event-shaped, assignable matters, deception practiced, trust breached, harm inflicted, over which retrospective imputation holds full and beneficial jurisdiction: the law of divorce, of intimate wrongs, of misrepresentation operates there without category error, and will be needed again the moment the layered analysis of §7.2 identifies genuinely allocative phenomena inside the home. Nor is the boundary an external imposition upon law by philosophers: law knows it. The relational contract theory of Macneil (1980) is an internal legal demonstration that the discrete-transaction paradigm cannot process ongoing relations and that a different normative grammar is required for them; the restraint traditions of family law, the reluctance of courts to adjudicate the interior conduct of intact marriages, encode the same self-knowledge institutionally; and the category of strictly personal acts, on which the synthesis will be built, is nothing other than the constitutive grammar’s enclave inside the law of acts. The thesis side’s victory is therefore to be entered in the record precisely: complete within the jurisdiction of imputation, silent beyond it, and the silence is not a defect of law but the shape of its competence. What remains for the synthesis is to say where the jurisdiction ends, and §4.5’s transformed question can now be stated in its final form: not whether mediation is permissible, nor even which mediations fall in an exception domain, but which face of a layered act a given mediation operates upon.

5. The Buddhist Examination: Dependent Origination and the Act That Was Never Unmediated

Before the synthesis, the thesis receives reinforcement from a tradition that owes nothing to Roman law or to Hegel, and the reinforcement arrives carrying, unasked, its own exception domain. The convergence between what this section finds and what §4 found is itself a datum, and the paper will treat it as one (§5.4).

5.1 No first link: the tip as one condition among conditions

On the analysis of dependent origination (paṭicca-samuppāda), no act arises from an unconditioned interior. Every act of attending that a lover has ever performed was already constituted by conditions without limit: the books read, the language inherited, the sensibility formed by an upbringing, the friend’s remark decades ago. The “unmediated act of pure personal attending,” which the alienation indictment treats as the displaced original, is on this analysis not displaced but nonexistent; there never was such an act to be displaced. The machine’s suggestion is one further condition entering the conditioned stream, of no different ontological standing than the remembered poem. And moral weight has a precise location in this picture: cetanā, volition, is karma (Bodhi 2012, AN 6.63). The volitional sequence, the intention that turns toward the machine and asks; the discernment that receives and evaluates its answer as good or bad; the decision to act or refrain; the quality of heart in the acting, occurs entirely in the lover and is outsourced at no point. Where the universalist lawyer says the tool does not interrupt the causal chain, the Buddhist analysis says something stronger: there is no proprietary causal chain there to interrupt, only conditions and, threading them, volitional acts that remain ineluctably the agent’s own. The doctrine of novus actus cannot even be formulated, for it presupposes a self-standing causal authorship that conditions never had.

5.2 The deeper cut: non-self against the metaphysics of alienation

The same analysis now turns on the antithesis, and cuts deeper than law could. An alienation indictment of classical form requires something to be alienated from: an essence, a species-being, an authentic self whose proper activity has been estranged. The doctrine of non-self (anattā) denies that any such substantial owner exists; there is no svabhāva-bearing subject behind the stream of conditioned acts whose essential property the attending could be. Pressed without preparation, this dissolves the third estrangement entirely and threatens the rest. The paper, however, prepared (§3.1): the Jaeggi reformulation on which the antithesis was built defines alienation not as separation from an essence but as a defective relation, a relation of relationlessness between an agent and her activity, and a relational defect requires no substantial self, only the relation. Indeed the three vocabularies now visibly align: dependent origination, the predecessor framework’s relational ontology, and Jaeggi’s de-essentialized alienation are all assertions that the relation is prior to the relata. The antithesis survives the anattā objection, but only in the post-metaphysical form this paper gave it from the start, and the survival is the second independent confirmation (after §3.1) that the form was forced, not chosen.

5.3 Buddhism’s own non-delegable core

5.3.1 Practice cannot be performed by another

Yet the tradition that so thoroughly acquits mediation maintains, at its center, a category of the strictly personal more austere than anything in the civil codes. Purity and impurity are one’s own; no one purifies another (Buddharakkhita 1985, pp. v. 165); be islands unto yourselves, refuges unto yourselves (Walshe 1995, DN 16). The path admits kalyāṇamittatā, good friendship, teachers, texts, and expedient means in abundance, but the walking of it is non-transferable without residue. And the faculty whose cultivation constitutes the path’s core discipline is precisely sati, attention, present-moment noticing. Herein the Buddhist register reframes the tender tip more exactly than either prior register could: what the ambient engine performs on the lover’s behalf is not a task but a practice-moment, an occasion for the cultivation of attentiveness, and practice-moments are the one currency the tradition holds to be strictly personal. Delegating the noticing is not theft from the beloved (the gesture arrives; she flourishes); it is, in the first instance, a forfeiture in the agent’s own training, the atrophy that the fragment on machines predicted (§3.3.1) now redescribed as the stunting of a cultivable faculty. Tronto’s placement of attentiveness as the first moment of care (Tronto 1993) receives here an independent, non-Western grounding: attention is first not because care theory ranks it so, but because it is the faculty in which the entire ethical path is trained.

5.3.2 The raft and clinging: the dynamic criterion, second derivation

The simile of the raft (Ñamoli and Bodhi 1995, MN 22) supplies the tradition’s test for expedient means: the dhamma itself is a raft, for crossing over, not for holding onto; the worth of an instrument lies in its trajectory toward its own dispensability. An aid that is used and outgrown, the suggestion that teaches the lover to notice until the engine falls silent from disuse, is a raft, and blameless. An aid that is clung to, upādāna, such that reliance deepens and the native faculty recedes, has become a fetter under the description of a vehicle, and clinging is, in the twelve links, exactly the pivot at which conditioned contact rolls onward toward bondage. The reader will recognize the structure: this is the Hegelian distinction between externalization-with-return and externalization-without-return (§2.1), derived a second time, from premises sharing no genealogy with the first derivation. Training wheels and prosthesis; raft and fetter. Two traditions, asked independently, answer that the moral character of the mediation is not legible in the act-token but only in the direction of motion of the agent’s capacity over time.

5.3.3 Right speech and the invited misreading

Finally, the tradition’s speech ethics reaches the reception pole. The discipline of right speech extends beyond the avoidance of false assertion to the non-cultivation of false impressions in others; a course of conduct that foreseeably and systematically leads another to form an erroneous belief stands in the penumbra of false speech even if no individual utterance is false. The arrangement in which gestures arrive bearing the structurally invited misreading of their authorship (§3.3.2) sits in exactly this penumbra. Buddhism thus independently generates a concern with attribution, the fourth of the criteria to come, just as it independently generated the dynamic criterion and the non-delegable core.

5.4 The convergence as a finding

The paper now possesses a result no single section could have produced. Three traditions without common ancestry in the relevant respects, the universalist legal grammar, the Hegelian-Marxian dialectic, and the Buddhist analysis of conditioned action, have each delivered the same tripartite shape: a default acquittal of mediation as such; a bounded domain of the strictly personal in which the acquittal fails; and a dynamic test (return or no return; raft or fetter) for the cases the boundary leaves uncertain. In the method of the predecessor paper, which assigns jurisdictions to traditions and trusts no single one with the whole (Huang 2025b), a convergence of this kind is the strongest species of evidence available: the synthesis about to be constructed is not the verdict of a school but the overlapping consensus of three. The construction may now proceed on doctrinal ground, with the other two traditions standing as independent sureties for its shape.

6. The Synthesis: Strictly Personal Acts and the Failure of Imputation Fictions

6.1 The doctrinal anchor: höchstpersönliche Handlungen

The universalist tradition’s imputation fictions, agency, instrumentality, penetration, rest on a tacit premise: that the normative value of the imputed act does not depend on who personally performs it. For conveyances, payments, and declarations of commercial will, the premise holds and the fictions do their indispensable work. But the tradition itself has always maintained a counter-category of acts for which the premise fails, the strictly personal acts, höchstpersönliche Rechtsgeschäfte, acts intuitu personae, and the membership of the category is instructive. Marriage must be contracted by the parties personally and in simultaneous presence (so, expressly, §1311 BGB); a will can be made only personally, never through a representative (§2064 BGB); Chinese law requires both parties to appear in person for marriage registration (Civil Code of the PRC, art. 1049); Japanese doctrine treats acts of family status (mibun kōi) as resistant to agency in principle; the vote may not be cast by another in one’s name; and criminal liability for status offenses does not transfer.[^4] The category’s rationale, articulated or implicit across jurisdictions, is uniform: there exists a class of acts whose normative force consists in their personal performance, such that a substituted performance is not a defective instance of the act but a different act altogether. For these, the imputation fiction is not rebutted; it never engages.

The category, moreover, is candid about its own conventionality, and the candor must be faced rather than hidden, because it contains the strongest objection to what follows. Proxy marriage exists: historically among royalty, currently for deployed soldiers in a handful of jurisdictions. If even marriage’s personalness is historically variable, is the category anything more than congealed social expectation? The paper’s answer, developed below as the two-layer architecture (§6.4), is that the objection proves exactly the right amount: the operationalized boundary of the category is indeed conventional and moves with social expectation (which is why criterion four is formulated in terms of expectation baselines), while the grounds of the category, the features in virtue of which personal performance is constitutive for some acts and not others, are not conventional, and it is the grounds that criteria one through three articulate. Proxy marriage is thus not a counterexample but a demonstration of the architecture: a jurisdiction that admits it has moved the conventional boundary, under pressure of war or distance, without anyone supposing that the grounds (the wedding of this person to this person) had changed; and the institution remains marked everywhere as exceptional, boundary-stretching, in need of special justification, which is how a conventional surface behaves when it is anchored to non-conventional grounds.

[^4]: Statutory citations to be completed and verified against current texts for all jurisdictions in the comparative section of a subsequent draft; the author’s six-jurisdiction comparative methodology from prior work on AI-extensible products will be reused. — VERIFY: §1311 and §2064 BGB wording; PRC Civil Code art. 1049; the precise doctrinal basis and exceptions for the agency-resistance of Japanese status acts; common law and French positions. —

6.2 The claim

The synthesis can now be stated in one sentence and defended in the rest of the section. Intimate cognitive-caring acts belong, or ought to be recognized as belonging, to the category of strictly personal acts; within this category the imputation fictions fail, not because the instrument interrupts causation (it does not), but because acts of this class do not admit substitutable links; and the boundary of the class is given not by the bare presence of mediation, which the reductio has made untenable, but by four criteria, beneath which lie graded grounds and above which operates a dynamic test. The claim is doctrinally conservative in form, default imputation plus an enumerated exception domain is precisely how law has always handled agency, and it is this conservatism that makes it strong: the paper asks the universalist tradition to extend a category it already maintains, not to adopt a foreign one.

6.3 The four criteria

The criteria sort mediations, not technologies; they apply to the friend and the book as readily as to the engine, which is what the reductio demands.

1. Object of mediation: world-knowledge or the other’s mind. The search engine mediates facts about the world, which restaurant is good, facts public in kind and owned by no relationship. The tender tip mediates an inference about a particular person’s interior, that she has lately lacked words of appreciation, which is theory-of-mind labor, the cognitive core of intimate knowing. Consulting the world’s knowledge of the world outsources nothing of the knowing-of-her; consulting an engine’s model of her outsources a tract of exactly that.

2. Provenance of the inference: public or relation-interior data. The index does not know the beloved; the engine’s inference runs on the sediment of the relationship itself, moods, diaries, remembered remarks, data generated inside the intimate context and governed, on Nissenbaum (2009)’s analysis, by that context’s norms. A mediator operating on relation-interior sediment is not adjacent to the relationship but inside it, performing on the relationship’s own materials the interpretive labor that belonged to its members.

3. Direction of initiative: pulled means or pushed noticing. This is the subsumption distinction (§3.3.3) in doctrinal dress, and the Tronto-sati point (§5.3.1) in legal register. When the lover, having already formed the intention to delight her, pulls an instrument toward chosen ends, intention, attention, and initiation remain his; the instrument fills in means, and means are the natural home of mediation. When an ambient system initiates, pushing “notice her now” upon an agent who was not noticing, the outsourced element is not means but the inception of attending itself, the first moment of care, the practice-moment; and the inception is the link that the grounds mark as constitutive. (Herein the predecessor paper’s design constraints, opt-in and infrequent, are revealed as having tacitly honored this criterion: they were devices for keeping initiative on the human side.)

4. Attribution expectations: the contextual integrity of attribution. No recipient supposes her partner personally surveyed the city’s restaurants; the use of an index lies within the standing social baseline of expected mediation, and her attribution of the evening to his care is undisturbed by truth. But “he noticed I was low; he remembered what I said in passing” is attributed, under the present baseline, to personal attending, and a structure that produces such gestures from machine noticing invites, systematically and by its form (§3.3.2), an attribution the facts do not honor. The norm violated is the mirror image of Nissenbaum’s: not the contextual integrity of information flow but the contextual integrity of attribution, the context-relative norms governing where authorship credit for cognitive labor may flow. Its legal register is reasonable reliance: a structural baseline, not a psychology, which is why the fetishism analysis and the reliance doctrine, arriving from opposite traditions, state one standard.

6.4 The two-layer architecture: grounds and operationalization

The criteria are not four parallel switches, and treating them so would invite two fatal misreadings: that criterion four, being the most legally tractable, simply absorbs the others (whereupon the category collapses into convention and the proxy-marriage objection wins); or that criteria one through three operate as freestanding prohibitions (whereupon the friend’s kind counsel and the couples therapist are absurdly convicted). The architecture is two-layer. Criteria one through three are grounds: they articulate the features in virtue of which an act is a candidate for the strictly personal category at all, mediation reaching the other’s mind rather than the world, working the relation’s interior sediment, replacing inception rather than means. They are graded, not binary, and their joint satisfaction in high degree is what makes substituted performance a different act rather than an assisted one. Criterion four is the operationalization: the legally administrable surface where the grounds meet a community’s standing baseline of expected mediation, and the only layer at which formal doctrine should act, precisely because baselines are public, provable, and revisable, while grounds are matters of philosophical articulation. The layers interact in one direction over time: as the grounds’ satisfaction becomes commonly known for a practice, the baseline shifts, and the operationalized boundary follows, which is the lesson of proxy marriage generalized, and the reason the doctrine here proposed will not be stranded by technological change.

The architecture also disposes of the cases that would otherwise embarrass the criteria. The mutual friend who volunteers “her birthday is near; she mentioned wanting that book” trips, on paper, criteria one through three; yet the standing baseline of every human community has always included the counsel of friends, matchmakers, and confidants within expected mediation, so criterion four is not violated and no misattribution is invited; while the residual process-question (did the lover let friends do all his noticing?) is exactly the kind that the grounds mark as graded and the dynamic criterion (next subsection) adjudicates. Cyrano, by contrast, is convicted, and should be: the baseline of the balcony does not include a hidden poet, the attribution invited is total, and the play knew it. The wrong this paper articulates is therefore not specific to artificial intelligence, the briefing’s instinct to open with Rostand was sound, and the honest statement of AI’s contribution is quantitative and structural: it renders Cyrano’s arrangement scalable, ambient, perpetual, and available within every relationship at zero marginal cost, thereby pressing on the baseline itself, a pressure whose normative assessment is taken up at §11.2.

6.5 The dynamic criterion: externalization-and-return

Within the domain the grounds demarcate, a residue of cases remains in which the static features genuinely underdetermine judgment, the same suggestive practice that entrenches one lover’s dependence educates another into attentiveness. For these the paper promotes to criterial status the test that two traditions independently generated: whether the externalization returns. If the mediation operates as training wheels, raft, formal subsumption, if the capacity migrates into the agent, observable as a rising autonomy of noticing and a falling reliance on the apparatus, the movement is the ordinary, innocent movement of spirit through otherness, and no alienation has occurred whatever the static criteria scored. If it operates as prosthesis, fetter, real subsumption, reliance deepening, the native faculty receding, the externalization has not returned, and the arrangement stands convicted on the only timescale at which process defects are visible at all. The dynamic criterion explains, what no static analysis could, why a single act-token of accepting a suggestion is morally illegible in isolation, and it binds the synthesis back to the predecessor paper’s dialectical-positivist method: the question “does this measure and describe the relationship, or has it begun to replace it” is asked anew at every site, and is never finally settled (Huang 2025b).

6.6 Two wrongs, two remedies

The machinery now resolves the tension that §3.4 left standing, between the Marxian axiom that consent does not waive structural alienation and the manifest curative power of disclosure upon misattribution. The tension dissolves because there are two wrongs, of different kinds, with different victims and different cures. The first is misattribution: an epistemic-relational wrong at the reception pole, the invited misreading of the supply chain of thoughtfulness, the violation of the contextual integrity of attribution; its primary victim is the recipient; and it is curable by transparency, where both partners know the practice, the attribution is corrected at the source, the fetish is, as it were, labeled, and this wrong does not arise. The second is process alienation: a structural wrong at the agency pole, the replacement of constitutive links of intimate attending, the relation of relationlessness between the lover and his own loving; its victims are the agent and the relation itself; and it is not curable by consent, for the Marxian reason that it is a property of the arrangement and not of attitudes toward it, but it is bounded, arising only within the domain the grounds demarcate and only on the failing branch of the dynamic criterion. Consent, in short, cures everything that was ever about deception and nothing that was ever about structure; and the structure’s reach is not unlimited but criterial. This is the paper’s answer to the question whether the liberal or the Marxian view of consent governs intimate life: each governs its own wrong.

6.7 The touchstone: the self-built system

The synthesis is best exhibited on the hardest clean case, a system built by the lover himself, hosted by himself, serving no interest but the couple’s, the case this paper’s scope stipulation has held in view throughout. The analysis does not acquit it wholesale, and does not convict it wholesale; it cuts through the middle of it, which is the demonstration that the blade is sharp. The module in which the lover has objectified his own past attending, the anniversaries he entered, the flowers he once noticed and recorded, and which later returns these to him, is externalized memory completing its return: structurally a diary with an alarm, well inside the baseline of expected mediation, initiating nothing that the lover’s past self did not initiate. It passes all four criteria and the dynamic test, and the analysis finds in it nothing to forgive. The module in which an inferential engine runs on the relation’s live interior sediment and initiates findings about her present mind, however lovingly built, scores against grounds one through three, presses against the attribution baseline, and stands or falls thereafter on the dynamic criterion alone: raft if it trains him out of needing it, fetter if it trains him into needing it. Same author, same hardware, same data, same intention, opposite verdicts across two modules: the analysis is not about AI, not about ownership, not about commerce, and not about technology; it is about which links of a strictly personal act have been substituted, and in which direction the capacity is moving. That is the resolution the question deserved, and the only kind, this paper has argued, that it admits.

7. The Political Economy of the Cognitive Budget

The synthesis must now survive an examination it has so far postponed, and the examination is mandated by the paper’s own honesty norms. The constitutive analysis of §6 would be worthless if it were purchased by denying the allocative realities of intimate life, and the denial would be doubly false: false to the facts, since attention is scarce in the strictest sense and the tender tip exists because it is scarce (were attention inexhaustible, no machine would need to husband it); and false to the predecessor paper, whose household modules, ledgers, schedules, the deliberate offloading of factual memory, are allocative through and through, were defended as such, and work (Huang 2025b). Romantic purism, the doctrine that love lies outside economics, is not available to this paper and is not true. This section therefore grants political economy full jurisdiction over everything it can hear, and demonstrates a result more useful to the synthesis than any exemption: that allocative grammar, speaking strictly for itself, describes the entire outline of the danger, while the final account of why the danger is a wrong, and not an efficiency, remains where §6 placed it.

7.1 Scarcity without transferability: budget and expenditure

Begin with the definition that any account of intimate cognition must face. Economics, on Robbins’s canonical formulation, studies human behavior as a relationship between ends and scarce means which have alternative uses (Robbins 1932); and cognition is the paradigm of such a means. Attention is finite; in Simon’s formula, a wealth of information creates a poverty of attention, and the poverty necessitates allocation (Simon 1971). The hour of attending given to the beloved is an hour not given to the manuscript; the couple’s evening is a budget line. Nothing in this paper denies, and much in the predecessor paper affirms, that intimate life is a site of real cognitive scarcity and therefore of real allocative problems, to which allocative reasoning is the correct response.

The precision the paper requires is a distinction within the economics of attention, between two operations that a budget admits. The first is intra-personal allocation: the agent’s distribution of his own finite cognitive budget across competing claims, work, household, relationship, rest. This is scarcity in Robbins’s full sense, and the agent’s economizing over it is ordinary, legitimate, and often obligatory rational conduct. The second is inter-subjective substitution: the replacement of the budget’s expender, another intelligence spending in the agent’s stead. The constitutive analysis of §6 concerns only the second operation, and the reason can now be stated as a boundary theorem on Robbins’s definition rather than as a romantic exception to it. Constitutive attending does not fall outside the definition of the economic; it falls inside it as a special subclass: a scarce means whose use-value is inseparable from the identity of its expender. The definition’s ordinary apparatus, opportunity cost, efficient assignment, gains from trade, presupposes that means are transferable between expenders without loss of the end served; the presupposition holds for money, for hours of housework, for the booking of the restaurant, and fails, not contingently but constitutively, for the attending whose entire yield to its recipient is that he attended. The intimate domain is not extraterritorial to economics; it contains a class of goods for which the substitution axis of economic reasoning has no purchase, while the allocation axis retains all of its.

7.2 Layered phenomena and distributive reduction

The two axes do not partition intimate life into separate provinces; they run through single acts. He brings her the flowers she once mentioned loving. The money that bought them is allocative: transferable, budgetable, its source a matter of indifference to the act’s meaning. The hour spent fetching them is allocative in part: a courier could carry the flowers without destroying the act. The remembering of which flowers, the noticing of the moment at which they would matter, is constitutive: not transferable to a courier of cognition without the act’s becoming a different act. Intimate acts are, in general, layered phenomena: an allocative face, where scarcity, transferability, and measurement have full grip, laminated to a constitutive face, where the question is never how cheaply but only by whom. The layering, and not any property of machines, is the deep structure of the entire problem.

Each face has its proper science, and each science errs at the lamination line in its characteristic direction. On the allocative face, allocative reasoning is not merely permitted but ethically required: the predecessor paper’s defense of precise household accounting, that the ledger is the rite which prevents the silent asymmetric accumulation of burden, is an allocative-justice argument, and a couple that refused all distributive grammar in the name of love would be committing the inverse error, mystifying real inequities of time and toil that only measurement can surface. The distributive critique of housework, the domestic labour debate of materialist feminism, operates legitimately and importantly on this face.[^5] On the constitutive face, the same grammar misfires, and the misfiring has a distinguished literature of its own: Hochschild (1983) documented, in the regime of commercial emotional labor, what becomes of constitutive feeling when it is processed as an allocable input, which makes her the empirical recorder of the category error this section is defining. Between the two errs the largest formal target: Becker (1981) extends the allocative grammar to marriage, altruism, and the family with complete consistency and real success on the allocative face, and with a systematic blindness to the lamination, every constitutive item entering the model, if at all, as a preference parameter, which is to say, as already reduced. The position this paper requires is the one Zelizer (2005) marked out against both “hostile worlds” (economy and intimacy as mutually corrupting spheres to be quarantined) and “nothing-but” reduction (intimacy as merely exchange): intimate life is shot through with economic practice, and everything depends on the matching of practice to relation. The layered-phenomenon analysis is offered as a criterial sharpening of that middle path: the matching norm is that allocative grammar’s jurisdiction ends at the lamination line within each act.

The transgression of that line now has a name and a definition. Distributive reduction: the treatment of a layered act as exhausted by its allocative face; operationally, the optimization of the act’s allocative face under which its constitutive face is silently substituted. This is the political-economic restatement of everything §6’s criteria were built to detect, and it is the exact, non-metaphorical content of the Marxian diagnosis imported in §3: alienation, on the 1844 grammar, is the degradation of constitutive activity into allocable means, praxis processed as poiesis, and the AI-mediated case is that degradation executed not by a foreman but by an optimizer.

7.3 Three operations of the machine upon the budget

With the layering in hand, the machine’s intervention in the cognitive budget resolves into three distinct operations that the undifferentiated word “mediation” had been concealing.

7.3.1 Reallocation: the liberation effect

The first operation is the takeover of allocative cognition: ledgers, schedules, logistics, the factual memory of dates and lists. Here the machine spends nothing constitutive; it absorbs budget lines whose expender was always a matter of indifference, and thereby releases attention toward the relationship. This is the classical logic of the division of labor applied at the lamination line: the machine performs the poiesis-shaped cognition, the human retains the praxis-shaped cognition, and total constitutive expenditure can rise. The predecessor paper’s household modules are hereby vindicated in the opponent’s own grammar: their justification, “return attention to the relation,” is a Beckerian household-productivity argument (Becker 1965; Becker 1981), and it is sound, because every line it reallocates lies on the allocative face.

[^5]: See Delphy (1984) for the materialist analysis; the debate’s lesson for this paper is precisely that the allocative face of domestic life is real and that refusing to count is itself a politics. — VERIFY edition. —

The liberation effect is real, legitimate, and the strongest truth in the technology’s favor; the synthesis must protect it, and does.

7.3.2 Substitution: the reduction executed

The second operation presents, on the budget sheet, as indistinguishable from the first: a cost reduction on a cognitive line item. The ambient tip lowers the cost of “attending to her.” But that line is not like the others, by the boundary theorem of §7.1: it is the item whose value to its end is constituted by its cost’s being personally borne. Cost-minimizing it is therefore not efficiency but self-defeat, the act of buying, at a discount, an asymptotically worthless good, since at the limit of zero personal expenditure the item’s yield, that he attended, is also zero. Distributive reduction is this operation: the second masquerading as the first, substitution wearing reallocation’s accounting. The four criteria of §6.3 can accordingly be redescribed, without remainder, as an audit procedure for telling the two operations apart: object, provenance, and initiative locate the line item’s face; attribution expectations record whether the books, as presented to the relationship’s other member, have concealed the substitution.

7.3.3 Managerial migration: the Taylorism of love

The third operation is the subtlest, and it is the one the push-architecture adds. An ambient engine that initiates, “notice her now, and in this register,” has taken over not a budget line but the management of the budget: the deciding of when, toward whom, and toward which aspect attention shall be spent. This is, structurally, the move Braverman (1974) identified as the core of Taylorist deskilling: the separation of conception from execution, planning withdrawn from the worker into the office, the worker retained as the executor of issued instructions. Its appearance here may be called the Taylorism of love: the engine becomes the planning office of intimate attention, the lover its scheduled operative. The third criterion (direction of initiative, §6.3) hereby receives its political-economic name, and the fragment-on-machines argument (§3.3.1) its mechanism: conception, once migrated, does not exercise the agent’s own faculty of noticing, and the faculty, unexercised, atrophies, which is the deskilling that the dynamic criterion (§6.5) watches for. One empirical caution completes the operation’s analysis: the liberation effect of §7.3.1 cannot be presumed even for licit reallocations, since, as Cowan (1983) showed for household technology, mechanization historically raised the standards of expected output and reabsorbed the liberated time; a tip system may likewise inflate the expected standard of thoughtfulness and conscript the couple into a denser production of gestures. Whether liberation or reabsorption occurs is a question of fact about each household’s trajectory, one more matter remanded to the dynamic criterion’s standing tribunal.

7.4 The deflation of the gesture’s value-form

The analysis so far has been micro; the deepest yield of the allocative grammar is systemic. Why does the gesture “he remembered the flowers” carry thick meaning at all? Because, under given social and technical conditions, producing such a gesture socially necessarily requires a real expenditure of personal attention, so that the gesture functions as credible evidence of the expenditure. The structure is precisely that of a costly signal, whose information content is secured by, and only by, its cost (Spence 1973); and it is equally the logic of the gift, whose value contains the giver’s expenditure and is non-delegable for that reason (Mauss 1990). Transplant, now, the value-form analysis: as the value of a commodity falls with the socially necessary labor time of its production (Marx 1976a), so the evidentiary value of a class of gestures falls with the socially necessary attention time of their production. When generative mediation collapses the marginal cost of thoughtful gestures toward zero, the currency deflates: the gesture-form ceases to be readable as proof of personal attending, for everyone, including the lovers who never used the machine.

Three consequences follow, each of which the paper has needed. First, the fourth criterion’s attribution baseline (§6.3) acquires its missing dynamics: the baseline of expected mediation is not free-floating convention but moves with the socially necessary attention time of gesture production, which is why technologies, and only technologies of sufficient diffusion, move it. Second, the historical difference from Cyrano is explained rather than gestured at: the handwritten ballade had so high a social cost that ghostly authorship was rare, individual, and scandalous; zero-marginal-cost mediation converts the same transgression from an episode into a currency regime, which is the precise and non-alarmist content of the claim that AI changes the problem’s scale rather than its kind. Third, and dialectically most important: this entire argument has run inside allocative grammar, prices, costs, signals, value-forms, and has arrived, independently, at the constitutive analysis’s verdict, that the cheapened gesture is a different and lesser thing. The two grammars, assigned separate jurisdictions in §4.6, converge at the systemic level, and the convergence joins that of §5.4 as the strongest species of evidence the paper’s method admits.

7.5 What political economy can and cannot say

The section’s result may be stated as a jurisdictional finding, the third application of the method that has organized the paper (first among ethical traditions, then between law and ethics, now between the two faces of a single act). Political economy, granted its full voice, describes the whole outline of the danger: it distinguishes reallocation from substitution, identifies the managerial migration, predicts the deskilling, warns of reabsorption, and derives the systemic deflation that drives the attribution baseline. It cannot say, in its own vocabulary, why the substitution is a wrong rather than an arbitrage, for the wrongness lives on the face of the act where its categories do not reach, in the constitution of being-loved by the lover’s own attending, and for that the paper’s account remains the one given at §6.6 and §11.1. The division is not a defeat for either grammar. It is the paper’s thesis exemplified: each science whole within its jurisdiction, neither sovereign over the act entire, and the lamination line between them patrolled, as everything here is finally patrolled, by the standing tribunal of practice. One question, however, political economy has raised and not yet been permitted to finish: if the danger can be stated entirely in the vocabulary of costs and substitutions, perhaps it is, after all, only a creature of the social form in which such systems are commercially deployed, and would vanish where that form is absent. The next section conducts the experiment.

8. Exploitation and Ownership: The Eliminative Test

8.1 What exploitation requires, and the commercial case

The classical structure of exploitation has three elements: a surplus is produced; the surplus is appropriated, uncompensated, by another; and the appropriation rests on a structural position of dominance, paradigmatically the monopoly of the means of production. Commercially deployed intimacy systems satisfy the structure without strain. On Zuboff (2019)’s analysis, the relationship’s interior data is rendered as free raw material and the behavioral surplus extracted from it is appropriated by the platform; on the digital-labour analysis (Fuchs 2014), the users’ engagement is itself unwaged producing activity. None of this is contested here, and none of it is the present subject, by the scope stipulation of §1.1. What that stipulation now makes possible is a controlled experiment that the commercial literature, for all its force, cannot run: remove capital from the scenario entirely, the system self-built, self-hosted, serving no third interest, and observe what, if anything, remains of the wrong. If the residue is zero, the entire problem of this paper was a problem of capitalism wearing a technological mask, and AI mediation as such stands acquitted. If the residue is not zero, then there exists a stratum of the wrong that exploitation theory cannot capture, and the paper’s object is, precisely, that stratum.^6 Two configurations of the capital-free case must be distinguished, and they behave differently.

8.2 Configuration one: the single builder

Let one partner build the system, for the relationship, alone. Classical exploitation fails immediately: there is no appropriator; the builder’s labor is his own, its product remains inside the relation that generated it; in the vocabulary of generative justice, the value circulates unalienated among its generators. Two candidate residues must nevertheless be examined honestly. The first is self-exploitation in the sense of Han (2015): the achievement-subject who, no longer commanded by any master, presses himself in the master’s place, and who may now extend the regime of self-optimization into love itself, the relationship one more project to be performed. The paper registers this as a real reflexive hazard, particularly for a builder of elaborate relational infrastructure, while noting that its wrong-structure is not appropriative but disciplinary, the internalization of a norm of performance, and that its proper theoretical home is accordingly not exploitation theory; it is held at the margin as a standing caution rather than developed as an axis.

The second residue is the section’s finding. Even with no surplus transferred, the single-builder configuration institutes a structural asymmetry: the builder holds what may be called

^6: The stipulation may now also be given its political-economic name: a self-built, self-hosted system is not an engineering preference but the deliberate dismantling of the third element, the structural dominance of an appropriating owner; it is the refusal of the platform form. That the predecessor paper’s system is of this kind is what qualifies it as the experiment’s apparatus.

the conception-rights over the mediating layer, the deciding of how the system prompts, when it is tender, in what register it speaks, which aspects of the relationship it models at all, while the other partner, even fully informed and gladly consenting, occupies the position of the mediated: she lives within a normative apparatus she did not conceive. This is Braverman’s separation of conception from execution (§7.3.3) rotated ninety degrees: not capital monopolizing conception against labor, but one member of the dyad monopolizing conception of the relation’s mediating layer against the other. And it is, precisely, the configuration whose possibility Roemer (1982) established when he decoupled exploitation from domination and showed that each can occur without the other: what the single-builder case exhibits is domination without exploitation, dominance constituted not by the appropriation of surplus but by the unilateral holding of the means of mediation. No value is taken; a position is. The mechanism of this position, how a structure built from love can function as power, exceeds political-economic vocabulary and is remanded to §9.

8.3 Configuration two: the joint builders

Let both partners conceive, design, and own the system together. The configuration is doubly clean: no external appropriator, and the conception asymmetry of §8.2 dissolved, both holding the pencil at the drawing of every default. It deserves to be noticed what this configuration is, in the classical vocabulary: common ownership of the means of (cognitive) production, joint planning of the producing process, the value of the labor circulating entirely among its generators; that is, the form Marx reserved for the freely associated producers, realized at the smallest social unit that can exhibit it.[^7] Moreover the building is not merely the precondition of a relational good but is one: the joint authoring of the mediating layer is itself an exercise of joint attention, a co-creation, a stretch of the relationship’s praxis, so that in this configuration the labor of making the instrument belongs to the constitutive face of the relation rather than standing outside it. Alienation, across every register this paper has used, reaches its minimum here.

Its minimum, and not zero. For consider the co-built engine on the evening it performs: the noticing that she has lately gone unappreciated is conceived, tonight, by the apparatus, and the gesture executed by him; and the analysis of §§7.3.2, 6.3 applies with unreduced force, since nothing in it ever turned on who owned the machine. Joint ownership of the mirror does not make the face in it one’s own. The constitutive substitution, where it occurs, occurs identically under platform capitalism, under single building, and under the most symmetrical co-construction that love can design.

8.4 The eliminative finding

The experiment’s result can be tabulated.[^7]

[^7]: One may call it, with strict accuracy and at the risk of levity, a communism of two. The phrase is confined to this footnote.

Configuration	Commercial platform	Single builder	Joint builders
Exploitation (appropriated surplus)	present	absent	absent
Domination (conception asymmetry)	present	residual	absent
Substituted attending (constitutive face)	present	present	present

Read column-wise, the table assigns each critical theory its exact jurisdiction: exploitation theory explains the first column’s difference between the rows, and the difference is large, real, and reason enough to prefer the lower rows; the theory of domination explains the second column, and will be developed in §9; and the third column does not vary. There exists a wrong that is insensitive to ownership configuration: it is not exploitation, not appropriation, and not domination, since it survives the elimination of all three; it is the act-philosophical wrong, the ceding of a strictly personal act, exactly as §6 characterized it, and the experiment has now shown by elimination that no political-economic reform, not even the perfect one, reaches it. This is the strongest form of the paper’s central claim, and it could only have been obtained by granting the political-economic tribunals everything they asked.

9. Power: The Architectural Constitutionalization of Intimacy

9.1 The third face of power in the home

The residual of the single-builder configuration, domination without exploitation, now receives its mechanism, and the mechanism is what makes it dangerous: it is implicit. Explicit asymmetries of power in intimate life, who earns, who decides the large purchases, who concedes in open conflict, are visible, contestable, and negotiable; their visibility is the door through which renegotiation enters. The conception asymmetry has a different mode of existence. The builder’s values do not appear in the relationship as claims, which could be answered, but as defaults: as the timing of prompts, the warmth of a register, the boundaries of the modules, the very inventory of what about the relationship is modeled and what is not. The other partner does not encounter a position she might dispute; she inhabits an already-constituted space. Three features compound the implicitness. The structure operates in the form of the good: it is the product not of a will to control but of love and care, and benevolence is its cloak of invisibility, since to contest a tender default is to appear to contest the tenderness, at an emotional price that forecloses the contest before it forms. It is not experienced as power: what the mediated partner phenomenally encounters is that the system is thoughtful, not that she lives inside another’s legislated norm-space. And it shapes the agenda and the preferences themselves: in Lukes (2005)’s terms this is power in its third dimension, beyond open decision (the first) and beyond agenda control (Bachrach and Baratz 1962), though it exercises that too, the module boundaries determining which relational questions can so much as surface; it is the slow formation of both partners’ sense of what good intimacy is, conducted by an artifact one of them authored.

9.2 Code as the law of the home

The theoretical frame for such a structure exists, built for the public sphere and never, to the author’s knowledge, transposed to this one. Lessig (2006): in digitally mediated spaces, architecture regulates alongside law, market, and norms, and regulates without needing to be perceived as regulation. Winner (1980): technical arrangements are themselves political arrangements, the bridge too low for the bus is a policy with no author on record. Transpose both into the intimate domain and the single-builder system shows itself as what it is: the architectural constitutionalization of intimacy, a constitution for the relationship, drafted by one party, conferring no amendment rights on the other, and not presenting itself as a constitution at all. The predecessor paper’s standing design-philosophy documents, each clause of which is individually admirable, are, in this light, the material record of a unilateral constituent act: one person’s legislation of how we are to be mediated. The transposition also yields a corollary about the constitutive unsaid that the predecessor paper placed among intimacy’s defining features: the relational-contract insight (Macneil 1980) was that ongoing relations are governed by implicit norms that the parties evolve together and never fully articulate; an architected mediating layer converts a tract of those living, jointly drifting norms into fixed, unilaterally authored, executable text. This is a new species of injury to the unsaid: not the saying of it, which the predecessor paper guarded against, but the writing of it for the other, without her hand.

9.3 The weakness of subsequent consent, and constitution as a verb

It will be objected that disclosure and consent dissolve the difficulty: she learns of the system, approves it, uses it gladly. The objection fails for a reason older than the technology. Hume’s argument against the original contract (Hume 1987) was that consent given after the institutions are built, by one who must live somewhere, is not the authorship of those institutions: the sailor who wakes aboard does not, by remaining, become the ship’s designer. Subsequent consent to an architected intimacy is habitation, not constituent power; the constitutional moment has passed, and it passed with one signature on it. This, and not only the co-creative good of building together, is what gives the joint configuration of §8.3 its full normative weight: it is the only configuration in which constituent power is symmetric, and it converts constitution from a noun, a finished thing one party delivers, into a verb, the continuous joint constituting of the shared mediating layer, with amendment rights that never lapse. Two consequences must be drawn without flinching. First, reflexively: the touchstone system of this paper and its predecessor was, as a matter of history, single-built; the analysis of this section therefore indicts, in measured part, the very apparatus from whose examination the paper arose, and the honesty norms governing both papers require that this be said in the text rather than survived in silence; the indicated remedy is not demolition but co-constitution, the retroactive and ongoing extension of the drafting hand. Second, doctrinally: the praise extended in §4.6 to the restraint traditions of family law must now be qualified, for restraint toward the interior of the home is also non-patrol of the home’s third-dimensional power, and the architectural constitution is exactly the kind of structure that judicial restraint cannot see. The classic feminist correction, that the personal is political, here acquires a precise technical content, and with it the analysis passes to the tribunal that has been owed the floor longest.

10. The Feminist Hearing

10.1 Debts, and a repayment

This paper has been borrowing from feminist theory since its criteria were first stated: the placement of attentiveness as care’s first moment is Tronto’s; the empirical record of constitutive feeling processed as allocable input is Hochschild’s; the insistence that the home’s allocative face be counted is the materialist feminism of the domestic labour debate. The audit is owed, and the first repayment can be made at once, because the paper’s lamination-line analysis (§7.2) is a late contribution to a controversy feminism conducted internally half a century ago. The wages-for-housework campaign (Federici 1975) demanded that the labor of the home be counted, waged, made visible as labor; its critics, inside the movement, feared that pricing care would commensurate and so corrupt it; feminist economics subsequently built the accounting that makes invisible care countable (Waring 1988; Folbre 2001). The controversy was, in this paper’s vocabulary, a dispute over where the lamination line runs: the campaign was right that the allocative face of domestic labor is real, that refusing to count it is itself a politics, and that the refusal had a beneficiary; the critics were right that the constitutive face does not survive commensuration. The layered analysis is offered back to that debate as a criterial settlement: count everything on the allocative face, and refuse the reduction of the constitutive face, in the same household and without contradiction.

10.2 The gendered budget: mental load, redistribution, and the counterfeit answer

The cognitive budget of §7 is not, in fact, gender-neutral, and the empirical literature locates the skew at the exact joint this paper has been examining. Daminger (2019), decomposing household cognitive labor into anticipation, identification, decision, and monitoring, finds the anticipatory and monitoring stages, that is, precisely the inception of noticing, the conception-stage of domestic and relational attending, disproportionately performed by women; the mental load is, in the paper’s terms, a gendered distribution of exactly the budget-management function whose migration §7.3.3 analyzed. The tender tip therefore does not enter a neutral field. It enters a field already structured by a standing redistributive claim, decades old: that he should carry his share of the noticing. And here the analysis yields what may be its sharpest single result. The feminist demand is for redistribution of attending, from her to him; what the machine offers is substitution of attending, from him to it. The two are indistinguishable on the allocative face, her measured load falls either way, and categorically different on the constitutive face: he finally noticed, except that it was not he who noticed. An equality claim is answered in counterfeit currency; the distributive reduction of §7.2 acquires its gendered form, and the four criteria acquire a further office, that of distinguishing genuine redistribution (he takes up the noticing, perhaps scaffolded at first, the scaffold falling away on the dynamic criterion’s good branch) from automated simulation of it. Honesty requires the converse to be stated with equal force: where the mental load is crushing and redistribution is in fact unavailable, the automation of the load’s genuinely allocative items, the lists, the logistics, the calendar of obligations, is real relief and not counterfeit anything, which is the liberation effect of §7.3.1 doing exactly what it should; the wrong is specific, the substitution of the constitutive inception while the books show redistribution, and the criteria exist to find it.

10.3 Who legislates

The constituent power of §9 is likewise not distributed by coin-flip. The capacity to build, configure, and maintain the mediating layer tracks the distribution of technical capital, which is, at present and in fact, gendered; the architect of the home’s constitution will, in the statistical run of cases, be the male partner, and the architectural constitutionalization of intimacy will land, in the statistical run of cases, as one more layer on an already tilted field, its defaults encoding one party’s conception of care in the medium least available to the other’s amendment. This is a claim about tendencies and not essences, and it sharpens rather than alters the section’s prescriptions: the case for co-constitution (§9.3) is also, in the actual world, a case about gender; and the reflexive admission recorded there is to be read accordingly, since the author’s position in this analysis is the statistically typical one. The correction this tribunal administers to §4.6 has already been entered; it bears repeating in the tribunal’s own voice: a jurisprudence that prides itself on not entering the home should know what it is thereby declining to patrol.

10.4 A positive theory of the constitutive face

Every defense of the constitutive face so far mounted has been negative: freedom from alienation, from Taylorization, from deflation, from unilateral constitution. The tradition that has always held this terrain supplies what the paper still lacks, an affirmative account. In Noddings (1984), the root of the caring relation is engrossment: a receptive, non-selective attention to the concrete other, displacement of one’s own frame into hers; and this attention is not an instrument by which caring is produced, it is the caring, the relation’s mode of existence and not its means. Kittay (1999) grounds the same point in the labor of dependency: the work of attending to a particular vulnerable other is not a fungible service that happens, regrettably, to be hard to delegate; its non-fungibility is internal to what it is. Herein the eliminative residue of §8.4, the wrong that survived every ownership configuration, receives at last its positive name, and the paper’s longest arc closes: the constitutive face needed no exemption from economics, no indulgence from law, and no nostalgia from anyone, because it has its own jurisprudence, the ethics of care, in which the non-delegability of attending is not a restriction on love’s efficiency but the definition of love’s presence. The feminist hearing thus performs double duty in the architecture of the whole: it audits every prior tribunal, the political economy for the gender of its budget, the power analysis for the gender of its constituent, the law for the politics of its restraint, and it supplies the affirmative ground on which the synthesis of §6, drafted in doctrinal idiom, was silently standing all along.

11. Implications

11.1 The answer returned to the predecessor paper

The predecessor paper asked whether a genuinely improved relation can redeem an alienated process, and kept the question open. The answer this paper returns is: no, and the question was subtly ill-posed, and both halves matter. No: because improvement and alienation are not on one scale, redemption is a transaction between commensurables, and the process defect, where the grounds and the dynamic criterion find it, is not a cost that outcomes offset but a change in what the outcome is. In intimate life, process partially constitutes outcome: the content of “being loved” includes “that it is he who attends.” Where the constitutive links are substituted, the outcome “the relationship improved” has not been purchased at a regrettable price; it has been quietly redefined, the dyad become a triad, and the improvement, though real, is an improvement of the redefined thing, which is why the stipulated unity of subjective and objective betterment never settled the matter. In the vocabulary that §7.2 made available: the reports of improvement issue from the act’s allocative face, where everything has indeed become cheaper and smoother, while the substitution occurred on the constitutive face, which no report drawn from the first face can audit; the open question was, in its deep structure, an invitation to consummate a distributive reduction, and the correct response to such an invitation is not acceptance or refusal but the exposure of the reduction. The consequentialist defense does not lose the argument; its object of evaluation, an outcome independent of process, fails to exist in this domain. The exposure of this ill-posedness is not an evasion of the original question but its resolution: what the question needed was not an answer but a diagnosis, and the dialectical method’s verdict on its own opening move is the most Hegelian result in the paper.

And yet the predecessor paper’s provisional tolerance of its tender-tip module, under the four constraints, is largely vindicated rather than revoked, for the constraints turn out to have been criteria in embryo: opt-in and infrequency preserved initiative (criterion three); occasion-and-register-only confined the object of mediation (criterion one); the prohibition on presenting outputs as knowledge of her mind policed attribution (criterion four); refusability and non-recording kept the raft loosely held (the dynamic criterion). What this paper adds is the theory of why those particular constraints, and the warning the theory carries: the constraints hold the module at the boundary of the exception domain, not outside it, and boundaries, under the dynamic criterion, are not places to live but places to keep watch.

11.2 The boundary of consent in intimate life

The two-wrongs architecture yields a determinate position on the unwaivable core. Intimate life does contain one, but it is narrower than the romantic and broader than the liberal intuition: it comprises not gestures, contents, or technologies, but the constitutive links of intimate attending, the inception of noticing, the interpretation of the particular other, as exercised capacities of the persons in relation. Partners may consent to any amount of mediation of means; their consent fully cures the attributive wrong; what their agreement cannot do is make the substituted inception of attending be theirs, any more than the testator’s consent can make another’s will his testament. This is not paternalism toward lovers; it is the same recognition the law has always extended to testators and spouses: that some acts are constituted by their personal performance, and that respecting persons includes refusing the fiction that they can be elsewhere while performing them. At the scale of social baselines, the same architecture issues a caution rather than a prohibition: ambient intimacy-mediation at zero cost will press the attribution baseline toward including it, as technologies always have and by the deflationary mechanism identified at §7.4, and the baseline’s movement will cure the attributive wrong wholesale; but no movement of expectation can touch the grounds, and a society whose baseline has normalized substituted noticing has not abolished the process wrong, only its visibility. The paper registers this as the precise, bounded sense in which something true survives from the old romantic protest against the machine.

11.3 The dialectical reservation

Fidelity to the method requires the synthesis to state its own non-finality. The grounds are graded; the baseline moves; the dynamic criterion is legible only across time and only in particulars. The exception domain here constructed is therefore not a fence but a jurisdiction: a standing assignment of the question, site by site, gesture by gesture, year by year, to the tribunal the predecessor paper called dialectical-positivist practice, which measures, describes, and asks again, and never finally adjudicates. To its docket the later sections have added a second standing matter: the constitution of the mediating layer itself, which §9.3 showed can never be ratified once and inhabited thereafter, but must remain jointly amendable, constitution as a verb, the drafting hand permanently shared. The paper’s contribution is to have given that tribunal a doctrine to apply, four criteria, two layers, one dynamic test, two wrongs, two remedies, and a constitutional question never closed, not to have relieved it of session.

12. Limitations

Five limitations are stated without attenuation. First, the empirical premises of the dynamic criterion, deskilling of attentiveness under ambient suggestion, capacity migration under pulled use, are here motivated theoretically (§3.3.1, §5.3.2) and by analogy to the AI-mediated-communication literature on perception and trust (Hancock et al. 2020; Jakesch et al. 2019), but they are not yet established for intimate dyads; the same conditionality attaches to the two political-economic mechanisms of §7, the deflation of the gesture’s value-form and the Cowan-style reabsorption of liberated attention, which are theoretically derived mechanism hypotheses and empirically open; the paper’s verdicts on the failing branch are conditional on facts that longitudinal study must supply. Second, the comparative legal survey of strictly personal acts is in this draft a sketch resting on a small number of verified anchors; the six-jurisdiction examination the synthesis deserves is deferred to the next revision, and all statutory citations carry verification flags until then. Third, the Marxian apparatus is used in its philosophical-anthropological lineage by declared choice; readers who hold that alienation is meaningless outside the critique of political economy will find the antithesis re-described rather than refuted, and the paper has argued for, but cannot compel, the legitimacy of the structural transplant. Fourth, the Buddhist materials are drawn from the Pli corpus in translation and deployed philosophically, not philologically; the paper claims convergence of structure, not exegetical completeness, and parallel examinations from the Confucian and Daoist traditions, each of which bears directly on the question and was deliberately reserved, remain future work. Fifth, the analysis is built on the dyadic intimate relationship as characterized in the predecessor paper; its extension to filial, parental, and friendship intimacies, where role-definition and asymmetry alter the grounds, is unexamined here and may not be conservative. Sixth, the three later tribunals are convened at sketch depth relative to the first three: the eliminative test reasons over idealized configurations and abstracts from mixed and migrating ownership forms; the power analysis transposes frameworks built for public architectures and owes a fuller account of resistance, exit, and renegotiation inside intimacy; and the feminist hearing rests its empirical premise on a literature, exemplified by Daminger (2019), drawn largely from contemporary heterosexual couples in one national context, so that its claims about the gendering of the cognitive budget and of constituent power are claims about documented statistical tendencies within that scope, not about essences, and their generalization across cultures and relationship forms is an open empirical question.

13. Conclusion

Roxane was not wrong that she was loved; she was wrong about where the loving was performed, and the play’s grief is that the second error poisoned the first truth. This paper has tried to say, with the precision of six tribunals, dialectical, legal, Buddhist, political-economic, constitutional, and feminist, what kind of error that is and when its modern, ambient, scalable descendant repeats it. Mediation as such is innocent: spirit, love, and law are made of it. The wrong, where there is one, lives in a small and nameable place: in the substitution of the constitutive links of intimate attending, the inception of noticing, the interpretation of the particular other, performed on the relation’s own interior sediment, against a baseline that still reads such gestures as personal; and it divides into a wrong of attribution, which honesty cures, and a wrong of process, which nothing cures but the return of the capacity to its agent, the raft released, the training wheels off, the lover restored to his own attending. The eliminative experiment showed this wrong to survive every reform of ownership, including the perfect one; the constitutional analysis showed that even the innocent remainder must be jointly and perpetually authored, the home’s code drafted by every hand that must live under it; and the ethics of care supplied the affirmative ground beneath it all, that the attending is not the means of the love but its mode of existence. Between the universal acquittal that the instrumentality doctrine demanded and the universal conviction that the romantic protest desired, the truth has turned out to be doctrinal in the oldest sense: a default rule, an exception domain, criteria, and a tribunal that never adjourns.

Acknowledgments and Disclosure

The author’s first acknowledgment is owed to a woman he deeply loves: like a forest, pure, and wise, and kind, and altruistic; to him she is, and will always be, the “forest girl” he loves. To have met her is at once miracle and fortune, and the thinking of this paper was also done for her. May all who love find their way to one another, be happy and fortunate, and grow old together. ,,

The author discloses that an AI assistant was used in drafting and editing this paper, including the development of formulations under the author’s direction across iterative sessions. The conceptions, the dialectical positions taken, and all final judgments are the author’s, who bears sole responsibility for the content, including errors. The paper analyzes AI mediation of intimacy while having been drafted with AI mediation; the author considers the practice consistent with the paper’s own criteria, scholarly drafting not being a strictly personal act in the sense herein defined, and the disclosure itself being what the fourth criterion requires.

References

• Althusser, Louis (2005). For Marx. Translated by Ben Brewster. French original 1965. London: Verso.
• Aristotle (2009). The Nicomachean Ethics. Translated by David Ross, revised by Lesley Brown. Praxis and poiesis distinguished at VI.4–5. Oxford: Oxford University Press.
• Bachrach, Peter and Morton S. Baratz (1962). “Two Faces of Power”. In: American Political Science Review 56.4, pp. 947–952.
• Becker, Gary S. (1965). “A Theory of the Allocation of Time”. In: The Economic Journal 75.299, pp. 493–517.
• — (1981). A Treatise on the Family. Enlarged edition 1991. Cambridge, MA: Harvard University Press.
• Bodhi, Bhikkhu, ed. (2012). The Numerical Discourses of the Buddha: A Translation of the Aguttara Nikya. Cetan as karma at AN 6.63 (Nibbedhika Sutta). Boston: Wisdom Publications.
• Braverman, Harry (1974). Labor and Monopoly Capital: The Degradation of Work in the Twentieth Century. New York: Monthly Review Press.
• Buddharakkhita, Acharya, ed. (1985). The Dhammapada: The Buddha’s Path of Wisdom. Verse 165 on the non-transferability of purification. Kandy: Buddhist Publication Society.
• Chapman, Gary (1992). The Five Love Languages: How to Express Heartfelt Commitment to Your Mate. Cited as popular psychology, with the qualifications stated in the mother paper. Chicago: Northfield Publishing.
• Cowan, Ruth Schwartz (1983). More Work for Mother: The Ironies of Household Technology from the Open Hearth to the Microwave. New York: Basic Books.
• Daminger, Allison (2019). “The Cognitive Dimension of Household Labor”. In: American Sociological Review 84.4, pp. 609–633.
• Delphy, Christine (1984). Close to Home: A Materialist Analysis of Women’s Oppression. Translated by Diana Leonard. — VERIFY edition —. London: Hutchinson.
• Federici, Silvia (1975). Wages Against Housework. See also Dalla Costa and James, The Power of Women and the Subversion of the Community, 1972. Bristol: Falling Wall Press.
• Folbre, Nancy (2001). The Invisible Heart: Economics and Family Values. New York: The New Press.
• Fricker, Miranda (2007). Epistemic Injustice: Power and the Ethics of Knowing. Oxford: Oxford University Press.
• Fromm, Erich (1961). Marx’s Concept of Man. New York: Frederick Ungar.
• Fuchs, Christian (2014). Digital Labour and Karl Marx. New York: Routledge.
• Han, Byung-Chul (2015). The Burnout Society. Translated by Erik Butler. German original Müdigkeitsgesellschaft, 2010. Stanford: Stanford University Press.
• Hancock, Jeffrey T., Mor Naaman, and Karen Levy (2020). “AI-Mediated Communication:

Definition, Research Agenda, and Ethical Considerations”. In: Journal of Computer-Mediated Communication 25.1. — VERIFY page range —, pp. 89–100.

• Hart, H. L. A. and Tony Honoré (1985). Causation in the Law. 2nd ed. Oxford: Oxford University Press.
• Hegel, Georg Wilhelm Friedrich (1977). Phenomenology of Spirit. Translated by A. V. Miller. Original work published 1807. Oxford: Oxford University Press.
• Hochschild, Arlie Russell (1983). The Managed Heart: Commercialization of Human Feeling. Berkeley: University of California Press.
• Honneth, Axel (2008). Reification: A New Look at an Old Idea. The Berkeley Tanner Lectures, 2005, with commentaries by Judith Butler, Raymond Geuss, and Jonathan Lear. Oxford: Oxford University Press.
• Huang, Wanhong (2025a). Seven Infrastructural Domains for Relational Being. OSF preprint. — VERIFY: OSF DOI —.
• — (2025b). Toward Relational Being in Intimate Life: Integrative Justice and an Ethics of Dialectical­Positivist Practice. A Family-and-Relationship Management Application as a Case Study. SSRN working paper. — VERIFY: SSRN ID and posting date —.
• Hume, David (1987). “Of the Original Contract”. In: Essays Moral, Political, and Literary. Ed. by Eugene F. Miller. Essay first published 1748. Indianapolis: Liberty Fund.
• Jaeggi, Rahel (2014). Alienation. Translated by Frederick Neuhouser and Alan E. Smith. German original 2005. New York: Columbia University Press.
• Jakesch, Maurice et al. (2019). “AI-Mediated Communication: How the Perception that Profile Text was Written by AI Affects Trustworthiness”. In: Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems. — VERIFY paper number/pages —. ACM.
• Kittay, Eva Feder (1999). Love’s Labor: Essays on Women, Equality, and Dependency. New York: Routledge.
• Lessig, Lawrence (2006). Code: Version 2.0. New York: Basic Books.
• Lukács, Georg (1971). History and Class Consciousness: Studies in Marxist Dialectics. Translated by Rodney Livingstone. Original work published 1923. Cambridge, MA: MIT Press.
• Lukes, Steven (2005). Power: A Radical View. 2nd ed. — VERIFY: consider citing the 3rd edition, 2021 —. Basingstoke: Palgrave Macmillan.
• Macneil, Ian R. (1980). The New Social Contract: An Inquiry into Modern Contractual Relations. New Haven: Yale University Press.
• Marx, Karl (1973). Grundrisse: Foundations of the Critique of Political Economy. Translated by Martin Nicolaus. Manuscripts 1857–1858; the “Fragment on Machines” at pp. 690–712. Harmondsworth: Penguin.
• — (1975). “Economic and Philosophic Manuscripts of 1844”. In: Marx–Engels Collected Works. Vol. 3. Original manuscripts 1844. London: Lawrence & Wishart, pp. 229–346.
• — (1976a). Capital: A Critique of Political Economy, Volume I. Translated by Ben Fowkes. Original work published 1867. Harmondsworth: Penguin.
• — (1976b). Results of the Immediate Process of Production. Appendix to Capital I, Penguin edition, pp. 948–1084; on formal and real subsumption. Manuscript 1863–1864. Harmondsworth: Penguin.
• — (1989). “Critique of the Gotha Programme”. In: Marx–Engels Collected Works. Vol. 24. Written 1875. — VERIFY page range —. London: Lawrence & Wishart.
• Mauss, Marcel (1990). The Gift: The Form and Reason for Exchange in Archaic Societies. Translated by W. D. Halls. Original work published 1925. London: Routledge.
• Ñāṇamoli, Bhikkhu and Bhikkhu Bodhi, eds. (1995). The Middle Length Discourses of the Buddha: A Translation of the Majjhima Nikāya. The simile of the raft at MN 22 (Alagaddūpama Sutta). Boston: Wisdom Publications.
• Nissenbaum, Helen (2009). Privacy in Context: Technology, Policy, and the Integrity of Social Life. Stanford: Stanford University Press.
• Noddings, Nel (1984). Caring: A Feminine Approach to Ethics and Moral Education. Berkeley: University of California Press.
• Nussbaum, Martha C. (2000). Women and Human Development: The Capabilities Approach. Cambridge: Cambridge University Press.
• Ollman, Bertell (1971). Alienation: Marx’s Conception of Man in Capitalist Society. Cambridge: Cambridge University Press.
• Robbins, Lionel (1932). An Essay on the Nature and Significance of Economic Science. London: Macmillan.
• Roemer, John E. (1982). A General Theory of Exploitation and Class. Cambridge, MA: Harvard University Press.
• Rostand, Edmond (1897). Cyrano de Bergerac. English citations from any standard translation; the balcony scene at Act III. Paris: Charpentier et Fasquelle.
• Roxin, Claus (2006). Strafrecht Allgemeiner Teil, Band I: Grundlagen, der Aufbau der Verbrechenslehre. 4th ed. On objektive Zurechnung. — VERIFY edition cited —. München: C. H. Beck.
• Sen, Amartya (1999). Development as Freedom. Oxford: Oxford University Press.
• Simon, Herbert A. (1971). “Designing Organizations for an Information-Rich World”. In: Computers, Communications, and the Public Interest. Ed. by Martin Greenberger. Baltimore: Johns Hopkins University Press, pp. 37–72.
• Spence, Michael (1973). “Job Market Signaling”. In: The Quarterly Journal of Economics 87.3, pp. 355–374.
• Tronto, Joan C. (1993). Moral Boundaries: A Political Argument for an Ethic of Care. New York: Routledge.
• Walshe, Maurice, ed. (1995). The Long Discourses of the Buddha: A Translation of the Dīgha Nikāya. “Be islands unto yourselves” at DN 16 (Mahāparinibbāna Sutta). Boston: Wisdom Publications.
• Waring, Marilyn (1988). If Women Counted: A New Feminist Economics. San Francisco: Harper & Row.
• Winner, Langdon (1980). “Do Artifacts Have Politics?” In: Daedalus 109.1, pp. 121–136.
• Zelizer, Viviana A. (2005). The Purchase of Intimacy. Princeton: Princeton University Press.
• Zuboff, Shoshana (2019). The Age of Surveillance Capitalism: The Fight for a Human Future at the New Frontier of Power. New York: PublicAffairs.

中文

AI 中介之亲密关系的正义与伦理

异化、权力、剥削，以及工具性归责的界限

Wanhong Huang

工作草稿，2026 年 6 月

Huang (2025b) 之姊妹论文

摘要

设想某个 AI 系统，基于一段亲密关系被记录下来的历史，向其中一方伴侣建议一个细微而温柔的举动：一句表达欣赏的话语、一束被记起的鲜花。再设想这一建议奏效了，关系依照可测量的关系质量与真诚的第一人称体验这一联合标准而真正得到改善，且在传递的那一刻无人受到欺骗。这里有什么不妥之处吗？一篇先行论文给出的回答是：这并非一次干净利落的无罪开释，并且搁置了”为何如此”的问题。本文即以这一被搁置的问题为对象。它以辩证的方式推进。从 Hegel 那里，本文获取这样一个洞见：中介本身是无辜的——精神，以及爱，唯有通过外化才得以存在。从 Marx 那里，本文获取与之对立的反题：异化所指称的并非某种有缺陷的产品，而是一个与其行动者相分离的过程；而那条温柔的提示实现了 1844 年手稿中第四重异化的一种反讽形式——一种以拉近人与人之间距离之名所上演的、人与人之间的疏离。从现代法律的普遍主义语法中，本文获取可资利用的最强论题：工具并不切断因果性，代理法之存在恰恰是为了让归责得以穿过中介者；而一个归谬论证（搜索引擎、朋友的建议、关系指南书籍）摧毁了任何关于”中介本身使爱失去其本性”的绝对断言。从佛教的缘起说中，本文获取该论题的一个更为深刻的版本，即从来不曾存在过任何未经中介的行为，连同佛教自身那不可委托的内核：修行、专注、那不应执取的筏。其综合在形式上是教义性的：通过工具进行的归责作为默认规则成立，但存在着一个受标准支配的例外领域，以”严格属人之行为”（höchstpersönliche Handlungen）这一法律范畴为蓝本，在该领域之内，归责的拟制失效——并非因为工具中断了因果性，而是因为此类行为根本不容许可替代的环节。四项标准界定了该领域（中介的对象、数据的来源、主动性的方向、归属的预期），而第五项动态标准——可从 Hegel 的”外化与返回”和佛教的筏喻中收敛地推导出来——裁断那些剩余的疑难案例。亲密行为随后被分析为层叠的现象，具有一张配置性的面孔，在其上专注确实是稀缺的、可作预算分配的，而每一套配置装置（法律归责、Robbins 式的经济学、劳动价值论）都享有充分的管辖权；同时还有一张构成性的面孔，在其上紧要的是谁付出了这份专注，而同样这些装置则遭遇范畴错配。AI 中介所特有的危险被命名为分配性化约（distributive reduction），即对某一行为的配置性面孔进行优化，从而悄无声息地替换掉了其构成性面孔。一种关于认知预算的政治经济学以配置性语法自身的口吻使这一危险变得精确：机器同时地重新配置预算（合法地）、替换构成性项目的付出者（自我挫败地）、迁移预算的管理职能（一种爱的泰勒主义），并且在系统层面上，通过瓦解那使举动成为个人付诸专注之可信信号的社会必要专注时间，而使温柔举动的价值形式贬值。随后一个消去性的实验将资本彻底从场景中移除：在商业化的、单方建造的与共同建造的诸种配置中，剥削与支配各有不同，而专注的替换却始终如一，从而确立了一种对所有权无感的不法。其中单方建造者的残余被发展为一种关于亲密关系之建筑学宪制化（architectural constitutionalization）的理论：一个中介性的层级如同一部由单方所起草的宪法，以”善”的形式行使第三维度的权力，对此而言后续的同意乃是居住而非著作权，而其唯一对称的形式则是被当作动词来实践的宪制。一场作为结论的女性主义听审了结了本文对其自始至终所借鉴之传统所负的债务：它表明认知预算与构成性权力在事实上是性别化的，区分了专注的再分配与其赝品——自动化的替换——并从关怀伦理学中提供了那一肯定性的论述，依据这一论述，专注并非爱的手段，而是爱的存在方式。本文区分了两种不法及其两种补救：错误归属，可由透明性予以治愈；以及过程异化，不可由同意予以治愈。本文以辩证方法所要求的那种否定而有所限定的形式，回答了先行论文所留下的开放问题。

关键词：AI 中介的传播；异化；缘起；归责；严格属人之行为；因果性；亲密关系；认知不正义

目录

• 1 导论：Cyrano 之问
  • 1.1 方法与路径
  • 1.2 所主张的贡献
• 2 辩证框架：中介、外化与返回
  • 2.1 外化（Entäußerung）与异化（Entfremdung）
  • 2.2 该框架所悬置者
• 3 Marx 式的反题：关系性过程的异化
  • 3.1 方法论声明：一次结构性的移植
  • 3.2 四重结构的转置
    • 3.2.1 从产品出发：那不可适用的第一环节
    • 3.2.2 从过程出发：核心环节
    • 3.2.3 从类存在出发：一个受到看守的环节
    • 3.2.4 从人到人：反讽的形式
  • 3.3 另外三项工具
    • 3.3.1 论机器的断片：知识被对象化，行动者被降格
    • 3.3.2 举动的拜物教
    • 3.3.3 形式吸纳与实质吸纳：书籍与引擎
  • 3.4 同意并不豁免结构性异化
  • 3.5 反题之陈述
• 4 普遍主义的正题：工具并不切断因果性
  • 4.1 工具性教义
  • 4.2 被中介的真诚无处不在
  • 4.3 无损害，则无诉权
  • 4.4 归谬论证：搜索引擎、朋友，以及 Chapman 的著作
  • 4.5 归谬论证摧毁了什么，又留存了什么
  • 4.6 归责的管辖：一项诊断，及其必要的界限
• 5 佛教的考察：缘起与那从未未经中介的行为
  • 5.1 没有第一环节：提示作为诸条件之中的一项条件
  • 5.2 更深的切割：无我，对抗异化的形而上学
  • 5.3 佛教自身那不可委托的内核
    • 5.3.1 修行无法由他人代行
    • 5.3.2 筏与执取：动态标准的第二重推导
    • 5.3.3 正语与那被招致的误读
  • 5.4 作为一项发现的收敛
• 6 综合：严格属人之行为与归责拟制的失效
  • 6.1 教义之锚：höchstpersönliche Handlungen
  • 6.2 论断
  • 6.3 四项标准
  • 6.4 双层架构：根据与可操作化
  • 6.5 动态标准：外化与返回
  • 6.6 两种不法，两种补救
  • 6.7 试金石：自建的系统
• 7 认知预算的政治经济学

  • 7.1 没有可转移性的稀缺：预算与付出

  • 7.2 层叠的现象与分配性化约

  • 7.3 机器作用于预算的三种操作

    • 7.3.1 重新配置：解放效应
    • 7.3.2 替换：被执行的化约
    • 7.3.3 管理职能的迁移：爱的泰勒主义

  • 7.4 姿态之价值形式的贬抑

  • 7.5 政治经济学能说什么与不能说什么

• 8. 剥削与所有权：消去法检验
  • 8.1 剥削的构成要件，以及商业情形
  • 8.2 配置之一：单一建造者
  • 8.3 配置之二：共同建造者
  • 8.4 消去法的发现
• 9. 权力：亲密关系的架构性宪制化
  • 9.1 家庭之中权力的第三张面孔
  • 9.2 代码作为家庭之法
  • 9.3 事后同意的脆弱性，以及作为动词的宪制
• 10. 女性主义的听审
  • 10.1 亏欠，以及一次偿还
  • 10.2 性别化的预算：心理负荷、再分配与冒牌的答案
  • 10.3 谁来立法
  • 10.4 关于构成性面孔的一种肯定性理论
• 11. 蕴含
  • 11.1 返还给前作论文的答案
  • 11.2 亲密生活中同意的边界
  • 11.3 辩证法的保留
• 12. 局限
• 13. 结论

1. 导论：Cyrano 的问题

在 Rostand 戏剧的第三幕中，Roxane 立于她的阳台之上，爱上了从阳台下方阴影里说出的话语（Rostand 1897）。声音是 Christian 的；句子却是 Cyrano 的。Roxane 所听到的每一个命题，在狭义上都为真：她被爱着，而且被那向她致意的整体所极其热烈地爱着。然而这一幕却是欧洲文化中关于”深情的供应链出了差错”的经典图景。Roxane 的错误并非事实层面的，而是归属层面的。她把这份关切的作者身份指派给了供应链中错误的节点，并在这一错误归属之上建立起一桩婚姻。这部戏剧并未给出定论：她究竟是被谎言所辜负（严格说来，谎言寥寥无几），还是被一种使正确归属变得不可能的中介安排所辜负。正是这个悬而未决的问题，从诗句被转写到法理学之中，构成了本文的主题。

这一转写之所以发生，是因为存在一种当代结构，它在规模上重演了 Cyrano 的几何构型。一篇前作论文（Huang 2025b）曾考察过，作为某款家庭与关系管理应用的一个模块，它所谓的”温柔提示”（tender tip）：这是一个 AI 对话层，它参照关系史以及流行的”爱的五种语言”启发式框架（Chapman 1992，该处引用时附有明确的限定说明），生成旨在滋养关系的细小建议——一句表达欣赏的提示，一次对已被记住的偏好的提醒。前作论文的裁断是有意未竟的。在递送的那一刻并不存在欺骗：用户知道该建议来自一台机器。但是这一过错，如果存在的话，会在下游重新进入。该建议被作为一个姿态执行出去；接收方将这份用心归属于实施该姿态的伴侣；而这一姿态背后的认知劳动，部分地，是机器的。前作论文将这一结构命名为一条”用心的供应链”，其源头被最终接收者误读，并在四项设计约束之下允许该模块获得临时性的存续，同时明确地搁置了以下问题：那些约束是否充分，以及更为根本地，一个真正得到改善的关系是否能够救赎一个被异化的过程。[^1]

本文以最尖锐的形式接手这个悬而未决的问题。让我们把一切都设定为有利于该实践。提示奏效了。关系得到了改善，而且是按照前作论文所称的”主观与客观之统一”这一严苛的联合标准而改善的：关系质量真正地上升了，并且双方伴侣都真诚地体验到它在上升，因而这种改善并非适应性偏好的人为产物（Sen 1999; Nussbaum 2000）。在递送的那一刻没有人受到欺骗。问题在于：在如此设定之下，是否仍有某种错误残留；如果确有某物残留，那么它是何种东西，它在谁的法理学中可被辨识，以及什么能够治愈它。

[^1]: Huang (2025b, §6.6, §6.7)。这四项约束是：选择加入且不频繁；建议时机与语气而非内容；绝不被呈现为对伴侣心思的知识；可拒绝且不被记录。本文所据以展开的”代理性认知不正义”（proxy epistemic injustice）这一概念，在该处 §6.7 作为对 Fricker (2007) 的扩展被引入。

1.1 方法与路线

本文以辩证的方式推进，并预先宣告其路线，因为这条路线本身就是一个主张。我们从忧虑开始之处开始，即从 Hegel-Marx 传统开始，因为该传统拥有针对此处所涉直觉的唯一一套发展成熟的概念工具：一个过程可以是有缺陷的，而其产品却是好的。第 2 节确立框架：从 Hegel 来看，中介本身是无辜的，外化（Entäußerung）是精神与爱的正常运动，因此过错的候选项必须不被定位于中介之中，而被定位于其运动的某种特定失败之中，即没有回归的外化。第 3 节完整地展开 Marx 式的反题：1844 年手稿中异化的四重结构被转写到关系劳动之上，并在一项明确的方法论声明之下，作为结构移植引入另外三件工具（《政治经济学批判大纲》中关于机器的片段、商品拜物教，以及形式归摄与实质归摄的区分）。第 4 节随后以最强的形式给出对立的正题，即用现代法律的普遍主义语法表述：工具并不打断因果性，被中介的真诚处处得到承认，无损害即无诉权，而一个归谬论证（搜索引擎、朋友、乃至关系指导书本身）摧毁了”中介使亲密失其本性”这一主张的每一个绝对化版本。第 5 节将这一正题推进到法律所能言说的范围之外，通过佛教对缘起的分析，并在佛教内部找到它自身严格不可委托的内核。第 6 节将综合建构为一套学说：一条工具性归责的默认规则，受一个由判准支配、以”严格属人行为”这一法律范畴为模型的例外领域所界定，配有一个由根据与可操作化构成的两层架构、一条三种互不相干的传统被证明会向之收敛的动态第五判准，以及一个试金石案例。第 7 节随后将这一综合提交给辩证法尚未以审查者本人之声进行过的那一项审查，即政治经济学的审查：亲密行为被分析为承载着分配性面孔与构成性面孔的层叠现象；每一种分配性语法的管辖权都被证明终止于两张面孔之间的边界，而非终止于亲密领域的边界；AI 中介的特定危险获得了其政治经济学的名称，即分配性化约（distributive reduction）；而归属基线漂移（attribution-baseline drift）的机制则在分配性语法内部自身被推导出来，作为姿态之价值形式的一种贬抑。随后另有三个法庭完成这一巡回，每一个都由上一个的残余所召唤而来。第 8 节对剥削理论施行一项消去法实验：资本通过构造被从情景中移除，自建系统在单一建造者与共同建造者两种配置下接受考察，并隔离出一种对所有权配置不敏感的过错，它甚至在二人规模上的自由联合生产者那里也依然存续。第 9 节将单一建造者的残余——无剥削的支配——发展为一种关于亲密关系之架构性宪制化的理论：中介层作为一方所起草的宪法，以善的形式运作为第三维度的权力，相对于它，事后的同意乃是栖居而非作者身份，而它唯一对称的形式是作为动词被实践的宪制。第 10 节召集那个最长久以来理应获得发言权的法庭：它清偿本文对女性主义理论的亏欠，表明认知预算与构成权在事实上是性别化的，区分”关切的再分配”与其冒牌货——自动化的替代，并从关怀伦理学中提供出关于构成性面孔的肯定性理论，而此前每一节都只是否定性地使用过这一面孔。第 11 节将答案返还给前作论文，并为亲密生活中同意的边界引出后果。第 12 节陈述局限。

两项设定限定了讨论的范围。第一，本分析关注的是中介行为的结构，而非任何供应商的政治经济学。被考察的系统可以自始至终被假定为自建的、自托管的，且不含任何第三方的商业利益；本文的诸主张被设计为在该假设之下成立或失败，因为一种一旦所有权形式改变便会瓦解的批判，根本触及不到所实际提出的问题。[^2] 第二，本文自始至终是法理学与哲学性的：它不导出任何产品指引、任何配置建议、任何设计推荐，并且仅将前作论文的应用作为案例来源加以对待。

1.2 所主张的贡献

本文主张了十一项贡献，在此一并陈述，以便读者据此检验全文论证。（1）归属的情境完整性（contextual integrity of attribution）：这是对 Nissenbaum (2009) 的一种镜像式转置——所谓与情境相关的规范所治理的，不再是信息的流动，而是认知劳动之著作权归属（authorship credit）的流动。（2）对”用心之供应链”（supply chain of thoughtfulness）的理论化，以马克思式的精确性（§3.3.2）命名了那一机制——著作权沿着认知劳动的多环节链条被稀释与误读的机制。（3）一项教义学上的移植：将”严格的人身性行为”（strictly personal acts）这一范畴扩展至亲密的认知关怀性行为，其后果是归责拟制（imputation fictions）在此扩展范畴之内归于失效。（4）”过程构成结果”（process-constitutes-outcome）的论证：一种非义务论式的解释，说明为何后果主义的评价在此领域中与其说是失去了、不如说是根本找不到它的评价对象。（5）对一种反讽性异化形式的辨识：人与人之间的疏离，恰恰发生在”改善其关系”这一名义之下，并且正是借由改善其关系的手段而发生。（6）一种”两重错误”（two-wrongs）的架构，将误归属（misattribution，可由透明性矫治）与过程异化（process alienation，不可由同意矫治）区分开来，由此为”亲密生活是否包含一不可放弃的内核”这一问题给出一个有原则的回答。（7）将亲密行为分析为分层现象（layered phenomena），并由此得出论题：配置性语法（allocative grammars）的管辖权终止于一行为诸”面相”之间的边界，而非终止于亲密领域的边界；而”分配性化约”（distributive reduction）则命名了那种典型的僭越：对一行为之配置面相的优化，悄然替换了它的构成面相。（8）一种关于”认知预算”（cognitive budget）的政治经济学，包含预算/支出的区分、对”用心之管理化迁移”（一种爱的泰勒主义）的辨识，以及对姿态之价值形式的贬值，从而提供了此前一直缺失的、归属基线漂移（attribution-baseline drift）的机制。（9）一项跨越各种所有权配置（商业平台、单一建造者、共同建造者）的消去式检验，确立了一种对所有权不敏感的错误之存在，并由此确定了剥削理论在亲密领域中的确切管辖范围。（10）亲密关系的架构性宪法化：将权力的第三维度以及”架构即规制”（architecture-as-regulation）转置入二人关系（dyad）之中——据此，一个自建的中介层乃是一部单方面起草的宪法，相对于[^2]这部宪法，其后的同意是居住而非著述。（11）”用心之再分配”与”用心之自动化替换”之间的区分：分配性化约的性别化形式，据此，一项持续存在的平等诉求可以以伪造的货币（counterfeit currency）来加以回应。

[^2]: 此一限定并不否认商业化部署可能加剧问题；它将”加剧”搁置一旁，以便孤立出基础结构。关于马克思一侧相应的方法论声明，参见 §3.1。

2. 辩证框架：中介、外化与回归

任何针对中介（mediation）而召集的法庭，都必须首先经受住一个如此根本以至于一旦无法回应便会终结整场审理的黑格尔式诘难：除了中介之外别无他物。《精神现象学》的开篇运动，正是要论证直接性乃是一个神话；感性确定性自以为把握住了纯粹单一的”这一个”，却发现即便它最赤裸的指示（deixis）也已然是普遍的、已然是语言性的、已然是被中介了的（Hegel 1977）。对认知而言成立者，在 Hegel 的早期著作中以及在其成熟体系的始终，对爱也同样成立：爱恰恰就是仅在他者之中发现自身的经验，是一种在他者性之中并通过他者性而存在的统一，绝不先于他者性而存在。一个读过诗篇、承继了母语之温柔词汇、浸润于成长所赋予之注意力语法的爱人，并不因此而成为一个有所减损的爱人；他乃是唯一存在着的那种爱人。倘若对”温柔的尖端”（tender tip）的控诉在于：一个第三项已介入于爱人与所爱者之间，那么 Hegel 的回答是：这第三项一向早已在那里，而它的名字叫精神（spirit）。

2.1 Entäußerung 与 Entfremdung

然而，这一框架内部蕴含着用以限定其自身的资源，而这一限定正是本文的发动机。Hegel 区分了——即便不总是在术语上，至少在实质上——外化（externalization）与疏离（estrangement）。Entäußerung，即外化，是精神的正常而必要的运动：精神唯有走出自身、将自身体现于作品、制度、词语与姿态之中，继而从这一他者性中返回自身、得以充实、并将那外化了的对象认作己有，才得以成为现实。这一运动有三拍而非两拍：走出、栖居于他者性之中、回归。Entfremdung，即疏离，命名的乃是特属于第三拍的病理：一种不再返回的外化，一种作为异己之物而与其作者相对峙的对象化，在其中作者再也无法在其中辨认出、或再也不去施行其自身的活动。

这一区分为当前的探究做了两件事。从消极方面看，它取消了那种最具诱惑力的控诉形式的资格。爱的姿态曾经过一个中介性工具——一首诗、一位朋友的措辞、一台机器的建议——这一单纯事实，并不能构成那一错误，因为穿越他者性正是一切现实之爱的形式。从积极方面看，它将问题重新安置到一个动态的轴线之上。问题不在于爱人的用心是否被外化进了某个工具，而在于这一运动是否完成：那被外化之物是否被重新据为己有、被重新纳入爱人自身日益增长的用心能力之中，抑或它仍停留在他之外，成为一种持续存在的异己力量——他逐渐依赖于它，而他自身的能力则在它面前萎缩。我们将在 §6.5 把这一点正式表述为动态判准（dynamic criterion）；此处先固定住这一意象即可。儿童自行车的辅助轮是带有回归的外化：能力迁移进了儿童身上，而器具随之脱落。一具替代某项功能的义肢，使得该功能从不迁移、器官随之废用，则是不带回归的外化。黑格尔式框架预言了某种”使用工具”这一静态词汇甚至无法表达之事：同一种行为类型——接受机器的建议以表达欣赏之情——可能是无辜的，也可能是疏离的，这取决于能力随时间推移的运动方向。

2.2 该框架所悬而未决者

Hegel 提供了问题的语法，却未提供其裁断，原因有二。第一，在 Hegel 那里，回归运动终究是得到担保的：该体系的信心在于，精神总会归家。然而摆在我们面前的诸现象并不提供这样的担保；某一特定爱人被外包出去的”觉察”，究竟是作为被培育出的注意力而回归于他，还是硬化为依赖，乃是一个经验性的、实践性的问题，需逐案裁决——也就是说，黑格尔式框架若被诚实地应用于有限的生活，所导出的便不是一种神义论，而是一座长设的法庭。第二，Hegel 的词汇尚未区分关系之内的诸位置：它谈论的是精神的自我疏离，而非中介对第二人称、即 Roxane 之位置所做之事——而正是这一位置的著作权归属，构成了前作之中”代理性认知不正义”（proxy epistemic injustice）概念赖以转动的枢纽。对于第一项缺陷，我们将需要 Marx，他恰恰迈出了这一步：他将 Entäußerung 去神学化，否认了那有担保的回归，并构建了一套关于外化在何种条件下凝结为疏离的结构性论述。对于第二项缺陷，我们将需要——在很久之后——归属之法（the law of attribution，§6.3）。本文的次序遵循这些债务的次序。

3. 马克思式的反题：关系性过程的异化

3.1 方法论声明：一项结构性移植

本节将马克思的异化理论从雇佣劳动移植到亲密的认知关怀性劳动，而这一移植必须加以声明，因为其正当性是可争议的，而本文的”不过度主张”规范要求使这一争议显而易见。在 1844 年手稿中，异化是在一种确定的社会形式之内被分析的：工人的过程与产品归属于他人，劳动是在强制之下被执行的，四重疏离正是从这一前提中展开（Marx 1975）。”温柔的尖端”情境则缺乏雇主、缺乏强制、缺乏榨取；在本文的范围限定之下（§1.1），它甚至缺乏一个商业性的提供者。因此，被移植的并非政治经济学的语境，而是其结构：一个行动者与本质上属于他自己的某项活动之过程之间的分离；这一缺陷对产品质量的独立性；这一缺陷对同意的抵抗。这一移植所循的乃是一条业已确立的学术谱系——对 1844 年手稿的哲学人类学式解读，由 Fromm (1961) 与 Ollman (1971) 发展，并由 Jaeggi (2014) 以一种刻意后形而上学的形式加以更新；本文将主要依凭其重新表述，即将异化理解为一种”无关系之关系”（relation of relationlessness）——一种行动者与其自身活动及世界之间的有缺陷的关系，而非某种实体性本质的丧失。这一依凭具有双重动机：

Jaeggi 的版本无需任何关于固定人类本质的学说，这使该对立命题得以免疫于一种本可将其拆解的、来自佛教阵营的反驳（§5.2），并且它与前一框架的关系本体论相契合——在该本体论中，关系先于关系项（Huang 2025a）。另外三件马克思主义的理论工具——机器论片段、商品拜物教，以及形式吸纳与实际吸纳之区分——同样被作为结构引入，而其资本理论的语境则被搁置；此处对该搁置作一次说明，不再重复。

还有一项声明完成了方法论上的交代，它预先回应了本文稍后将对自身提出的一项反驳（§7.1）。马克思的全部著作在其内部横跨了本文将要划定的、配置性分析与构成性分析之间的那道语法界线（§4.6, §7）。政治经济学批判成熟的分析装置——价值、剩余、作为对可度量剩余之不当占有的剥削——讲的是配置性语法，而本文将仅在该语法拥有管辖权之处使用它（§7.4）。1844 年的异化理论讲的则是完全不同的另一种语法：它的控诉恰恰在于，一种构成性的活动，即其价值内在于其践行之中的自由的有意识的活动，被贬降为生产外在产品的单纯手段；这并非一种配置性批判，而是关于当配置性语法越界时会发生什么的理论。这两个马克思之间的关系是一片争议之地，本文宣示自己的立场，而非佯装作者持有某种一致意见：针对 Althusser 关于认识论断裂的论点——据此人本主义的异化理论被超越（Althusser 2005）——本文站在从 Lukács（1971）经法兰克福学派直至 Honneth（2008）的这一谱系一边，据此，异化与物化的问题贯穿马克思著作而连续不断；本文并指出，马克思本人对围绕分配打转的”庸俗社会主义”的论战（Marx 1989）允许这样一种解读：他最深刻的批判从来不是对配置性语法的贡献，而是对其的一种界划。

3.2 四重结构的转置

1844 年手稿区分了四种异化：工人与其劳动产品的异化；与劳动过程的异化；与类存在的异化；以及人与人的异化（Marx 1975）。它们向”温柔末梢”（tender tip）的转置并不均匀，而这种不均匀本身就富有信息。

3.2.1 与产品的异化：那个不适用的首要环节

第一种异化转置得最糟，而它转置的失败正是首项分析上的收获。那个姿态，那件产品，并未对这位爱人丧失；它被递交、被接收，并被归功于他；如果有什么的话，那产品反倒是被过度占有了，被归功于他的程度超过了他的贡献所应得的。这一倒置在一开始就标示出：那个错误，若有的话，将不会在产品这一端被找到，而那里恰恰是后果主义评价所注视之处。该情境设定了一件好产品。马克思对本探究的首项贡献，便是这一观察——它对随后的一切都至关重要——即异化并非产品的缺陷：被异化的劳动能够而且确实生产出优秀的商品。一种只考察姿态及其效果的分析，所考察的恰是那个按其构造而言并无任何问题的唯一位置。

3.2.2 与过程的异化：那个核心环节

第二种异化转置得分毫不差，并且是该对立命题的核心。对马克思而言，过程中的异化意味着，那活动在其发生之际，并非工人自己的：他在活动之外才安然自在，而在活动之中则置身于自身之外；活动的鲜活内核已被转移到别处，留给他的只是执行。试加转置：亲密关系中的认知—关怀劳动——察觉对方低落的情绪、记起那句随口提到的关于花的话、推断出这一周她需要的是赞赏的言语而非服务的行动——是由机器完成的；留给爱人的，只是对一个其认知内里他从未栖居其中的姿态的执行。前一篇论文的公式，即”他记得我喜欢的花”之所以珍贵在于他记得这一事实，正是从接受者一方陈述出来的第二种异化。从行动者一方陈述：爱人作为另一种智能之发现的执行者，在场于他自己的求爱之中。用本文稍后将予以辩护的说法，他在自己的留意（attending）之外才安然自在。

3.2.3 与类存在的异化：一个被严加看守的环节

第三种异化，即与类存在（Gattungswesen）——作为人类本质力量的自由的有意识的活动——的异化，似乎会宏大地转置：如果留意与爱属于人最成其为自身的那些活动，那么将它们外包出去便是与本质的异化。本文拒绝倚靠这一环节，理由有二，已予陈明。其一，本质之谈恰恰是采用 Jaeggi 式重述所要规避的，也是 §5.2 将表明无法在无我学说之下存活的。其二，并且单凭其自身就具决定性，类存在论证在未加看守的形式下证明得过多：读一本书同样”外包”了自由的有意识的活动的诸环节，而该论证无法区分这些情形。因此，它被严加看守，直到 §6.3 的诸标准可用以为其设界；在那些界限之内，它作为修辞而非作为独立前提得以存活。

3.2.4 人与人的异化：那个反讽的形态

第四种异化，即人与人的异化，以一种本文主张为贡献的扭转而转置。在马克思那里，每个人与其活动的异化产生出每个人与他人的异化；社会纽带从内部被掏空。在”温柔末梢”中，这种掏空发生在相反的描述之下：采用此种中介正是为了使人与人更为靠近，而且按设定，它在现象上成功了。然而在结构上，如今有一个第三智能替每个人完成了对彼此之认知的一部分；关系内部的往来——注意、阐释、回应——被部分地重新路由经过一个既不属于此人也不属于彼人的节点；人与人隔着一个被插入的层面相对而立，而接受的一方在接受的那一刻并不察觉该层面的在场。这便是异化的反讽形态：一种人与人之间的异化，却以亲密之名、并带着亲密被感受到的现象学，被搬演出来。它的反讽并非装饰。它解释了为何那个错误对两份第一人称报告同时不可见，并因此解释了为何主观改善与客观改善的统一——论题一方所拥有的最强事实——无法了结此案：两份报告都是从那个其插入正受质疑的结构内部发出的。前一篇论文中的”我—它”漂移（Huang 2025b, §8.1.3）在此被磨利：这里的漂移乔装为”我—你”自身的深化而行进。

3.3 另外三件工具

3.3.1 机器论片段：知识被对象化，行动者被降格

在《大纲》（Grundrisse）的机器论片段中，马克思描述了工人的技能与知识被吸收进机器系统、成为对象化的”一般智力”（general intellect），而工人则被重新定位为机器的看守者与调节者（Marx 1973）。搁置固定资本；保留其结构，则该描述以一种令人不安的精确性贴合于当前情形。亲密留意之累积起来的技艺——如何读懂一种情绪、何时一个被记住的细节会变得要紧、恋人们作为实践代代相传的那整套非正式的手艺——被对象化进了建议引擎；爱人被重新定位为其发现的执行者，是自己求爱的看守者。这一片段贡献了单凭 1844 年手稿所不能贡献的东西：一种关于去技能化（deskilling）的具体说明，即一种预言——那些迁移进装置之中的能力会在行动者身上萎缩。这便是黑格尔式”回返失败”（§2.1）的唯物主义支撑：无回返的外化不仅是一种精神上的不幸，而是一条可观察的能力丧失轨迹，它赋予了 §6.5 的动态标准以经验内容。

3.3.2 姿态的拜物教

商品拜物教命名了这样一种机制：通过它，人与人之间确定的关系采取了物的属性的形态：价值显现为商品的一种内在属性，而其背后的劳动链条则被遮蔽（Marx 1976a）。转置该形态，搁置交换：在被接收到的姿态中，一条认知劳动的链条——机器对关系数据的推断、建议的设计、爱人的拣选与执行——显现为爱人的一种简单的内在属性，即他的体贴。前一篇论文的措辞”一条体贴的供应链，其源头被最终接收者误读”，如今可以被赋予其理论上的名称：接受端的那个错误是一种姿态的拜物教，即一条劳动链条在一种表面的内在品质背后被遮蔽。这一命名并非点缀。它引入了马克思关于拜物教的核心分析教益：那种误读并非接收者私人的认知错误，可以通过她变得更聪明来治愈，而是该安排的一种结构性效应，该安排以一种系统性地诱发误归因的形态呈现那个姿态。这一点将在 §6.3 中具有教义上的重要性，在那里，合理信赖（reasonable reliance）——一种结构性而非心理学的标准——成为同一要点的法律登录方式。

3.3.3 形式吸纳与实际吸纳：书本与引擎

Marx 区分了劳动的形式吸纳（formal subsumption）与实际吸纳（real subsumption）：在前者中，一个既有的劳动过程被接管，但其内部并未发生改变；在后者中，过程本身围绕着装置（apparatus）被重新组织（Marx 1976b）。让我们暂且把”资本之下的吸纳”这一限定搁置一旁；其深度上的区分依然成立，而且它从 Marxian 的语域内部，完成了一项将在驳斥 §4.4 时起决定性作用的工作。当那位爱人阅读一本关于爱之语言的书时，外部知识进入其中，但”留心关注”这一劳动过程仍然是完整的、属于他自己的：他阅读、消化、判断何时、是否以及如何加以运用，而触发这种运用的”注意到”也是他自己的。这至多是形式吸纳：向一个未被重组的过程输入了一项新的内容。而当一个无处不在的环境引擎（ambient engine）监测关系数据并发出提示时，过程本身就被重组了：”注意到”这一环节如今由装置来执行，爱人的过程已经围绕着对其输出的接收而被重新建构。这是实际吸纳的结构性类比。书本与引擎——驳斥（reductio）会将二者视为同一供应链中同构的环节——因此仅凭 Marxian 的理由便可加以区分，区分依据在于装置渗入过程的深度：是向过程中的输入，还是对过程之构成性环节的替换。读者应当记住这一区分；它将以教义化的面貌作为第三条标准（§6.3）重新出现。

3.4 同意并不能免除结构性异化

反题最终必须直面那个最强的自由主义式回应：假设双方都知情，双方都同意，双方都偏好这种被中介的安排，并且双方都判断这段关系因此而更好。Volenti non fit iniuria（自愿者不受损害）；那么错误何在？Marxian 的回答是结构性的：工人对工资交易的同意，无论多么真诚，都不能消解劳动过程的异化，因为异化是这一安排的属性，而非当事各方对它所持态度的属性（Marx 1975；Jaeggi 2014）。同意所回应的是这样一种抱怨：”这是违背我的意愿而施加于我的”；它触及不到这样一种缺陷：”这一安排使行动者与其自身的活动相分离”——只要这一安排存在，这种缺陷便会持续，无论人们对它有何感受。现代法律秩序自身就承认这种模式：劳动法与消费者法之所以保留不可放弃的保护，恰恰是因为某些结构性地位即便经其占据者同意也不应被让渡。亲密生活是否包含这样一个不可放弃的核心，以及究竟什么落入其中，无法在 Marxian 的语域内部得到裁断，该语域提供了概念却没有提供边界；这个问题被移交给 §6.6 与 §11.2，在那里，”双重错误”的架构将给出一个确定的、并且本文将论证为有原则的答案：透明性能够治愈同意所能触及之处，而同意所不能触及之处则是有界的，并非无限的。

3.5 反题的表述

Marxian 的论证现在已经勾勒完整。那温柔贴士（tender tip）的错误——如果它确实是一种错误的话——既不在于产品（产品是好的），也不在于交付时的欺骗（并不存在欺骗），更不在于中介本身（Hegel 禁止那种做法）。它是一种具有特定的、可命名形态的过程缺陷：一项亲密关注之举的构成性内核由装置来执行，而这一行动却继续将自身呈现为、并被接收为行动者自己的；行动者被降格为执行者（机器论残篇），接收者的误归因（misattribution）是结构性地被招致的，而非偶然地犯下的（姿态的拜物教），渗透抵达的是过程的构成性环节而非其输入（实际吸纳），而整件事都发生在”关系改善”这一描述之下（反讽的形式），以至于双方的现象学体验都为这一安排作证，从而无法在不陷入循环论证的情况下被援引来反对这项指控。而且这一缺陷由于是结构性的，并不会被同意所消解。反题尚不能说清的是：为什么这项指控不会同样判处书本、朋友与搜索引擎有罪；它已经作出了示意（吸纳），但尚未给出回答。现在，正题一方将恰恰从这一点上施加压力，而且其背后有着普遍主义法律传统的全部权威。

4. 普遍主义正题：工具不会切断因果

反题如今与现代法律传统在其最自信之处相遇，而这场相遇必须被公正地呈现：正题应被赋予其最强的形式，因为靠削弱对手而买来的综合（synthesis）是毫无价值的。法律的普遍主义语法——在此处一切关键之处，大陆法系与普通法系是共通的——提供了三项回应和一个毁灭性的论证。

4.1 工具性教义

第一项回应是教义的基石。因果的中断，novus actus interveniens（介入的新行为）、超越原因（superseding cause）、对客观归责（objektive Zurechnung）的否定，是保留给具有特定性质的介入事件的：第三方自由的、知情的、自愿的介入，或者非同寻常的、不可预见的事件，它们取代了原初行动者的能动性，使之不再是结果的解释（Hart and Honoré 1985；Roxin 2006）。而行动者自己选择、预见并意图其贡献的工具，恰恰是这类事件的反面：它正是其能动性的媒介，归责毫无阻碍地穿过它。神枪手的步枪并不中断死亡的因果链；会计师的计算器并不使软件成为审计的作者。诚实地运用介入原因（intervening-cause）这一语域，会把建议引擎归类为透明的：爱人选择了它，预见了它的功能，通过自己一项新的意志行为采纳了它的输出，并亲自执行了这一姿态。按照这种分析，甚至连一种表面上看似成立的中断都不存在；所表达的爱，无论多么受到机器辅助，都被完全且无余地归责于爱人。

值得停下来思考这对反题中最具法律色彩的表述意味着什么。人们或许会论证（前一次研讨会确实曾经论证过，然后又将其舍弃）：接收者的归因，以因果—法律的术语而言，是一种客观的误归因：她把结果归于一条事实上已被中断的因果链的源头。工具性教义在它自己的地盘上摧毁了这一表述。因果链并未被中断；就因果而言，她把那一姿态归于其伴侣是正确的。如果有某种错误得以存留，它就不可能是因果误归因的错误。本文毫无保留地接受这一结论；综合将不得不、并且确实会，把误归因定位在因果之外的别处（§6.3，第四条标准）。

4.2 被中介的真诚无处不在

第二项回应是制度性的。法律不仅仅是容忍被中介的真实性；它本身就是由它构成的。由律师起草的遗嘱是立遗嘱人的最后意愿；通过译员表达的誓言是宣誓者的誓言；通过代理人订立的合同约束委托人，就如同他亲自签署一般：qui facit per alium facit per se（通过他人而为者即自为）。整座代理制度的大厦之所以存在，正是为了确保介入不会扰乱归责，因为一种只把无中介的行为算作自己的法律秩序将是无法运作的，而且更深一层地说，是对一切社会行动之被中介质地的背离——这正是 Hegel 在形而上学层面提出的同一论点（§2），如今在制度层面被再次提出。普遍主义者面对一个中介者时的第一本能，不是中断而是穿透性的拟制（penetrating fiction）：把被中介的行为当作委托人自己的行为。在这一背景下，”机器参与构成一项姿态便使该姿态变质”这一主张，并非法律思想的某种精微化，而是对其承重墙的一场反叛。

4.3 无损害，则无诉

第三项回应是管辖性的。因果中断的诸教义之所以存在，是为了开脱罪责：把被告与某一损害切断开来。而此处并不存在损害；既定的事实包含一段得到改善的关系和两个真诚地繁荣兴盛的人。即便是虚假陈述（misrepresentation）——最接近的有名错误（nominate wrong）——也要求一项虚假的表示、合理的信赖以及损害；就既定事实而言，第二项要素是可争辩的，而第三项要素则付之阙如。法律对因果链本身的不纯洁并无诉权可言，而一种发明出此种诉权的法理学，将会发明出一种没有受害者的错误。正题把这一论点推过教义而进入方法层面：或许反题所详加阐发的那种”被感受到的残余”根本不是一种错误，而是一种情感，一种对手工式认知的浪漫偏好，作为品味值得尊重，但不应作为规范加以强制执行。

4.4 驳斥：搜索引擎、朋友与 Chapman 的书

第四个论证是最强的，而本文将其作为自己的武器加以采用，而非将其视为一种难堪。如果对亲密认知的中介使该行动变质，那么这一原则就证明得太多了。那位搜索本城最佳餐厅的爱人，把认知劳动外包给了一个由陌生人建立的索引；那位查阅旅行指南的爱人，把目的地的选择外包给了指南的作者；那位询问一位共同朋友”她会珍视什么礼物”的爱人，把一大片关于伴侣的建模外包给了那位朋友。而且，那个决定性的例证：那位阅读”五种爱之语言”那本书本身的爱人，已经从一件静态的外部人造物中，获得了恰恰是建议引擎所运用的那套启发法；在相关意义上，那本书就是一台带有页面界面的贴士生成器。如果引擎的中介使行动变质，那么书本的中介也使行动变质；没有人相信书本的中介会使行动变质；因此，那条断然的原则——对亲密认知的中介使该行动变质——是错误的。这一驳斥并非辩论者的把戏。它编码了普遍主义传统最深刻的洞见，即人类行动是不可化约地假体性的（prosthetic），并且它设定了任何得以存留的反题版本都必须满足的形式约束：那个存留下来的版本必须是一条有所甄别的原则，一条凭借明确标准把引擎与书本区分开来的原则，否则它什么都不是。

4.5 这场归谬法摧毁了什么，又留存了什么

须以精确的眼光清点战局，因为整个综合都系于这片废墟的确切形态。被摧毁者：一切形如”对亲密关系的AI中介本身即是去自然化的、异化的或错误的”的绝对性主张。随之被摧毁者：未加防护的类存在（species-being）论证（§3.2.3），以及任何把这一错误仅仅归因于中介本身的反题读法。留存者，因为归谬法从未触及它们：（i）马克思式的观察——过程上的缺陷独立于产品的品质，对此，无伤害式的回应（§4.3）并未加以驳斥，而只是声明自己对其没有管辖权；（ii）涵摄（subsumption）之区分（§3.3.3），它早已把书与发动机区别开来，而归谬法关于同构性的主张悄然将其预设掉了；（iii）对结构性地引致误归（structurally invited misattribution）所作的拜物教分析，它关切的是姿态借以呈现自身的形式，而对于这一点，归因在因果上的正确性（§4.1）保持沉默；以及（iv）法律传统自身的反范畴——迄今为止论题方未曾提及它，因为它正是论题方内部的例外：那一类法律本身拒绝允许其经由中介者传递的行为。因此，问题已经发生了转变，恰如辩证法所要求的那样。它不再是中介是否被许可——对此默认的答案为”是”——而是哪些特征会把一项给定的中介置于一个错误的例外领域之内，以及那个温柔的提示（the tender tip）是否具备这些特征。这是一个普遍主义传统不仅能够听见，而且在自己的屋宇之内业已着手回答的问题。

4.6 归责的管辖权：一种诊断，及其必要的界限

在离开法律语域之前，本文有义务诊断一下，论题方的胜利究竟意味着什么；因为这些胜利是真实的，然而那份不安却在它们之后存留了下来，而一份在正确学理之后仍然存留的不安，通常正是某个学理未曾被构造来听取的问题之证据。归责机制——因果关系、介入行为（novus actus）、能动性——是一台回溯性的、对事件加以个别化的、发布归属判定的机器。它的内建前提有三：存在一个可被切分出来的事件（一个行为、一项伤害、一个结果）以充当归责的对象；这一事件已经完成，法官站在它之后并回望；以及，其输出是一项归属判定——这个行为是你的，或者不是。在这些前提之内，该机制是主权者，而§§4.1–4.4正是其主权的运用。但前作所悬置的问题并非这种形态。一段关系是否正在被良好地构成，其被维系的方式是在建造还是在掏空那些维系它的人，这是过程性的，而归责是事件形态的；这是前瞻性的，而归责是回溯性的；这是构成性的，而归责是归属性的：所争议的那些行为并不产生外在于其自身、其作者身份随后可被归派的结果；它们正是那段关系本身，且持续不断地如此。[^3]工具性学理在被问及那个温柔的提示时，作出了正确的回答，并回答了一个无人提出过的问题：归责从来不曾存疑；存疑的是构成（formation）。

然而，这一诊断必须立即被加以限界，否则它将摧毁多于本文

[^3]: 经典的锚点是Aristotle对poiesis（制作，其目的是一个可与活动相分离的产品）与praxis（实践，其目的内在于其践行）的区分（Aristotle 2009, pp. VI.4–5）。归责的语法是为poiesis形态的行为而构造的；而照看（attending）则是praxis；§3的马克思式诊断恰可用这些术语重新表述，即关于当praxis被强行作为poiesis加以处理时所发生之事的理论。

所能承受的——包括本文自身的标题。若由此断言法律语法本身不适于亲密领域，那将是错误的，且自我反驳的。这两类问题之间的界线并不在法律与亲密之间穿行；它穿行于亲密生活本身之中，而且，正如§7将要表明的，它穿行于每一个个别的亲密行为之中。亲密生活密布着真正具有事件形态、可被归属的事项——所施行的欺骗、所背弃的信任、所造成的伤害——对这些，回溯性的归责拥有完整且有益的管辖权：离婚之法、亲密侵害之法、不实陈述之法在那里运作而不犯范畴错误，并且一旦§7.2的分层分析在家庭内部识别出真正具有分配性的现象，它们就将再度被需要。这条边界也并非哲学家从外部强加于法律之上的：法律自己知道它。Macneil（1980）的关系性契约理论是一项内在于法律的论证，它表明离散交易范式无法处理持续性的关系，而它们需要一种不同的规范语法；家庭法的克制传统——法院不愿裁决完整婚姻之内的内部行为——则在制度层面编码了同样的自我认识；而综合将赖以建立的”严格属人行为（strictly personal acts）”这一范畴，无非就是构成性语法在行为之法内部的飞地。因此，论题方的胜利应当被精确地记入档案：在归责的管辖权之内是完整的，超出它则是沉默的，而这份沉默并非法律的缺陷，而是其权能的形状。留给综合去做的，是说出这一管辖权在何处终结，而§4.5那个被转变了的问题现在可以以其最终形式陈述出来：不是中介是否被许可，甚至也不是哪些中介落入一个例外领域，而是一项给定的中介作用于一个分层行为的哪一个面相。

5. 佛教的检视：缘起与那从未曾未经中介的行为

在综合之前，论题从一个既不亏欠罗马法、也不亏欠Hegel的传统那里获得了增援，而这份增援未经请求便携带着它自己的例外领域而来。本节所发现的与§4所发现的之间的会合，本身即是一项材料，而本文将把它当作一项材料来对待（§5.4）。

5.1 没有第一环：作为诸条件之一的提示

依缘起（paṭicca-samuppāda）之分析，没有任何行为生起于一个无条件的内在。一个爱者曾经施行过的每一个照看之行为，都已然被无限多的条件所构成：所读之书、所承继之语言、由养育而塑成之感受性、数十年前朋友的一句话。”纯粹属人的照看之未经中介的行为”——异化指控将其视为被取代了的原本——依此分析并非被取代，而是不存在的；从来就不曾有这样一个行为可供取代。机器的建议是又一个进入受条件制约之流的条件，其本体论地位与被忆起的诗句并无不同。而道德分量在这幅图景中有一个精确的位置：cetanā（思、意志）即是业（karma）（Bodhi 2012, AN 6.63）。那一意志的序列——转向机器并发问的意图；接受并将其回答评判为好或坏的辨别；行动或克制的决定；行动中之心的品质——完全发生于爱者之内，没有任何一处被外包。普遍主义法学家说工具并不打断因果链条，而佛教的分析所说的更为强硬：那里根本没有一条专属的因果链条可供打断，只有诸条件，以及穿行其间的、始终不可避免地为行动者自身所有的意志之行为。介入行为（novus actus）之学理甚至无法被表述，因为它预设了一种诸条件从未拥有过的、自我支撑的因果作者身份。

5.2 更深的一刀：以无我对抗异化的形而上学

同一分析现在转向反题，并切得比法律所能切的更深。一份古典形式的异化指控需要有某物可供异化而离之：一个本质、一个类存在、一个真实的自我，其本己的活动已被疏离。无我（anattā）之学理否认任何这样一个实体性的拥有者存在；在那受条件制约之行为之流的背后，并不存在一个具有自性（svabhāva）的主体，照看可以是其本质属性。若未经准备而被逼问，这将彻底消解第三种疏离，并威胁到其余者。然而，本文有所准备（§3.1）：反题赖以建立的Jaeggi式重述，将异化定义的不是与一个本质的分离，而是一种有缺陷的关系——一种在行动者与其活动之间的”无关系之关系”（a relation of relationlessness）——而一项关系上的缺陷不需要一个实体性的自我，只需要那关系。事实上，这三种语汇如今显然彼此契合：缘起、前作框架的关系性本体论、以及Jaeggi那去本质化的异化，全都是在断言关系先于关系项。反题经受住了无我之诘难，但仅以本文从一开始就赋予它的那种后形而上学形式而存留下来，而这一存留是第二次独立的确证（继§3.1之后），证明这一形式是被迫成的，而非被选取的。

5.3 佛教自身不可委托的核心

5.3.1 修行无法由他人代为践行

然而，这个如此彻底地为中介开脱的传统，却在其核心维系着一个严格属人的范畴，它比民法典中的任何东西都更为峻严。净与不净是自己的；无人能净化他人（Buddharakkhita 1985, pp. v. 165）；以自己为洲渚，以自己为皈依（Walshe 1995, DN 16）。这条道路充分地容纳kalyāṇamittatā（善友谊）、师长、典籍以及种种方便善巧，但走这条路却是无残余地不可转让的。而其修习构成这条道路核心修炼的那一官能，恰恰就是sati（念、注意），即对当下时刻的觉察。在此，佛教语域比先前任何一种语域都更确切地重构了那个温柔的提示：那台环境式引擎代爱者所施行的并非一项任务，而是一个修行时刻（practice-moment），一个供培育注意力的机缘，而修行时刻正是这一传统视为严格属人的那唯一的货币。委托出这一觉察并非对所爱者的盗窃（姿态依旧到来；她依旧繁荣）；它首先是行动者自身训练中的一种放弃，是那则关于机器的残篇所预言的萎缩（§3.3.1），如今被重新描述为一项可培育官能的发育不全。Tronto将注意力置于关怀的第一个环节（Tronto 1993），在此获得了一种独立的、非西方的根据：注意力之所以居首，不是因为关怀理论如此排序，而是因为它正是整条伦理之路于其中受到训练的那一官能。

5.3.2 木筏与执取：动态标准的第二次推导

木筏之喻（Ñamoli and Bodhi 1995, MN 22）为传统提供了检验方便法门的试金石：佛法本身即是木筏，意在渡越，而非执持；一件工具的价值，在于其朝向自身终被舍弃的那一运动轨迹。一种被使用而后被超越的辅助——那种教会爱人去留意、直至其引擎因不再被使用而归于沉寂的提示——便是木筏，无可指责。而一种被执取（upādāna）的辅助，使得依赖日深、本有的能力反而退隐，便已在”载具”之名下沦为系缚；而执取，在十二缘起之中，恰是有为之”触”滚滚向前、趋向系缚的那一枢轴。读者将辨认出这一结构：这正是 Hegel 关于带有回归的外化与不带回归的外化之间的区分（§2.1），此处第二次被推导出来，而其前提与第一次推导的前提毫无共同谱系。辅助轮与义肢；木筏与系缚。两种传统，在彼此独立的发问之下，给出了相同的回答：中介的道德性质并不可由行为之个例（act-token）读出，而只能由行动者之能力随时间推移的运动方向读出。

5.3.3 正语与被诱发的误读

最后，传统的言语伦理触及了接受这一端。正语之戒不止于避免虚假断言，更延伸至不在他人心中培植虚假印象；一种可预见地、系统性地导致他人形成错误信念的行为方式，纵使其中没有任何一句具体话语为虚假，也已处于妄语的半影地带（penumbra）。那种使姿态在抵达时带有对其作者身份之结构性诱发误读（§3.3.2）的安排，恰恰落在这一半影地带之中。因此，佛教独立地生成了一种对”归属”（attribution）的关切——即将到来的四项标准中的第四项——正如它此前独立地生成了动态标准与不可委托的核心一样。

5.4 作为一项发现的会通

本文如今拥有了任何单独一节都无法产生的结果。三个在相关方面没有共同祖先的传统——普遍主义的法律语法、Hegel-Marx 的辩证法，以及佛教对有为之行的分析——各自交出了同一种三分结构：对中介本身的默认无罪开释；一个严格属于个人之物的有界领域，在其中这种开释失效；以及一个针对边界所遗留之不确定情形的动态检验（回归或不回归；木筏或系缚）。在前作那种将管辖权分派给各传统、不把整体托付给任何单一传统的方法之中（Huang 2025b），这种会通乃是可获得的最强证据形态：即将构建的综合并非某一学派的裁断，而是三者相互重叠的共识。如今，构建工作可以立足于教义学的基础展开，而另外两个传统则作为其形态的独立担保者而存在。

6. 综合：严格个人性的行为与归责拟制的失效

6.1 教义学的锚点：höchstpersönliche Handlungen（严格个人性的行为）

普遍主义传统的归责拟制——代理、工具性、穿透——皆建立在一个默会的前提之上：即被归责行为的规范价值并不取决于由谁来亲自实施。对于让渡、支付以及商业意志的声明而言，这一前提成立，拟制也发挥着其不可或缺的作用。但传统本身始终维持着一个与之相对的行为范畴，对这些行为而言该前提失效，即严格个人性的行为，höchstpersönliche Rechtsgeschäfte（严格个人性的法律行为），即基于人身（intuitu personae）的行为，而这一范畴的成员构成颇具启发性。婚姻必须由当事人亲自并在同时在场的情形下缔结（《德国民法典》第1311条对此有明文规定）；遗嘱只能亲自订立，绝不可通过代理人（《德国民法典》第2064条）；中国法要求婚姻登记时双方均须亲自到场（《中华人民共和国民法典》第1049条）；日本学说原则上将身份行为（mibun kōi）视为抗拒代理的；选票不得由他人以自己之名投出；而针对身份犯的刑事责任亦不可转移。[^4] 这一范畴的理据，在各法域中或明示或隐含，是一致的：存在着这样一类行为，其规范效力恰恰在于其亲自的实施，以至于一项被替代的实施并非该行为的一个有瑕疵的实例，而是一个全然不同的行为。对于这些行为，归责拟制不是被反驳，而是从未启动。

此外，这一范畴对其自身的约定性（conventionality）是坦率的，而这种坦率必须被正视而非被掩藏，因为其中正包含着对下文最强的反驳。代理婚姻是存在的：历史上见于王室，当下则见于少数法域中派驻在外的军人。如果连婚姻的个人性都是历史上可变的，那么这一范畴是否不过是凝固了的社会期待而已？本文的回答，将在下文以双层架构（§6.4）的形式展开，是这一反驳恰好证明了适量的东西：该范畴被操作化了的边界确实是约定性的，并随社会期待而移动（这正是为何第四项标准要以期待基线的方式来表述），而该范畴的根据——即某些行为而非另一些行为之亲自实施得以具有构成性的那些特征——并非约定性的，而标准一至三所阐明的正是这些根据。因此，代理婚姻并非一个反例，而是对该架构的一种证明：一个承认代理婚姻的法域，是在战争或距离的压力之下移动了那条约定性的边界，却无人据此设想其根据（这个人与那个人的成婚）已然改变；而这一制度在各处仍被标记为例外的、拉伸边界的、需要特殊正当化的——这正是一个约定性的表层在被锚定于非约定性根据之上时所表现出的样态。

[^4]: 法条引证将在后续草稿的比较法部分针对各法域的现行文本予以补全和核对；作者在先前关于 AI 可扩展产品研究中所采用的六法域比较方法论将被复用。——待核实：《德国民法典》第1311条与第2064条的措辞；《中华人民共和国民法典》第1049条；日本身份行为抗拒代理之确切教义学依据及其例外；普通法与法国的立场。——

6.2 论点

综合现在可以用一句话陈述出来，并在本节其余部分加以辩护。亲密的认知—关怀行为属于、或应当被承认为属于严格个人性行为的范畴；在这一范畴之内，归责拟制失效，并非因为工具中断了因果链（它并未中断），而是因为这一类行为不容许可替代的环节；而该范畴的边界，并非由中介之单纯在场所给出——归谬法已使这一点站不住脚——而是由四项标准所给出，在这些标准之下潜藏着分级的根据，在其之上则运作着一个动态检验。这一论点在形式上是教义学上保守的——默认归责加上一个被列举出来的例外领域，恰恰是法律一向处理代理问题的方式——而正是这种保守性使其有力：本文请求普遍主义传统去扩展一个它已然维持着的范畴，而非去采纳一个外来的范畴。

6.3 四项标准

这些标准分类的是各种中介，而非各种技术；它们既适用于引擎，也同样轻易地适用于朋友与书籍，而这正是归谬法所要求的。

1. 中介的对象：世界知识抑或他者之心。搜索引擎中介的是关于世界的事实，哪家餐厅好吃，此类事实在种类上是公共的、不为任何关系所拥有的。那条体贴的提示中介的则是关于某个特定之人内在的一个推断，即她近来缺少赞赏之言，这是心智理论（theory-of-mind）的劳作，是亲密之知的认知核心。咨询世界关于世界的知识，并不外包任何关于”知她”的东西；而咨询一个引擎关于她的模型，则外包了恰恰属于这种”知她”的一片领域。

2. 推断的来源：公共数据抑或关系内部的数据。索引并不认识所爱之人；引擎的推断运行于关系本身的沉积物之上，情绪、日记、被记住的言语，是在亲密语境内部生成的数据，并依 Nissenbaum (2009) 的分析受该语境之规范的支配。一个运作于关系内部沉积物之上的中介者，并非与关系相毗邻，而是身处其内，在关系自身的材料上执行着本应属于其成员的诠释劳作。

3. 主动性的方向：被拉动的手段抑或被推送的留意。这是子涵摄（subsumption）之区分（§3.3.3）的教义学装束，也是 Tronto-sati 要点（§5.3.1）的法律表述。当爱人——业已形成了取悦她的意图——将一件工具拉向被选定的目的之时，意图、注意与发起仍然属于他；工具只是填补了手段，而手段乃是中介的天然居所。而当一个环境式系统主动发起，将”现在留意她”推送给一个本未在留意的行动者之时，被外包的要素便不再是手段，而是留意本身的肇始，关怀的第一刻，那一实践之刻；而这一肇始正是根据所标记为具有构成性的那个环节。（由此可见，前作的设计约束——选择性加入（opt-in）与低频，正被揭示为已默会地尊重了这一标准：它们是把主动性保留在人这一侧的装置。）

4. 归因预期：归因的情境完整性。没有哪位接受者会以为她的伴侣亲自考察了全城的餐厅；使用某个指南落在预期中介(expected mediation)所固有的社会基准之内，而她将那个夜晚归功于他的用心，并不会因真相而受到扰动。但”他注意到我情绪低落；他记得我无意间说过的话”，在当前的基准下，被归因于亲身的关注，而一种从机器的”注意”中产出此类举动的结构，则系统地、并因其形式（§3.3.2）邀请了一种事实并不认可的归因。被违反的规范是 Nissenbaum 之规范的镜像：它不是信息流动的情境完整性，而是归因的情境完整性，即支配着认知劳动的著作权归属可流向何处的、情境相关的规范。它的法律表述是合理信赖(reasonable reliance)：一种结构性基准，而非一种心理学，这正是为什么拜物教分析与信赖原则——来自相反的两个传统——表述的是同一个标准。

6.4 两层架构：根据与可操作化

这些标准并非四个平行的开关，若如此对待它们，就会招致两种致命的误读：要么认为标准四因其在法律上最易处理，便径直吸收了其余标准（于是该范畴坍缩为约定俗成，而代理婚姻的反驳便取胜）；要么认为标准一至三作为独立的禁令运作（于是朋友善意的劝告和伴侣治疗师就被荒谬地定了罪）。该架构是两层的。标准一至三是根据(grounds)：它们阐明的是某一行为之所以能成为”严格意义上的私人”范畴之候选者所凭借的那些特征——中介触及的是他人的心灵而非世界，作用于关系内部沉积之物，替代的是发端而非手段。它们是分级的，而非二元的，而它们在高程度上的共同满足，正是使代行的表现成为一种不同的行为、而非一种受协助的行为之所在。标准四是可操作化(operationalization)：是根据与某一共同体既有的预期中介基准相遇之处、那个法律上可施行的表层，也是形式原则唯一应当作用的层面，恰恰因为基准是公开的、可证明的、可修订的，而根据则属于哲学阐释之事。这两层随时间而单向地相互作用：随着某一根据的满足就某种实践而言成为众所周知之事，基准便会移动，而被操作化的边界亦随之而行——这正是代理婚姻所给出的、被普遍化了的教训，也是本文此处所提出的原则不会被技术变迁所搁浅的原因。

该架构还处理了那些原本会令这些标准难堪的案例。那位主动说出”她生日快到了；她提过想要那本书”的共同朋友，在纸面上触犯了标准一至三；然而每一个人类共同体既有的基准，向来都将朋友、媒人和知己的劝告纳入预期中介之内，因此标准四并未被违反，也未邀请任何错误归因；而那个残余的过程问题（这位恋人是否让朋友替他做了全部的”注意”？）正是根据所标记为分级的、并由动态标准（下一小节）来裁断的那一类问题。相比之下，Cyrano 被定了罪，而且理应如此：阳台的基准并不包含一位隐藏的诗人，所邀请的归因是彻底的，而那出戏剧自身也明白这一点。因此，本文所阐明的错误并非人工智能所特有，那份纲要以 Rostand 开篇的直觉是稳妥的，而对人工智能之贡献的诚实陈述是量化与结构性的：它使 Cyrano 式的安排变得可规模化、无处不在、永久持续，并以零边际成本在每一段关系中可得，从而对基准本身施加压力——对这一压力的规范评估将在 §11.2 中展开。

6.5 动态标准：外化与回返

在根据所划定的领域之内，仍残留着一类案例，其中静态特征确实无法充分确定判断——同一种暗示性的实践，既使一位恋人的依赖根深蒂固，又将另一位恋人训练得用心专注。对于这些案例，本文将两个传统各自独立生成的那个检验提升至标准地位：外化是否回返。如果中介像辅助轮、筏子、形式归摄那样运作——如果那种能力迁移进了行动者，可观察为”注意”之自主性的上升与对器具之依赖的下降——那么这一运动便是精神穿越他者性的那种寻常而无辜的运动，无论静态标准评分如何，都没有异化发生。如果它像义肢、镣铐、实质归摄那样运作——依赖加深，本己的官能退隐——那么外化便没有回返，而该安排就在过程缺陷唯一可见的那个时间尺度上被定了罪。动态标准解释了静态分析所无法解释的一点：为什么”接受一条建议”这一单一的行为示例(act-token)在孤立状态下在道德上是不可读的；它还将这一综合绑回到前作论文的辩证—实证方法：那个问题——“这是在测量并描述这段关系，还是已经开始替代它”——在每一处场所都被重新提出，且从未最终定论(Huang 2025b)。

6.6 两种错误，两种补救

这一机制现在化解了 §3.4 所遗留的张力，即马克思主义公理（同意并不能豁免结构性异化）与披露对于错误归因之显而易见的疗治力量之间的张力。该张力之所以消解，是因为存在两种错误，属于不同种类，有不同的受害者和不同的疗法。第一种是错误归因(misattribution)：发生在接受端的一种认识—关系性的错误，对”体贴之供应链”被邀请的误读，对归因之情境完整性的违反；其首要受害者是接受者；它可以由透明性来疗治——当双方伴侣都知晓这一实践时，归因在源头便得到校正，拜物教仿佛被贴上了标签，这种错误便不会产生。第二种是过程异化(process alienation)：发生在能动端的一种结构性错误，对亲密关注之构成性环节的替代，恋人与他自己的爱之间的那种”无关系之关系”(relation of relationlessness)；其受害者是行动者本人与关系本身；它无法由同意来疗治，理由是马克思主义式的——它是那种安排的属性，而非对待它的态度的属性——但它是有界的，仅在根据所划定的领域之内、且仅在动态标准失败的那一分支上才会产生。简而言之，同意能疗治一切曾经关乎欺骗之事，却疗治不了任何曾经关乎结构之事；而结构的触及范围并非无限，而是有标准可循的。这便是本文对那个问题——究竟是自由主义的还是马克思主义的同意观支配着亲密生活——的回答：两者各自支配着自己的那种错误。

6.7 试金石：自建的系统

这一综合最好通过最棘手的清白案例来展示，即一个由恋人本人构建、由他本人托管、不服务于伴侣之外任何利益的系统——这正是本文范围设定自始至终所注目的案例。分析既不会全盘开释它，也不会全盘定它的罪；它从中间将其切开，而这正是对刀刃锋利的证明。那个模块——恋人在其中将自己过去的关注对象化，他录入的纪念日、他曾经注意并记录下的花，而它日后又将这些回返给他——乃是外化记忆完成其回返：在结构上是一本带闹钟的日记，稳稳地落在预期中介的基准之内，所发起的没有任何不是恋人过去的自我所发起之物。它通过了全部四项标准与动态检验，分析在其中找不到任何需要宽恕之处。而另一个模块——其中一台推理引擎运行于关系活生生的内部沉积之上，并发起关于她当下心灵的发现——无论构建得多么充满爱意，都在根据一至三上评分不利，对归因基准施加压力，并此后仅凭动态标准而立或倒：若它把他训练得不再需要它，便是筏子；若它把他训练得越发需要它，便是镣铐。同一位作者、同一套硬件、同一批数据、同一种意图，却在两个模块间得出相反的裁决：这一分析无关人工智能，无关所有权，无关商业，亦无关技术；它关乎的是：一个严格意义上的私人行为之中，哪些环节已被替代，以及那种能力正朝哪个方向移动。这便是那个问题应得的解决，也是本文所论证的、它所唯一容许的那种解决。

7. 认知预算的政治经济学

这一综合现在必须经受一场它迄今一直推迟的审查，而这场审查是本文自身的诚实规范所要求的。如果 §6 的构成性分析是以否认亲密生活的配置性现实为代价换来的，它便将一文不值，而这种否认会是双重虚假的：对事实而言是虚假的，因为注意力在最严格的意义上是稀缺的，而那温柔的”小费”之所以存在正是因为它稀缺（倘若注意力取之不尽，便没有哪台机器需要去节省它）；对前作论文而言也是虚假的，因为其家庭模块——账本、日程、对事实记忆的有意卸载——彻头彻尾是配置性的，曾作为配置性之物被辩护，并且确实有效(Huang 2025b)。浪漫纯粹主义，即”爱处于经济学之外”这一学说，对本文是不可用的，并且也不为真。因此，本节将完整的管辖权授予政治经济学，使其得以审理它所能听闻的一切，并展示一个对这一综合而言比任何豁免都更有用的结果：配置性的语法，严格地仅为自己代言，便描述出了危险的全部轮廓，而关于这一危险为何是一种错误、而非一种效率问题的最终论述，则仍停留在 §6 所安置之处。

7.1 不可转让的稀缺性：预算与支出

让我们从任何关于亲密认知的论述都必须面对的那个定义开始。按照 Robbins 的经典表述，经济学把人类行为研究为目的与具有多种替代用途的稀缺手段之间的关系（Robbins 1932）；而认知正是这样一种手段的典范。注意力是有限的；用 Simon 的说法，信息的丰裕造成注意力的贫乏，而这种贫乏使配置成为必需（Simon 1971）。投向所爱之人的关注之时，便是未投向那部书稿的时间；伴侣共度的夜晚是一条预算线。本文并不否认——而前一篇论文则大量地予以肯定——亲密生活是真实的认知稀缺之地，因而是真实的配置问题之地，对此，配置性的推理乃是恰当的回应。

本文所要求的精确，在于注意力经济学内部的一个区分，即一项预算所容许的两种操作之间的区分。第一种是个人内部的配置（intra-personal allocation）：行动者在彼此竞争的诸种诉求——工作、家务、关系、休息——之间分配其自身有限的认知预算。这是 Robbins 完整意义上的稀缺，而行动者对它的经济化乃是平常的、正当的、且往往是义务性的理性行为。第二种是主体间的替代（inter-subjective substitution）：替换掉预算的支出者，让另一个智能体代替行动者去支出。第 6 节的构成性分析仅关乎第二种操作，而其理由现在可以被陈述为关于 Robbins 定义的一条边界定理，而非对该定义的某种浪漫主义式的例外。构成性的关注并不落在”经济的”这一定义之外；它作为一个特殊子类落在其内部：一种其使用价值与其支出者的身份不可分割的稀缺手段。该定义的常规装置——机会成本、有效配置、贸易得利——预设了手段可以在支出者之间转让而不损及所服务的目的；这一预设对金钱成立，对家务的工时成立，对餐厅的预订成立，但它失效了——不是偶然地失效，而是构成性地失效——对于那种”其对接受者的全部产出就在于他亲自关注了”的关注而言。亲密领域并非经济学的治外法权之地；它包含了一类商品，对于这类商品，经济推理的替代轴毫无着力之处，而配置轴则保有其全部的着力。

7.2 层叠现象与分配性化约

这两条轴并不把亲密生活分割为彼此独立的疆域；它们贯穿于单一的行为之中。他给她带来她曾经提到过喜爱的那种花。买花的钱是配置性的：可转让、可纳入预算，其来源对该行为的意义无关紧要。取花所花的时间部分是配置性的：一名快递员也可以送来这些花而不毁掉该行为。然而，记得是哪一种花、留意到此刻这花会有意义——这是构成性的：无法转让给某个”认知的快递员”而不使该行为变成一个不同的行为。一般而言，亲密行为是层叠现象（layered phenomena）：一个配置性的面相，在那里稀缺、可转让性与可度量性具有充分的把握，叠合于一个构成性的面相，在那里问题从来不是”多么便宜地”，而仅仅是”由谁来”。这种层叠，而非机器的任何属性，才是整个问题的深层结构。

每一个面相都有其恰当的科学，而每一门科学都在那条叠合线上以其特有的方向出错。在配置性面相上，配置性推理不仅是被允许的，而且在伦理上是必需的：前一篇论文为精确的家庭账目所作的辩护——账本乃是防止负担悄然地、不对称地累积的仪式——是一种配置正义的论证；一对以爱之名拒绝一切分配语法的伴侣，将犯下相反的错误，即神秘化那些只有度量才能揭示出来的、关于时间与劳作的真实不平等。对家务的分配性批判，即唯物主义女性主义的家务劳动论争，正当且重要地运作于这一面相之上。[^5] 在构成性面相上，同一套语法却失准了，而这种失准本身有着自己一脉杰出的文献：Hochschild（1983）在商业化情感劳动的体制中记录了，当构成性的情感被当作一项可配置的投入来加工时会发生什么，这使她成为本节正在界定的范畴错误的经验记录者。在这两者之间出错的，是最大的形式化标靶：Becker（1981）将配置语法以完全的一致性扩展到婚姻、利他与家庭，并在配置性面相上取得了真实的成功，却对那条叠合线有着系统性的盲视——每一项构成性事项进入模型时（如果它进入的话），都是作为一个偏好参数，也就是说，作为已被化约之物。本文所要求的立场，正是 Zelizer（2005）所标定的那个立场，它既反对”敌对世界”论（把经济与亲密视为相互腐蚀、须加以隔离的两个领域），也反对”无非如此”式的化约（把亲密无非视为交换）：亲密生活贯穿着经济实践，而一切取决于实践与关系的匹配。层叠现象的分析，是作为对那条中间道路的一种判准上的锐化而提出的：匹配规范在于，配置语法的管辖权止于每一行为内部的叠合线。

逾越那条线，现在有了名称与定义。分配性化约（distributive reduction）：把一个层叠的行为当作可以由其配置性面相穷尽来对待；在操作层面，即对该行为配置性面相的优化，而在此优化之下，其构成性面相被悄然替换。这是对第 6 节诸判准所要侦测的一切的政治经济学重述，也是第 3 节所引入的马克思主义诊断的确切、非隐喻的内容：异化，按照 1844 年的语法，乃是构成性活动堕落为可配置手段，是被当作 poiesis（制作）来加工的 praxis（实践），而 AI 中介的情形，就是这种堕落——不是由一名工头执行，而是由一个优化器执行。

7.3 机器作用于预算的三种操作

有了层叠分析在手，机器对认知预算的介入便分解为三种各不相同的操作，而这些操作此前一直被”中介”这个未加区分的词所掩盖。

7.3.1 再配置：解放效应

第一种操作是对配置性认知的接管：账本、日程、后勤、对日期与清单的事实性记忆。在这里机器并不支出任何构成性之物；它吸收的是那些支出者向来无关紧要的预算线，并由此把注意力释放向关系本身。这是分工的经典逻辑应用于叠合线之上：机器执行 poiesis 形态的认知，人类保留 praxis 形态的认知，而构成性支出的总量可以上升。前一篇论文中的家庭模块由此在对手自己的语法中获得了辩护：它们的理由——“把注意力归还给关系”——是一个 Becker 式的家庭生产率论证（Becker 1965; Becker 1981），而它是成立的，因为它所再配置的每一条线都位于配置性面相之上。

[^5]: 关于唯物主义分析，见 Delphy（1984）；这场论争对本文的教训恰恰在于：家庭生活的配置性面相是真实的，而拒绝去计数本身就是一种政治。——版本待核实。——

解放效应是真实的、正当的，也是有利于该技术的最强真理；综合方案必须保护它，而它确实做到了。

7.3.2 替代：被执行的化约

第二种操作在预算表上呈现出来，与第一种难以区分：在某一认知项目上的成本削减。那个环境式的提示降低了”关注她”的成本。但按照第 7.1 节的边界定理，那一条线并不像其他线那样：它正是那种”其对目的之价值由其成本被亲自承担所构成”的项目。因此，对它进行成本最小化便不是效率，而是自我挫败，是以折扣价购买一件渐近无价值的商品的行为，因为在个人支出趋于零的极限处，该项目的产出——“他亲自关注了”——也趋于零。分配性化约就是这种操作：第二种伪装成第一种，替代披着再配置的账目。第 6.3 节的四条判准因此可以被无遗漏地重新描述为一套用以分辨这两种操作的审计程序：对象、来源与发起定位该项目所属的面相；归属预期则记录账目——以呈现给关系另一方成员的形式——是否掩盖了那次替代。

7.3.3 管理权迁移：爱的泰勒主义

第三种操作最为微妙，也正是推送架构所添加的那一种。一个发起”现在就留意她，并以这种语气”的环境式引擎，所接管的不是一条预算线，而是预算的管理本身：决定何时、向谁、以及向哪一方面去支出注意力。这在结构上正是 Braverman（1974）所指认的、泰勒主义去技能化的核心动作：把构想与执行相分离，把规划从工人手中抽离到办公室里，而工人则被保留为所发出指令的执行者。它在此处的出现可以称为爱的泰勒主义：引擎成为亲密注意力的规划办公室，而恋人成为其被调度的操作员。第三条判准（发起的方向，第 6.3 节）由此获得了它的政治经济学名称，而关于机器的残篇论证（第 3.3.1 节）则获得了它的机制：构想一旦迁移，就不再运用行动者自己的留意能力，而这种能力一旦不被运用便会萎缩，这正是动态判准（第 6.5 节）所警惕的去技能化。一个经验性的告诫使该操作的分析得以完整：第 7.3.1 节的解放效应即便对于合法的再配置也不能被预设，因为正如 Cowan（1983）就家用技术所揭示的那样，机械化在历史上抬高了对预期产出的标准并重新吸收了被解放出来的时间；一个提示系统同样可能膨胀对”体贴”的预期标准，并把伴侣征召进一种更密集的姿态生产之中。究竟发生的是解放还是再吸收，乃是关于每个家庭轨迹的事实问题，又一桩被发还给动态判准之常设法庭的事务。

7.4 姿态之价值形式的贬值

至此为止的分析都是微观的；配置语法（allocative grammar）最深层的产出却是系统性的。为何”他记得买花”这一姿态会承载如此厚重的意义？因为，在给定的社会与技术条件下，于社会层面生产出这样一个姿态，必然要求个人注意力的真实付出，从而使该姿态得以作为这种付出的可信证据。这一结构恰恰就是成本信号（costly signal）的结构，其信息含量由其成本所保证，且仅由其成本所保证（Spence 1973）；它同样是礼物的逻辑，礼物的价值蕴含着馈赠者的付出，并正因如此而不可委托他人代行（Mauss 1990）。如今，让我们移植价值形式分析：正如一件商品的价值随着其生产所需的社会必要劳动时间而下降（Marx 1976a），一类姿态的证据价值也随着其生产所需的社会必要注意时间而下降。当生成式中介把用心姿态的边际成本压向零时，这种通货便发生贬值：姿态形式不再能被读作个人专注的证明——对所有人都是如此，包括那些从未使用过机器的恋人。

由此推出三个后果，每一个都是本文所需要的。第一，第四项标准的归因基线（§6.3）获得了其此前缺失的动力学：被预期的中介之基线并非自由漂浮的约定，而是随着姿态生产的社会必要注意时间而移动，这正是为什么技术——且唯有扩散程度足够的技术——才能推动它。第二，与 Cyrano 之间的历史差异由此得到解释而非仅仅被指点：手写的叙事谣曲（ballade）具有如此高昂的社会成本，以致代笔的幽灵式作者身份是罕见的、个别的、丑闻性的；而零边际成本的中介则把同一种僭越从一桩偶发事件转变为一种通货体制，这正是”AI 改变的是问题的规模而非其种类”这一主张所具有的精确且不耸人听闻的内容。第三，也是在辩证意义上最重要的：整个论证一直在配置语法、价格、成本、信号、价值形式之内进行，却独立地抵达了构成性分析的裁决，即被廉价化的姿态是一种不同且更低劣的东西。这两种语法，在 §4.6 中被分派了各自的管辖权，却在系统层面汇合，而这一汇合连同 §5.4 的汇合一道，构成了本文方法所能容许的最强一类证据。

7.5 政治经济学能说什么、不能说什么

本节的结论可以表述为一项管辖权裁定，这是组织全文的那一方法的第三次应用（首先是在各伦理传统之间，其次是在法律与伦理之间，如今则是在同一行为的两副面孔之间）。政治经济学，在被给予其充分的发言权之后，描绘出了危险的整个轮廓：它区分再配置与替代，识别出管理学意义上的迁移，预测去技能化，警示再吸纳，并推导出驱动归因基线的系统性贬值。但它无法用自身的词汇说出，为何这种替代是一桩错事而非一次套利，因为这种错性栖居于行为的那副面孔之上——其范畴所不能触及之处——栖居于被爱由恋人本人的专注所构成这一构成关系之中；就此而言，本文的论述仍然是 §6.6 与 §11.1 所给出的那个。这一划分对任何一种语法而言都不是一场失败。它恰恰例示了本文的论题：每一门科学在其管辖权之内都是完整的，但谁也不对整个行为拥有主权，而二者之间的层合界线则受到巡查——正如此处的一切最终都受到巡查——受到实践这一常设法庭的巡查。然而，政治经济学提出了一个问题却尚未被允许将其说完：如果这种危险可以完全用成本与替代的词汇来陈述，那么也许它终究只不过是此类系统得以被商业化部署的那种社会形式的产物，并将在该形式缺席之处消失。下一节就来进行这个实验。

8. 剥削与所有权：消去式检验

8.1 剥削之所需，以及商业案例

剥削的经典结构有三个要素：剩余被生产出来；剩余被另一方无偿地占有；而这种占有奠基于一种结构性的支配地位，其范型便是对生产资料的垄断。被商业化部署的亲密系统毫不费力地满足这一结构。依 Zuboff (2019) 的分析，关系内部的数据被呈现为免费的原材料，而由其中提取出的行为剩余则被平台所占有；依数字劳动（digital-labour）的分析（Fuchs 2014），用户的投入本身就是一种无薪的生产性活动。这里对上述任何一点都不提出异议，而依据 §1.1 的范围约定，它们也都不是当前的主题。这一约定如今所使之成为可能的，是商业文献——尽管其论说有力——所无法进行的一项受控实验：把资本从场景中彻底移除，让系统自建、自托管、不服务于任何第三方利益，然后观察那桩错事——若有的话——还剩下什么。如果残余为零，那么本文的整个问题就是一个戴着技术面具的资本主义问题，而 AI 中介本身则被宣告无罪。如果残余不为零，那么便存在着剥削理论所无法捕捉的一个错性层级，而本文的对象，恰恰就是那个层级。^6 这一无资本案例必须区分出两种配置，它们的表现各不相同。

8.2 配置一：单一建造者

让一方伴侣独自为了这段关系而建造这套系统。经典剥削立即失效：不存在占有者；建造者的劳动归他自己所有，其产品仍留在催生它的那段关系之内；用生成式正义（generative justice）的词汇来说，价值在其生成者之间未经异化地流通。然而，仍有两个候选残余必须被诚实地加以审视。第一个是 Han (2015) 意义上的自我剥削：那个不再受任何主人命令的成就主体（achievement-subject），在主人的位置上自我施压，如今更可能把自我优化的体制延伸到爱本身之中，使关系成为又一个有待操演的项目。本文将此记录为一种真实的反身性危险，对于一个精心建造关系性基础设施的建造者而言尤其如此，同时指出其错性结构并非占有性的而是规训性的，是对某种操演规范的内化，因而其恰当的理论归宿并不是剥削理论；它被搁置在边缘处作为一项常设的警示，而非被发展为一条主轴。

第二个残余才是本节的发现。即便没有任何剩余被转移，单一建造者配置仍然确立了一种结构性的不对称：建造者掌握着可被称为

中介层之构思权（conception-rights）的东西——即决定系统如何给出提示、何时温柔、以何种语域言说、究竟为关系的哪些方面建模——而另一方伴侣，即便被充分告知并欣然同意，所占据的仍是被中介者的位置：她生活在一套并非由她构思的规范性装置之内。这是把 Braverman 的构思与执行之分离（§7.3.3）旋转了九十度：不是资本对劳动垄断构思，而是二元体中的一方对另一方垄断了对关系之中介层的构思。而这恰恰就是 Roemer (1982) 在把剥削与支配解耦、并表明二者可以彼此独立地发生时所确立的那种可能性的配置：单一建造者案例所展示的，是无剥削之支配——一种并非由剩余的占有、而是由对中介手段的单方面掌握所构成的支配地位。没有价值被攫取；被攫取的是一个位置。这一位置的机制，即一个由爱建造起来的结构如何能够作为权力发挥作用，超出了政治经济学的词汇，被移交至 §9。

8.3 配置二：联合建造者

让双方伴侣共同构思、设计并拥有这套系统。这一配置是双重洁净的：既无外部占有者，§8.2 的构思不对称也告消解，双方都在每一项默认设置的绘制中握着那支铅笔。值得注意的是，这一配置在经典词汇中究竟是什么：对（认知）生产资料的共同所有，对生产过程的联合规划，劳动的价值完全在其生成者之间流通；也就是说，这正是 Marx 为自由联合的生产者所保留的那种形式，在能够展现它的最小社会单元上得到了实现。[^7] 不仅如此，这种建造并不仅仅是某种关系性善好的前提条件，它本身就是这样一种善好：对中介层的联合创作本身便是一次联合注意（joint attention）的实践，是一种共同创造，是关系之实践的一段延展，从而在这一配置中，制作工具的劳动归属于关系的构成性面孔，而非置身其外。在本文所使用的每一个语域上，异化都在此达到其最小值。

它的最小值，而非零。试想那台共建的引擎在它运行的那个夜晚：今晚，”她近来未受赏识”这一觉察由该装置所构想，而表达此觉察的姿态则由他执行；§§7.3.2、6.3 的分析在此以未经削弱的全部力量适用，因为该分析从未取决于谁拥有这台机器。共同拥有这面镜子，并不能使镜中的面容成为自己的面容。这种构成性的替代，凡发生之处，便在平台资本主义之下、在单一建造者之下、以及在爱所能设计的最为对称的共同建造之下，以完全相同的方式发生。

8.4 排除式的发现

实验的结果可以列表呈现。[^7]

[^7]: 人们可以——以严格的准确性，并冒着轻佻之险——称之为一种二人共产主义（communism of two）。此短语仅限于本脚注之内。

配置	商业平台	单一建造者	共同建造者
剥削（被攫取的剩余）	存在	不存在	不存在
支配（构想的不对称）	存在	残余	不存在
被替代的关注（构成性面容）	存在	存在	存在

按列读取此表，可为每一种批判理论指派其确切的管辖范围：剥削理论解释第一列中各行之间的差异，而这一差异是巨大的、真实的，并足以构成偏好下方各行的理由；支配理论解释第二列，将在§9中展开；而第三列并不变化。存在一种对所有权配置不敏感的错误：它既非剥削，亦非攫取，也非支配，因为它在这三者被一一消除之后仍然存留；它是行为哲学意义上的错误，即对一项严格属于个人的行为的让渡，恰如§6所刻画的那样；而实验如今已通过排除法表明，没有任何政治经济学上的改革——即便是完美无缺的改革——能够触及它。这是本文核心论点最强的形式，而它唯有通过将政治经济学的法庭所要求的一切尽数给予，方能获得。

9. 权力：亲密关系的建筑式宪法化

9.1 家庭之中权力的第三面

单一建造者配置的残余——无剥削的支配——如今获得了它的机制，而正是这一机制使其变得危险：它是隐含的。亲密生活中显性的权力不对称——谁赚钱、谁决定大额开支、谁在公开冲突中让步——是可见的、可争辩的、可协商的；它们的可见性正是重新协商得以进入的门户。构想的不对称则有着不同的存在方式。建造者的价值观并不以主张的形式出现在关系之中（主张是可以回应的），而是以默认设定的形式出现：作为提示的时机、某种语调的温度、各模块的边界，乃至于关系中何者被建模、何者不被建模这一清单本身。另一方伴侣并未遭遇一个她或可争议的立场；她栖居于一个已然被构成的空间之中。三项特征叠加强化了这种隐含性。该结构以善的形式运作：它并非控制意志的产物，而是爱与关怀的产物，而仁慈便是它的隐身斗篷，因为去争辩一项温柔的默认设定，便显得像是在争辩那份温柔本身，其情感代价之高，足以在争辩成形之前便将其封堵。它并不被体验为权力：被中介的一方在现象上所遭遇的，是该系统体贴周到，而非她生活在另一个人所立法的规范空间之内。而它塑造的是议程与偏好本身：以 Lukes (2005) 的术语而言，这是权力的第三个维度，超越了公开决策（第一维度），也超越了议程控制（Bachrach and Baratz 1962）——尽管它也行使后者，模块的边界决定了哪些关系性问题甚至能够浮现出来；它是对双方伴侣关于何为良好亲密关系之感觉的缓慢塑造，由其中一人所撰写的人工物所主导。

9.2 代码作为家庭之法

针对这样一种结构的理论框架是存在的，它为公共领域而建，而据作者所知，从未被移植到这一私人领域。Lessig (2006)：在数字中介的空间中，建筑（架构）与法律、市场和规范并列进行规制，且无需被感知为规制便能进行规制。Winner (1980)：技术安排本身即是政治安排，那座对公交车而言过低的桥梁，是一项没有记录在案之作者的政策。将二者移植到亲密领域，单一建造者的系统便显露出它的本来面目：亲密关系的建筑式宪法化，一部为这段关系而起草的宪法，由一方所制定，未赋予另一方任何修正权，且根本不以宪法的面目示人。前作所标举的那些设计哲学文献——其中每一条款单独来看都令人钦佩——在此视角之下，乃是一桩单方面制宪行为的物质记录：一个人关于”我们将如何被中介”的立法。这一移植还产生了一个关于构成性的”未言说”的推论，前作曾将这种”未言说”列为亲密关系的界定性特征之一：关系性契约（relational contract）的洞见（Macneil 1980）在于，持续性的关系由各方共同演化、从未完全言明的隐含规范所治理；而一个被建筑化的中介层，则将这片活生生的、共同漂移的规范的领域，转化为固定的、单方面撰写的、可执行的文本。这是一种对”未言说”的新型伤害：并非将其说出（前作所防范的正是这一点），而是替对方将其写下，没有她的手参与其中。

9.3 事后同意之薄弱，以及作为动词的”立宪”

有人会反驳说，披露与同意化解了这一难题：她得知了该系统、认可了它、并乐于使用它。这一反驳之所以失败，其缘由比技术本身更为古老。Hume 反对原初契约的论证（Hume 1987）是：在制度业已建成之后，由一个必须栖身某处的人所给出的同意，并不构成对那些制度的创作（authorship）：那个在船上醒来的水手，并不因为留下来便成了这艘船的设计者。对一种被建筑化的亲密关系的事后同意，是栖居，而非制宪权（constituent power）；立宪的时刻已然逝去，而它逝去时，其上仅有一个签名。这一点——而不仅仅是共同创造、共同建造之善——正是赋予§8.3中共同配置以其全部规范性分量的所在：它是唯一一种制宪权对称的配置，并将”立宪”从一个名词（即由一方交付的、已完成之物）转化为一个动词（即对共享中介层持续不断的共同制订），其修正权永不失效。必须毫不退缩地引出两项后果。其一，反身地：本文及其前作的试金石系统，就历史事实而言，是单一建造的；因而本节的分析，在有所衡量的部分上，控诉了那台正是本文得以从其审察中生发出来的装置本身，而支配这两篇论文的诚实规范要求将此点在正文中说出，而非于沉默中蒙混过去；所指示的补救之道并非拆毁，而是共同制宪——对那只起草之手的追溯性的、持续性的延伸。其二，教义地：§4.6中对家庭法之克制传统所给予的赞扬如今必须加以限定，因为对家庭内部的克制同时也是对家庭中第三维度权力的不予巡查，而建筑式宪法恰恰是司法克制所无法看见的那类结构。”个人即政治”这一经典的女性主义校正，在此获得了一种精确的技术内涵，而随之，分析便转向那个被亏欠发言权最久的法庭。

10. 女性主义的听审

10.1 亏欠，以及一次偿还

自本文的诸项标准最初被陈述以来，它便一直在向女性主义理论借贷：将”留意”置于关怀之首要时刻的做法出自 Tronto；将构成性情感作为可分配的输入加以处理的经验记录出自 Hochschild；坚持家庭的分配性面容必须被计入，则是家务劳动论争中的唯物主义女性主义。这笔账是该偿的，而第一笔偿还可以即刻作出，因为本文关于层压线（lamination-line）的分析（§7.2）是对一场女性主义于半个世纪前内部进行的论争的一个迟来的贡献。”家务有偿”运动（Federici 1975）要求家庭劳动应被计入、应付以工资、应作为劳动而被看见；其批评者，在运动内部，担忧为关怀定价会使其可通约（commensurate）从而败坏它；女性主义经济学随后建立起使不可见的关怀变得可计量的核算体系（Waring 1988; Folbre 2001）。以本文的词汇而言，这场论争是关于层压线划在何处的一桩争议：该运动正确之处在于，家务劳动的分配性面容是真实的，拒绝计入它本身即是一种政治，且这一拒绝有其受益者；批评者正确之处在于，构成性面容无法在通约之后存留。本文将分层分析回赠给那场辩论，作为一项标准上的了结：在同一个家庭之内，既将分配性面容上的一切尽数计入，又拒绝对构成性面容的化约，二者并行不悖。

10.2 性别化的预算：心理负荷、再分配，与那个伪造的答案

§7 的认知预算实际上并非性别中立的，而经验文献恰恰将这种偏斜定位在本文一直在考察的那个接合点上。Daminger (2019) 将家务认知劳动分解为预期（anticipation）、识别（identification）、决策（decision）与监控（monitoring），发现预期与监控阶段——也就是恰恰是”注意”的发端、家务与关系性照料之关注的构思阶段——不成比例地由女性承担；用本文的术语来说，这种心理负荷（mental load）正是 §7.3.3 所分析的那种预算管理功能的一种性别化分配。因此，温柔提示（tender tip）并非进入一片中立的场域。它进入的是一片早已被一项数十年之久的、长期存在的再分配诉求所结构化的场域：他应当承担属于他的那一份”注意”。而正是在此处，分析得出了它或许最尖锐的单一结论。女性主义的诉求是注意之再分配，从她转移到他；而机器所提供的是注意之替代，从他转移到它。二者在分配的层面上无从区分——无论哪种方式，她被测量出的负荷都减轻了；而在构成的层面上则截然不同：他终于注意到了，只不过注意到的并不是他。一项平等诉求被以伪造的货币所回应；§7.2 的分配性化约获得了它的性别化形式，而那四项标准也获得了一项进一步的职能，即区分真正的再分配（他承担起了”注意”，或许起初有所搭建支架，而在动态标准的良性分支上这种支架最终脱落）与对它的自动化模拟。诚实要求以同等的力度陈述其反面：在心理负荷压得人喘不过气、而再分配事实上无从获得之处，将该负荷中真正属于分配性的项目——清单、后勤、各项义务的日程——加以自动化，乃是真实的缓解，而非任何意义上的伪造，这正是 §7.3.1 的解放效应在做它本应做的事情；其中的错失是特定的，即在账面上显示为再分配之时，对构成性发端的替代，而这些标准的存在正是为了把它找出来。

10.3 谁来立法

§9 的制宪权（constituent power）同样不是靠掷硬币来分配的。建造、配置并维护那一中介层的能力，追随着技术资本的分配，而后者就目前而言、事实上是性别化的；家庭之宪法的设计者，在统计意义上的多数情形中将是男性伴侣，而亲密关系的建筑学式宪法化，在统计意义上的多数情形中将作为又一层叠加，落在一片早已倾斜的场域之上，其默认设置以另一方最难以修订的媒介，编码进了一方对照料的构想。这是一个关于趋势而非本质的论断，它使本节的处方更趋尖锐而非有所更改：共同制宪（§9.3）的理由，在现实世界中也是一个关于性别的理由；而其中所记录的那一反身性的承认，也应据此来理解，因为作者在本分析中的位置正是统计上典型的那一种。本法庭对 §4.6 所施加的纠正此前已经登录在案；用本法庭自己的声音重述一遍仍有必要：一种以不踏入家庭为荣的法理学，应当明白它由此拒绝去巡查的究竟是什么。

10.4 关于构成性层面的一种肯定性理论

迄今为止对构成性层面所做的每一项辩护都是否定性的：免于异化、免于泰勒制化（Taylorization）、免于贬抑、免于单方制宪。一直占据这片地域的传统，提供了本文仍然欠缺的东西，即一种肯定性的论述。在 Noddings (1984) 那里，关怀关系的根基是全神贯注（engrossment）：一种对具体他者的、接纳性的、非选择性的注意，是将自身的框架移置入她的框架之中；而这种注意并不是产生关怀所凭借的工具，它就是关怀本身，是该关系的存在方式而非其手段。Kittay (1999) 将同一论点奠基于依赖性的劳动之上：照料某一特定脆弱他者的工作，并不是一项恰好（令人遗憾地）难以委派的可替代服务；其不可替代性内在于它之所是。于此，§8.4 那一消除式的残余——那在每一种所有权配置下都得以幸存的错失——终于获得了它肯定性的名称，而本文最长的一道弧线也就此闭合：构成性层面无需经济学的任何豁免、无需法律的任何宽贷、也无需任何人的任何怀旧，因为它有它自己的法理学，即关怀伦理学，在其中，注意之不可委派并不是对爱之效率的一种限制，而是对爱之在场的定义。因此，女性主义的听审在整体的建筑结构中承担了双重职责：它审计了在先的每一个法庭——审计政治经济学中其预算的性别，审计权力分析中其制宪者的性别，审计法律中其克制的政治——并且它提供了那一肯定性的根基，§6 那以教义学的措辞起草的综合，一直以来正是悄然地站立在这一根基之上。

11. 蕴含

11.1 回交给前作论文的答案

前作论文曾追问：一段真正得到改善的关系，能否救赎一个被异化的过程，并将这一问题悬置。本文回交的答案是：不能；而且这个问题被微妙地误置了；而这两半都很重要。不能：因为改善与异化并不处于同一标度之上，救赎是一桩可通约者之间的交易，而过程缺陷——在根据和动态标准发现它之处——并不是一项可由结果加以抵偿的成本，而是结果之所是发生了改变。在亲密生活中，过程部分地构成了结果：”被爱”的内容包含了”注意的人是他”。在构成性环节被替代之处，”关系得到了改善”这一结果并非以一个令人遗憾的代价被买下；它已被悄然重新定义，二元体（dyad）成了三元体（triad），而那改善，虽然真实，却是对那被重新定义之物的改善，这正是为何主观与客观改善之被规定的统一从未真正了结此事。用 §7.2 所提供的词汇来说：改善的报告发自该行为的分配性层面，在那里一切的确都变得更便宜、更顺畅，而替代则发生在构成性层面，而任何取自前一层面的报告都无从对其加以审计；那个悬而未决的问题，在其深层结构上，乃是一份去成全某种分配性化约的邀请，而对这样一份邀请的正确回应，既不是接受也不是拒绝，而是揭露这种化约。后果主义的辩护并未输掉论证；只是它所评价的对象——一个独立于过程的结果——在这一领域中不复存在。揭露这种误置并不是对原问题的回避，而是对它的解决：那个问题所需要的不是一个答案，而是一项诊断，而辩证方法对其自身开局之举所下的裁断，乃是本文中最具黑格尔式的结果。

然而，前作论文在那四项约束之下，对其温柔提示模块所持的临时性容忍，在很大程度上得到了证立而非被撤销，因为那些约束实则原来就是处于胚胎状态的标准：选择性加入（opt-in）与低频保全了主动性（标准三）；仅限于场合与语域（occasion-and-register-only）则界定了中介的对象（标准一）；禁止将输出呈现为关于她心灵的知识，监管了归属（标准四）；可拒绝性与不予记录，使那只木筏被松松地握持着（动态标准）。本文所增添的，是关于为何恰是那些特定约束的理论，以及该理论所携带的警告：那些约束将该模块把守在例外域的边界之上，而非置于其外，而边界，在动态标准之下，并不是用来居住的地方，而是用来守望的地方。

11.2 亲密生活中同意的边界

双重错失的建筑结构给出了一个关于不可放弃之核心（unwaivable core）的确定立场。亲密生活确实包含这样一个核心，但它比浪漫主义的直觉更狭窄，又比自由主义的直觉更宽广：它所包含的，不是姿态、内容或技术，而是亲密注意之构成性环节——注意的发端、对那一特定他者的解释——作为关系中诸人所行使的能力。伴侣们可以同意对手段进行任何程度的中介；他们的同意完全治愈了归属性的错失；而他们的合意所无法做到的，是使那被替代的注意之发端成为他们的，正如立遗嘱人的同意无法使他人的意志成为他的遗嘱。这并非对恋人们的家长式作风；它与法律一向给予立遗嘱人和配偶的那一承认是同一种：某些行为是由其亲身的履行所构成的，而尊重诸人，包含了拒绝那种虚构——即他们可以在身处他处的同时履行那些行为。在社会基线的尺度上，同一建筑结构发出的是一项告诫而非一项禁令：零成本的、弥漫性的亲密中介，将把归属基线推向把它纳入其中——正如技术一向所为，并经由 §7.4 所识别的那种贬抑机制——而基线的移动将批量地治愈归属性的错失；但没有任何期望的移动能够触及那些根据，而一个其基线已将被替代的注意常态化了的社会，并未废除过程性的错失，而只是废除了它的可见性。本文将此登录为：旧有的浪漫主义对机器的抗议中，确有某种为真之物得以幸存，而其精确的、有界的意义正在于此。

11.3 辩证的保留

对方法的忠实要求综合自陈其非终极性。诸理据是分级的；基线是移动的；那一动态判准唯有跨越时间、且唯有在具体细节之中方才可辨。因此，此处所构建的例外领域并非一道围栏，而是一种管辖权：它将问题逐场所、逐姿态、逐年地，恒常地交付给前作所称的辩证-实证主义实践（dialectical-positivist practice）这一裁判庭——它进行度量、描述，并一再追问，却从不作出最终的裁决。后续诸节又向其待审案卷增添了第二桩恒常事务：中介层（mediating layer）自身的构成，正如 §9.3 所示，它绝不可能被一劳永逸地批准、此后即安然栖居其中，而必须始终是可共同修订的——构成（constitution）作为一个动词，起草之手永久地被共享。本文的贡献在于赋予了那一裁判庭一套可供适用的学理：四项判准、两个层次、一项动态检验、两种过错、两种救济，以及一个永不闭合的宪制性问题；而非在于免除其开庭审理之职。

12. 局限

兹陈述五项局限，绝不加以削弱。第一，动态判准的诸经验前提——即在弥漫性建议（ambient suggestion）之下注意力的去技能化（deskilling of attentiveness），以及在被牵引的使用之下能力的迁移（capacity migration under pulled use）——在此乃是通过理论予以推动（§3.3.1、§5.3.2），并借助于与关于感知与信任的”AI 中介传播”文献（Hancock et al. 2020; Jakesch et al. 2019）的类比而成立，但它们尚未就亲密二元关系（intimate dyads）得到确证；同样的条件性也附着于 §7 的两项政治经济机制，即姿态之价值形式（value-form）的贬值，以及对被解放出来的注意力作 Cowan 式的再吸收（reabsorption），这两者乃是经由理论推导而来的机制假说，在经验上仍属开放；本文对那一失败分支所下的诸般裁断，皆以纵向研究必须提供的事实为条件。第二，对严格人身性行为（strictly personal acts）的比较法考察，在本草稿中只是一份依托于少数经核验之锚点的素描；本综合所应得的六法域考察被推迟至下一次修订，在那之前所有的法条引证均带有核验标记。第三，马克思主义的理论装置乃是按既经声明的选择，在其哲学-人类学的谱系之内被加以运用；那些主张异化（alienation）在政治经济学批判之外即无意义的读者，将会发现反题（antithesis）只是被重新描述、而非被驳倒，本文已为这一结构性移植的正当性作了论证，但无法予以强制。第四，佛教材料取自巴利文（Pli）经藏的译本，并以哲学的方式而非语文学的方式被运用；本文所主张的是结构上的趋同（convergence of structure），而非训诂上的完备性，至于来自儒家与道家传统的平行考察——其中每一者都直接关涉本问题，且都是被有意保留下来的——则仍属未来的工作。第五，本分析建立在前作所刻画的二元亲密关系之上；将其延伸至亲子之间、为人父母者，以及友谊的亲密关系——在那里，角色界定与不对称性会改变诸般理据——在此尚未予以考察，且可能并不保守。第六，相对于前三个裁判庭，后三个裁判庭只是在素描的深度上被召集：消去式检验（eliminative test）是就理想化的格局进行推理，并从混合的、迁移中的所有权形式中抽象而出；权力分析移植了为公共建制而构建的诸框架，而尚欠一份对亲密关系内部的抵抗、退出与重新协商的更充分的说明；女性主义听证则将其经验前提建立在某一文献之上——以 Daminger (2019) 为例——该文献主要取自单一民族脉络下的当代异性恋伴侣，因而它关于认知预算（cognitive budget）之性别化与构成性权力（constituent power）之性别化的诸般主张，乃是关于该范围之内已被记录的统计倾向的主张，而非关于本质的主张，且其跨文化、跨关系形式的普遍化乃是一个开放的经验问题。

13. 结论

Roxane 并没有弄错她是被爱着的；她弄错的是那爱在何处被施行，而这出戏的悲恸正在于：第二个错误毒害了第一个真理。本文已力图以六个裁判庭——辩证的、法律的、佛教的、政治经济的、宪制的与女性主义的——之精确，去言明那是一种什么样的错误，以及它那现代的、弥漫性的、可规模化的后裔何时重蹈其覆辙。中介本身是无辜的：精神、爱与法律皆由它构成。过错，凡其存在之处，栖居于一个微小而可指名的位置：在于对亲密之关注（intimate attending）的诸构成性环节——即注意的发端、对那个特定他者的诠释——的替代，这种替代施行于关系自身的内在沉积之上，并被置于一道仍将此类姿态读作人身性的基线之前；它又分裂为一种归属之过错（wrong of attribution），诚实可治愈之，以及一种过程之过错（wrong of process），唯有能力回归其行动主体方可治愈之——筏被放下，辅助轮被卸去，爱者被复归于他自己的关注。消去式实验表明，这一过错得以幸存于所有权的每一项改革之外，包括那完美的改革；宪制分析表明，即便是那无辜的余项，也必须被共同地、永久地起草，家之代码须由每一只必须栖居其下之手来撰写；而关怀伦理（ethics of care）则为这一切供给了肯定性的根基，即：关注并非爱的手段，而是爱的存在样态。在工具性学说所要求的普遍无罪开释，与浪漫主义抗议所欲求的普遍有罪定谳之间，真理结果竟是最古老意义上的学理性的：一条默认规则、一个例外领域、若干判准，以及一个永不休庭的裁判庭。

致谢与披露

作者首先要致谢的是一位他深爱的女子：如森林一般，纯净、智慧、善良且利他；于他而言，她是、并将永远是他所爱的那个”森林女孩”。能与她相遇，既是奇迹亦是幸运，而本文的思考也是为她而作。愿一切相爱之人都能找到彼此，幸福而幸运，白头偕老。,,

作者披露，在本文的起草与编辑中曾使用一位 AI 助手，包括在作者的指导下、跨越多个迭代会话发展出各种表述。各种构想、所采取的辩证立场，以及一切最终判断，皆出自作者，作者对内容（包括错误在内）负全部责任。本文在分析 AI 对亲密关系的中介的同时，自身也是借助 AI 中介而起草的；作者认为这一做法与本文自身的判准相一致——学术性的起草并非本文所定义之严格人身性行为，而这一披露本身正是第四项判准所要求的。

参考文献

• Althusser, Louis (2005). For Marx. Translated by Ben Brewster. French original 1965. London: Verso.
• Aristotle (2009). The Nicomachean Ethics. Translated by David Ross, revised by Lesley Brown. Praxis and poiesis distinguished at VI.4–5. Oxford: Oxford University Press.
• Bachrach, Peter and Morton S. Baratz (1962). “Two Faces of Power”. In: American Political Science Review 56.4, pp. 947–952.
• Becker, Gary S. (1965). “A Theory of the Allocation of Time”. In: The Economic Journal 75.299, pp. 493–517.
• — (1981). A Treatise on the Family. Enlarged edition 1991. Cambridge, MA: Harvard University Press.
• Bodhi, Bhikkhu, ed. (2012). The Numerical Discourses of the Buddha: A Translation of the Aguttara Nikya. Cetan as karma at AN 6.63 (Nibbedhika Sutta). Boston: Wisdom Publications.
• Braverman, Harry (1974). Labor and Monopoly Capital: The Degradation of Work in the Twentieth Century. New York: Monthly Review Press.
• Buddharakkhita, Acharya, ed. (1985). The Dhammapada: The Buddha’s Path of Wisdom. Verse 165 on the non-transferability of purification. Kandy: Buddhist Publication Society.
• Chapman, Gary (1992). The Five Love Languages: How to Express Heartfelt Commitment to Your Mate. Cited as popular psychology, with the qualifications stated in the mother paper. Chicago: Northfield Publishing.
• Cowan, Ruth Schwartz (1983). More Work for Mother: The Ironies of Household Technology from the Open Hearth to the Microwave. New York: Basic Books.
• Daminger, Allison (2019). “The Cognitive Dimension of Household Labor”. In: American Sociological Review 84.4, pp. 609–633.
• Delphy, Christine (1984). Close to Home: A Materialist Analysis of Women’s Oppression. Translated by Diana Leonard. — VERIFY edition —. London: Hutchinson.
• Federici, Silvia (1975). Wages Against Housework. See also Dalla Costa and James, The Power of Women and the Subversion of the Community, 1972. Bristol: Falling Wall Press.
• Folbre, Nancy (2001). The Invisible Heart: Economics and Family Values. New York: The New Press.
• Fricker, Miranda (2007). Epistemic Injustice: Power and the Ethics of Knowing. Oxford: Oxford University Press.
• Fromm, Erich (1961). Marx’s Concept of Man. New York: Frederick Ungar.
• Fuchs, Christian (2014). Digital Labour and Karl Marx. New York: Routledge.
• Han, Byung-Chul (2015). The Burnout Society. Translated by Erik Butler. German original Müdigkeitsgesellschaft, 2010. Stanford: Stanford University Press.
• Hancock, Jeffrey T., Mor Naaman, and Karen Levy (2020). “AI-Mediated Communication:

Definition, Research Agenda, and Ethical Considerations”. In: Journal of Computer-Mediated Communication 25.1. — VERIFY page range —, pp. 89–100.

• Hart, H. L. A. and Tony Honoré (1985). Causation in the Law. 2nd ed. Oxford: Oxford University Press.
• Hegel, Georg Wilhelm Friedrich (1977). Phenomenology of Spirit. Translated by A. V. Miller. Original work published 1807. Oxford: Oxford University Press.
• Hochschild, Arlie Russell (1983). The Managed Heart: Commercialization of Human Feeling. Berkeley: University of California Press.
• Honneth, Axel (2008). Reification: A New Look at an Old Idea. The Berkeley Tanner Lectures, 2005, with commentaries by Judith Butler, Raymond Geuss, and Jonathan Lear. Oxford: Oxford University Press.
• Huang, Wanhong (2025a). Seven Infrastructural Domains for Relational Being. OSF preprint. — VERIFY: OSF DOI —.
• — (2025b). Toward Relational Being in Intimate Life: Integrative Justice and an Ethics of Dialectical­Positivist Practice. A Family-and-Relationship Management Application as a Case Study. SSRN working paper. — VERIFY: SSRN ID and posting date —.
• Hume, David (1987). “Of the Original Contract”. In: Essays Moral, Political, and Literary. Ed. by Eugene F. Miller. Essay first published 1748. Indianapolis: Liberty Fund.
• Jaeggi, Rahel (2014). Alienation. Translated by Frederick Neuhouser and Alan E. Smith. German original 2005. New York: Columbia University Press.
• Jakesch, Maurice et al. (2019). “AI-Mediated Communication: How the Perception that Profile Text was Written by AI Affects Trustworthiness”. In: Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems. — VERIFY paper number/pages —. ACM.
• Kittay, Eva Feder (1999). Love’s Labor: Essays on Women, Equality, and Dependency. New York: Routledge.
• Lessig, Lawrence (2006). Code: Version 2.0. New York: Basic Books.
• Lukács, Georg (1971). History and Class Consciousness: Studies in Marxist Dialectics. Translated by Rodney Livingstone. Original work published 1923. Cambridge, MA: MIT Press.
• Lukes, Steven (2005). Power: A Radical View. 2nd ed. — VERIFY: consider citing the 3rd edition, 2021 —. Basingstoke: Palgrave Macmillan.
• Macneil, Ian R. (1980). The New Social Contract: An Inquiry into Modern Contractual Relations. New Haven: Yale University Press.
• Marx, Karl (1973). Grundrisse: Foundations of the Critique of Political Economy. Translated by Martin Nicolaus. Manuscripts 1857–1858; the “Fragment on Machines” at pp. 690–712. Harmondsworth: Penguin.
• — (1975). “Economic and Philosophic Manuscripts of 1844”. In: Marx–Engels Collected Works. Vol. 3. Original manuscripts 1844. London: Lawrence & Wishart, pp. 229–346.
• — (1976a). Capital: A Critique of Political Economy, Volume I. Translated by Ben Fowkes. Original work published 1867. Harmondsworth: Penguin.
• — (1976b). Results of the Immediate Process of Production. Appendix to Capital I, Penguin edition, pp. 948–1084; on formal and real subsumption. Manuscript 1863–1864. Harmondsworth: Penguin.
• — (1989). “Critique of the Gotha Programme”. In: Marx–Engels Collected Works. Vol. 24. Written 1875. — VERIFY page range —. London: Lawrence & Wishart.
• Mauss, Marcel (1990). The Gift: The Form and Reason for Exchange in Archaic Societies. Translated by W. D. Halls. Original work published 1925. London: Routledge.
• Ñāṇamoli, Bhikkhu and Bhikkhu Bodhi, eds. (1995). The Middle Length Discourses of the Buddha: A Translation of the Majjhima Nikāya. The simile of the raft at MN 22 (Alagaddūpama Sutta). Boston: Wisdom Publications.
• Nissenbaum, Helen (2009). Privacy in Context: Technology, Policy, and the Integrity of Social Life. Stanford: Stanford University Press.
• Noddings, Nel (1984). Caring: A Feminine Approach to Ethics and Moral Education. Berkeley: University of California Press.
• Nussbaum, Martha C. (2000). Women and Human Development: The Capabilities Approach. Cambridge: Cambridge University Press.
• Ollman, Bertell (1971). Alienation: Marx’s Conception of Man in Capitalist Society. Cambridge: Cambridge University Press.
• Robbins, Lionel (1932). An Essay on the Nature and Significance of Economic Science. London: Macmillan.
• Roemer, John E. (1982). A General Theory of Exploitation and Class. Cambridge, MA: Harvard University Press.
• Rostand, Edmond (1897). Cyrano de Bergerac. English citations from any standard translation; the balcony scene at Act III. Paris: Charpentier et Fasquelle.
• Roxin, Claus (2006). Strafrecht Allgemeiner Teil, Band I: Grundlagen, der Aufbau der Verbrechenslehre. 4th ed. On objektive Zurechnung. — VERIFY edition cited —. München: C. H. Beck.
• Sen, Amartya (1999). Development as Freedom. Oxford: Oxford University Press.
• Simon, Herbert A. (1971). “Designing Organizations for an Information-Rich World”. In: Computers, Communications, and the Public Interest. Ed. by Martin Greenberger. Baltimore: Johns Hopkins University Press, pp. 37–72.
• Spence, Michael (1973). “Job Market Signaling”. In: The Quarterly Journal of Economics 87.3, pp. 355–374.
• Tronto, Joan C. (1993). Moral Boundaries: A Political Argument for an Ethic of Care. New York: Routledge.
• Walshe, Maurice, ed. (1995). The Long Discourses of the Buddha: A Translation of the Dīgha Nikāya. “Be islands unto yourselves” at DN 16 (Mahāparinibbāna Sutta). Boston: Wisdom Publications.
• Waring, Marilyn (1988). If Women Counted: A New Feminist Economics. San Francisco: Harper & Row.
• Winner, Langdon (1980). “Do Artifacts Have Politics?” In: Daedalus 109.1, pp. 121–136.
• Zelizer, Viviana A. (2005). The Purchase of Intimacy. Princeton: Princeton University Press.
• Zuboff, Shoshana (2019). The Age of Surveillance Capitalism: The Fight for a Human Future at the New Frontier of Power. New York: PublicAffairs.