Toward Relational Being in Intimate Life - Integrative Justice and an Ethics of Dialectical-Positivist Practice

A bilingual deposit (English · 中文) of the SSRN working paper, with the full PDF of each language embedded, followed by the complete Markdown text of both.

English Version (PDF)

中文版（PDF）

---

English — Full Text

A Family-and-Relationship Management Application as a Case Study

Wanhong Huang
huangwanhong.g.official@gmail.com
2026-06-03

---

To the one I love deeply, a fellow native of the homeland we both left behind, yet whom I was to meet, by some miracle, for the first time only in a country far from the one that bore us.

两情若是久长时，又岂在朝朝暮暮
If two hearts are to hold each other for a lifetime, why should they grieve to be apart from one morning, one evening, to the next?
— after Qin Guan (秦观, 1049–1100)

但愿人长久，千里共婵娟
May we both be granted long years, and though the whole world lie between us, may we look up, each from afar, upon the same bright moon.
— after Su Shi (苏轼, 1037–1101)

---

Abstract

Positivist, evidence-based technologies increasingly mediate intimate life: couples coordinate, record, and remember through software whose default logic is measurement and optimization. This paper asks how such technologies ought instead to be governed if their aim is the flourishing of a relationship, the sustaining of what it calls relational being in intimate partnership.

It makes four kinds of claim. (1) Normative: an integrative-justice framework, resting on a relational ontology, draws on a dozen ethical and epistemic traditions and assigns each the questions it is competent to answer, in a defended ordering rather than a numerical average. (2) Practical: an ethics of dialectical-positivist practice warrants measurement differentially, site by site, according to whether a phenomenon is constituted partly by remaining unmeasured. (3) Engineering: a working application, designed by the author, instantiates the framework in software, with event-driven household coordination, a witness-only AI layer, co-creative exploration, and the deliberate absence of AI where it would intrude. (4) Diagnostic: it names wrongs specific to intimate computing, among them proxy epistemic injustice, the misattribution of an act of knowing, and the opposition between structural efficiency and relational development.

The limits are explicit. (i) Mediation’s deepest harms, relational alienation and epistemic distortion, can be reduced but never wholly removed, so the framework offers harm-reduction and a discipline of restraint. (ii) Its dialectical adjudication is a practice of judgement for which no algorithm is given. (iii) The evidence is a single, author-designed case, offered as illustration and existence proof. The framework is meant, finally, to bear on adjacent settings as well: the ethics of the societal use of data, and the design of mechanisms linking the family to public institutions.

Keywords: relational being; intimate computing; integrative justice; the ethics of dialectical-positivist practice; generative presence; data ethics; the ethics of flourishing

---

Introduction

A growing share of intimate life is now conducted through, or recorded by, software. Couples coordinate on shared calendars, keep joint ledgers, log meals and moods, and increasingly delegate to conversational agents the small administrative friction of a shared existence. The literature that examines this development has been overwhelmingly either empirical, concerned with how couples in fact use such tools, or design-methodological, concerned with how such tools should be built. The normative question, how ought the positivist, evidence-based mediation of intimate life be governed and toward what end, has received comparatively little sustained philosophical treatment. This paper is an attempt at that treatment.

The end proposed here is not efficiency, nor wellbeing as conventionally operationalized, but relational flourishing: the sustaining over time of a relationship understood, after the relational-ontological tradition, as ontologically prior to the individuals it relates. The paper restricts this large idea deliberately. It does not advance a general theory of relational being; it asks only what relational being requires in the specific domain of intimate partnership, and what follows for the technologies admitted into that domain. The restriction matters because intimate partnership has features that distinguish it from the family at large, the workplace, and the polity, and that bear directly on which forms of measurement and mediation are admissible.

The paper’s central constructive proposal is an integrative-justice framework. By this is meant a way of bringing several normative traditions (evidentialism in epistemology, the positivism–anti-positivism dialectic in the philosophy of the human sciences, deontological ethics, feminist care ethics, Confucian role-ethics and Daoist non-interference, and the capability approach to flourishing) to bear jointly on the ethics of intimate data, in such a way that each tradition is assigned the questions it can competently answer and denied the questions it cannot. The integration is not eclectic averaging; it is a division of normative labour. Evidentialism governs what the system is entitled to claim to know; deontology fixes the inviolable constraints on the treatment of the other; care ethics governs the texture of responsiveness; the capability approach supplies the criterion of flourishing against which the whole is measured; and the East Asian sources supply an account, largely absent from the Western canon, of the positive value of restraint and non-action.

From this framework the paper derives a practical posture it calls an ethics of dialectical-positivist practice. The positivist holds that domestic phenomena are legitimately quantifiable and therefore improvable through measurement. The anti-positivist, drawing on Diltheyan, Weberian, Geertzian, and Frankfurt-school sources, denies that the constitutive phenomena of intimate life submit without distortion to such operationalization. The posture defended here treats this opposition dialectically: measurement is neither universally licensed nor universally corrosive, but warranted differentially, site by site, according to whether the phenomenon at issue is partly constituted by not being measured. The dialectic is not resolved once and for all; it is re-adjudicated at each site of possible measurement, and the adjudication itself is treated as a first-order ethical act.

To keep the argument from floating free of practice, the paper takes as a sustained case study a working family-and-relationship management application that the author designed and maintains: roughly two dozen interlocking modules, among them a shared journal, a household ledger, a health log, a synchronized calendar, a conversational AI layer, and an etiquette aid, each of which embodies, in concrete software, a determinate answer to the question of where on the dialectical continuum it should sit. The application is used here not as the object of a design-research report but as an extended worked example: a place where the framework’s abstract commitments can be seen taking on operational form, and where their costs and tensions become visible.

One feature of the case is methodologically and ethically unusual, and the paper takes it up as a philosophical instrument. The application has been built for an intimate relationship in which the other party is, at the time of writing, unaware of the application’s existence. Designing for an absent or unknowing other is a limiting case that sharpens every question a relational ethics must answer: What may one party legitimately decide on behalf of a relationship the other has not been consulted about? When does preparation for another’s good become presumption upon it? What is owed to a person who is the subject of a system they have not consented to? The paper argues that the integrative-justice framework yields determinate and sometimes uncomfortable answers to these questions, answers that constrain the project significantly, and that a framework which could not address the absent-other case would be inadequate to the ordinary case as well.

The paper is accordingly organized around three questions. The first concerns normative foundations: what does relational flourishing, in intimate partnership specifically, require of the technologies admitted into that relationship, and which normative traditions are competent to specify those requirements? The second concerns integration: how may evidentialist, positivist and anti-positivist, deontological, care-ethical, Confucian-Daoist, and capability-theoretic considerations be integrated into a single coherent practice rather than left as competing and incommensurable demands? The third concerns the limiting case of the absent other: what does this framework imply for the case of designing for an intimate other who is unaware of, and has not consented to, the design, and what does that limiting case reveal about the ordinary one?

The paper’s contributions are correspondingly philosophical rather than technical, and may be stated as follows: (i) a restriction and specification of the relational-being thesis to the domain of intimate partnership (§2); (ii) an integrative-justice framework for the ethics of intimate data, with each constituent tradition given an explicit competence and an explicit limit (§3–§4); (iii) the articulation of an ethics of dialectical-positivist practice as the framework’s practical upshot (§5); (iv) a worked case study showing the framework instantiated in a real system (§6); (v) the identification of wrongs specific to intimate computing, including proxy epistemic injustice, the misattribution of an act of knowing, and the opposition between structural efficiency and relational development (§6.7, §8); (vi) a treatment of the absent-other problem as a test of relational ethics in general (§7); and (vii) a set of implications reaching beyond intimate partnership, for the ethics of the societal use of data and the design of mechanisms linking the family to public institutions (§10). Questions of software engineering, such as architecture, storage, and choice of frameworks, are treated only insofar as they bear on the normative argument, and are confined to a brief appendix; they are not the subject of this paper.

The remainder proceeds as follows. §2 specifies relational being for intimate partnership and situates its intellectual lineage. §3 sets out the constituent normative traditions, one per subsection. §4 argues for their integration. §5 develops the dialectical-positivist posture. §6 presents the case study. §7 takes up the absent other. §8 states limits. §9 places the paper within the surrounding literatures, and §10 concludes.

Relational Being in Intimate Partnership

Intellectual Lineage

The ontological commitments developed in this section are post-phenomenological in the lineage of , who reads artefacts as mediating rather than merely containing the practices around them, and relational in the lineages of , for whom the I-Thou relation is ontologically primary, and , for whom the self is constituted in its roles and relations, and , who develops the concept of focal practices whose value is intrinsic and erodable by instrumentalization. What this paper adds is a restriction of these large ontological claims to the specific domain of intimate partnership, and an argument that the restriction changes what the ontology demands. A fuller placement of the paper within the literatures of intimate technology, self-quantification, and autobiographical design is deferred to §9, after the framework it positions has been set out.

The Ontological Ground: Relationality and Generative Presence

The framework rests on a relational ontology, which the present author has developed at length elsewhere and which converges with a wide tradition: Buber’s primacy of the I-Thou relation , Gergen’s relational being , Whitehead’s process metaphysics of relational “occasions of experience” , Heidegger’s Mitsein or being-with , the Buddhist doctrine of dependent origination and emptiness , and Confucian role-ethics . The shared claim is that the basic units of being are not substances but relations: the self is not a self-standing atom that subsequently enters relations, but an effect that becomes perceptible only as it enters a relational structure. A useful formalization is algebraic: a group is defined not by the intrinsic nature of its elements but by the operation that relates them, so that prior to instantiation only the field of potential relations exists, not a collection of self-subsistent entities . On this view, what is primary is the relation, not the relata.

This ontology carries a consequence that the rest of the paper depends on. If the self is constituted relationally, then no symbolic structure, no record, no representation, ever coincides with the relational being it refracts; each captures the relation only partially and at the cost of a remainder it cannot hold. The author’s prior work names the lived, generative counterpart to this structural insight generative presence: the procedural and phenomenological mode through which relational being becomes active in experience, the manner in which existence manifests through practice, attention, and responsiveness rather than through representation . Presence contrasts with representation as enactment contrasts with abstraction: representation abstracts, presence enacts. The pairing is dialectical, since relation without presence collapses into abstract structure while presence without relation collapses into solipsistic immediacy. For an intimate technology this distinction is foundational. A system that traffics only in representation, in records, dashboards, and summaries, operates at the level of the symbolic and necessarily misses the generative presence in which the relation actually lives; the design question becomes how a system might make room for presence rather than substituting ever-finer representation for it.

The same ontology yields a precise account of what an artificial conversational agent is and is not, an account on which §6.8 and the case study later draw. Human and collective subjects are grounded in a pre-symbolic relational reality; an AI language layer is not. Its operations are confined to the symbolic, generated by conditional probability over prior symbols, and it therefore exhibits what the author’s prior work calls a pseudo-presence: it manifests within the symbolic as if it were a subject, while lacking the ontological grounding in relational being that genuine presence requires . This is neither praise nor dismissal of such systems but a placement of them: an AI in an intimate setting can mediate, prompt, and witness, but it cannot itself be the relational other, and a design that forgets this stages a counterfeit of presence where presence is what was wanted.

The Thesis and Its Restriction

The relational thesis, in its full generality, ranges over all relations whatever. This paper restricts it to the domain of intimate partnership and asks what, in that domain specifically, it requires. On the relational view the relevant unit of ethical concern in an intimate partnership is the relation itself, which has a reality and a trajectory of its own, rather than two individuals and the contract between them.

The restriction is consequential. A great deal that is true of relational being in the family at large, the workplace, or the polity does not transfer to intimate partnership, which is distinguished by at least three features. First, it is non-role-defined: unlike the parent–child or employer–employee relation, it is not structured by an antecedent distribution of authority, so that no party may invoke a role to justify deciding unilaterally on the other’s behalf. Second, it is constituted by voluntary vulnerability: each party discloses what they are under no obligation to disclose, and the relation’s depth is a function of that uncompelled disclosure. Third, it depends constitutively on the unsaid: a portion of what sustains the relation is precisely what is left unmeasured, unrecorded, and unspoken, so that exhaustive articulation would deplete rather than enrich it. This third feature is the intimate-partnership instance of the general ontological point that no representation coincides with relational being: the remainder that symbolization cannot reach is, in the intimate case, constitutive of the bond.

The Vulnerability of Intimate Partnership to Mediation

Each of these three features is directly vulnerable to the wrong kind of technological mediation. Non-role-definition is threatened by any system that silently arrogates decision authority to one party. Voluntary vulnerability is threatened by any system that converts disclosure from a gift into a metric. The constitutive unsaid is threatened by any system whose default is to record, surface, and summarize. The general diagnosis offered in the author’s prior work applies here with particular force: as symbolic mediation proliferates, the gap between the symbolic and the relational Real widens, and the I-Thou relation erodes into an I-It relation conducted through interfaces that do not respond as subjects . A technology admitted into an intimate partnership cannot, therefore, be evaluated solely by whether it performs its function efficiently; it must be evaluated by what it does to the three features that make the relation the kind of thing it is, and by whether it widens or narrows the gap between symbolic mediation and relational presence. This is the criterion against which the case study in §6 is assessed.

An Integrative-Justice Framework

This section sets out the normative traditions the paper brings to bear on the ethics of intimate data. The exposition of each follows a fixed pattern: an account of the tradition’s core claims and principal sources, developed at enough length to do the tradition justice; the question about intimate data over which the tradition is competent; the question it is not competent to answer and must cede to another tradition; and, where useful, an anticipation of how the tradition bears on the case study of §6. The traditions are presented in an order that is itself argued for: a methodological overframe, dialectics, first, then the epistemic traditions that govern what may be known and claimed, then the substantive ethical traditions that govern what may be done, and finally the political-economic tradition that governs where value comes to rest. The claim that the result is an integration rather than an eclectic heap is defended in §4.

One orienting commitment runs through the whole and deserves statement at the outset, since it shapes how every tradition below is read. The positive aim of the system, beneath all the constraints and refusals the framework imposes, is eudaimonic in the sense developed in §3.11: not the optimization of states but the provision of a generative field, a space in which the partners’ shared self-realization can unfold of its own movement. The design philosophy common to the system’s modules is to furnish such a field and then to withdraw, rather than to drive the relation toward any measured outcome; the constraints of the framework mark the boundaries of the field, and the criterion of flourishing judges whether the field is, in fact, generative. This commitment connects the developmental, eudaimonic, and capability-theoretic traditions below into a single positive orientation, against which the negative work of the constraints is to be understood.

Dialectics: The Method of the Whole

Before any single normative tradition is introduced, a word is needed about the form the integration will take, for that form is itself a philosophical commitment. The framework developed here is dialectical in a sense that descends from Hegel and is sharpened, for present purposes, by Adorno. The minimal Hegelian thought is that a concept, pressed hard enough, generates its own opposite, and that understanding advances not by choosing between the two but by grasping the more comprehensive structure in which each is a one-sided moment . Applied to the ethics of intimate data, the dialectical method refuses at the outset the two stances that dominate public discussion: techno-optimism, which holds that more measurement is more knowledge and therefore more good, and techno-romanticism, which holds that measurement as such desecrates what it touches. Each of these, the dialectical method insists, is a one-sided moment of a truth neither can state alone: measurement both reveals and deforms, and which it does depends on what is measured and how.

The Hegelian inheritance must, however, be qualified in one decisive respect, and here Adorno is the better guide. Classical dialectics aims at Aufhebung, a synthesis that sublates and preserves the opposed moments in a higher unity, and there is a standing temptation to treat such synthesis as a final resting point at which the tension is dissolved. Adorno’s negative dialectics resists exactly this temptation: it holds that the reconciling synthesis is often a premature peace that suppresses a contradiction still really present in the object, and that thought does better to dwell in the contradiction than to paper over it . The relevance to intimate data is direct. There is no stable synthesis at which the claims of intimacy and the claims of coordination are permanently reconciled; the diary will always resist the dashboard, and the dashboard will always be useful. The framework therefore does not seek a formula that, once found, ends the argument. It seeks a practice of perpetual re-adjudication, in which the contradiction between measuring and refraining is held open and decided afresh at each site, never abolished. The dialectical commitment is thus the deepest layer of the framework: it is the reason the framework is a procedure rather than a rule, and the reason §5 can speak of an ethics that is “never finally settled.”

The competence of dialectics is methodological and total: it governs not any particular question about intimate data but the manner in which all the questions are to be held together. Its limit is the obverse of this competence: precisely because it is a form rather than a content, dialectics specifies no determinate prohibition or permission on its own. It tells us that measurement reveals and deforms; it does not tell us, of this diary entry, which it does. For that, the substantive traditions are required.

Evidentialism: The Limits of Warranted Assertion

Evidentialism is the epistemological thesis that the justification of a belief is wholly a function of one’s evidence for it: a belief is justified to the precise degree that the evidence supports it, and to believe beyond the evidence is to believe without warrant . In its classic moralized form the thesis acquires an ethical edge: argued, in a phrase whose force survives its Victorian setting, that it is wrong always and everywhere to believe anything on insufficient evidence, on the ground that beliefs are not private but feed into shared action and so carry a responsibility to the community that will act on them. The contemporary refinement strips away the universal moralism while keeping the core: justification supervenes on evidence, and a system or person that asserts more than the evidence licenses commits an epistemic wrong even when, by luck, the assertion is true.

Transposed to a data system that mediates intimate life, evidentialism governs the modality of everything the system asserts: the confidence, the precision, and the hedging with which its outputs are framed. The point is easily underestimated because the violations are so familiar as to seem natural. A health module that infers caloric intake from a free-text description of a meal has evidence, at best, for a wide range; to display a single confident integer is to assert vastly more than the evidence supports, laundering a guess into a fact through the typographic authority of a clean number on a clean screen. A module that estimates a mood from the lexical content of a diary entry has, at best, weak probabilistic evidence about a fabulously complex inner state; to label the day with an emoji is to commit, in miniature, exactly the Cliffordian wrong. Evidentialism’s demand is therefore not that such estimates never be computed but that they be expressed honestly: ranges rather than points, “roughly” rather than “is,” the visible marking of an inference as an inference. The competence of evidentialism is over epistemic warrant: it tells the system what, given its evidence, it is entitled to claim, and it forbids the false precision by which interfaces routinely overstate their knowledge.

Evidentialism’s limit is equally sharp, and naming it motivates the very next tradition. Evidentialism is a theory of how to fit belief to evidence; it is silent on what ought to be inquired into at all, and silent on who counts as a credible source. It can rule that a sentiment score, once computed, must not be asserted beyond its warrant; it cannot rule that computing a sentiment score over a beloved’s diary is a wrong even when the computation is impeccable, nor can it detect the subtler wrong of systematically crediting one partner’s data-borne self-report over the other’s spoken word. The first of these belongs to substantive ethics, addressed below. The second belongs to a tradition that extends epistemology into the domain of justice, to which we turn immediately.

Epistemic Injustice: Credibility and Interpretive Fairness

Evidentialism asks whether belief fits evidence; it does not ask whether the social machinery of credibility is fair. That question is the subject of Miranda Fricker’s account of epistemic injustice, a wrong done to someone specifically in their capacity as a knower . Fricker distinguishes two forms. Testimonial injustice occurs when a hearer, under the influence of prejudice, gives a speaker less credibility than their word deserves: the speaker’s testimony is discounted not because it is unreliable but because of who they are taken to be. Hermeneutical injustice is deeper and more structural: it occurs when a gap in the collective resources of interpretation leaves some person or group unable to render an important part of their own experience intelligible, because the concepts that would make it sayable have not been developed or have been developed only from a dominant standpoint. The paradigm cases are social, but the structure transfers with unsettling precision to intimate data systems.

A system commits something closely analogous to testimonial injustice whenever it positions one channel of evidence as authoritative and thereby silently downgrades another. Consider a system that infers a partner’s wellbeing from sleep-tracker and activity data and presents the inference with quiet confidence; when the partner says, in words, that they are in fact unhappy, the architecture has already established the sensor stream as the credible witness and the spoken word as soft, unreliable, anecdotal. The person has been wronged as a knower of their own life: the system has arrogated to its instrumentation an authority over the person’s experience that belongs, in an intimate relation between equals, to the person. Worse still is the hermeneutical case. Much of what matters most in intimate life, ambivalence, the slow alteration of feeling, the kind of sadness that is also a form of love, has no clean operationalization; a system whose categories admit only what its sensors can register does not merely fail to capture these experiences but actively crowds out the interpretive space in which they might be articulated, training its users to recognize in themselves only what the schema can hold. This is hermeneutical injustice enacted in software: the impoverishment of a person’s means of understanding themselves, brought about by the dominance of a measuring vocabulary.

The competence of the epistemic-injustice tradition is therefore over a question neither evidentialism nor the substantive ethical traditions can pose in its own terms: whether the system, in the very structure of how it gathers and weights evidence, treats each partner fairly as a knower, both of the shared world and of themselves. It supplies a positive design imperative, that the person’s first-personal testimony retain standing against the machine’s inference, and a negative one, that the system not narrow the interpretive resources through which its users can understand their own lives. The case study will identify a third variant, distinct from the two in Fricker’s original taxonomy, in which the system performs an act of knowing or remembering on a partner’s behalf in such a way that the other is led to misattribute its source (§6.7); this proxy epistemic injustice, as the paper will call it, is a distinctively technological wrong worth marking. The limit of the tradition is that it diagnoses such wrongs without, by itself, fixing the inviolable boundaries whose breach constitutes the gravest wrongs; for those boundaries we need deontology, and for the account of the affective texture that hermeneutical generosity requires we need the ethics of care. Both are below.

Positivism and Anti-Positivism: The Question of Measurability

The opposition between positivism and anti-positivism is the axis along which the whole question of measurement turns, and it deserves fuller statement than the slogans usually allow. The positivist tradition in the human sciences, descending from Comte and consolidated in twentieth-century empirical social science, holds that social and psychological phenomena are in principle measurable, that the methods which succeeded in the natural sciences are the model for knowledge as such, and that measurement is the precondition of rational improvement: what can be counted can be compared, what can be compared can be optimized. Its appeal in the domestic setting is real and should not be caricatured. A household genuinely is, in part, a site of logistics, expenditure, and scheduling, and for that part the positivist promise holds: measurement clarifies, and clarity reduces the friction of shared life.

The anti-positivist or interpretivist counter-tradition is not a single doctrine but a convergent family of objections. Dilthey drew the foundational distinction between Erklären, the causal explanation proper to the natural sciences, and Verstehen, the interpretive understanding proper to the human sciences, on the ground that human life is meaningful in a way that nature is not and so calls for a method that grasps meaning from within rather than subsuming behaviour under external law . Weber built an entire interpretive sociology on the demand that social action be understood in terms of the subjective meaning the actor attaches to it . Geertz, in the anthropological register, argued that human conduct must be read through “thick description,” which records not the bare physical motion but the dense web of significance that makes a wink a wink rather than a twitch . And the Frankfurt School added the critical edge: the extension of instrumental, calculative reason into every domain of life is not a neutral gain in knowledge but a form of domination, one that reconstitutes its objects in its own image and forgets that it has done so.

The competence of this pair is over the prior question that evidentialism explicitly cedes: which phenomena admit of measurement without distortion. The interpretivist contribution is the recognition that for certain phenomena the very act of measuring is not a neutral registration but an intervention that alters the phenomenon, sometimes constituting a different phenomenon altogether. The empirical literature confirms what the philosophy predicts: real-time emotional measurement changes the emotion it purports to record , and the data of feeling is in important part constructed by the apparatus that elicits it . The limit of the axis, taken alone, is that it frames the decision, measure or refrain, without supplying a procedure for making it in a given case; it tells us that some phenomena are deformed by measurement without telling us, here and now, whether this one is. Supplying the procedure is the work of the dialectical posture in §5, which draws on all the traditions at once.

Structuralist Psychoanalysis: Desire, the Other, and the Unsymbolizable

The traditions surveyed so far share an assumption that the subject of intimate life is, at bottom, transparent to itself: it has experiences it could in principle articulate, preferences it could in principle report, a wellbeing that measurement might in principle track. Structuralist psychoanalysis, and pre-eminently the work of Jacques Lacan, denies this assumption at its root, and in doing so supplies the framework with a warning the other traditions cannot generate from their own resources. The structuralist background is Saussure’s: a sign takes its value not from a positive content of its own but from its differential position within a system, so that meaning is relational through and through, an effect of structure rather than of substance . Lacan’s decisive move was to claim that the unconscious is structured like such a language, and that the human subject is therefore not the self-present author of its meanings but an effect produced within a symbolic order that precedes and exceeds it .

Three Lacanian theses bear directly on intimate data. First, desire is the desire of the Other: what a person wants is not a fact lodged in them awaiting measurement but is constituted in relation to others and to the symbolic order, mediated, deferred, and often opaque to the desiring subject themselves . A system that purports to read a partner’s desires from behavioural data thus mistakes the very nature of desire, treating as a retrievable datum what is in truth a relational and shifting structure. Second, the subject is split: there is no harmonious inner unity for an interface to mirror back, but a divided subject marked by what escapes its own conscious grasp. Third, and most consequential for design, there is the objet petit a and the broader category of the Real: that which cannot be symbolized, the remainder that no signifier captures and no measurement reaches. In an intimate relation, a great deal of what binds two people belongs to this unsymbolizable remainder; it is precisely not available to articulation, let alone to quantification, and the attempt to capture it does not approximate it but substitutes for it a manageable token that is not the thing.

The critical force of this tradition, developed in the ideology-critical key by , is to expose a fantasy structuring much intimate technology: the fantasy of a relationship made fully transparent, fully known, its remainders abolished, its desires legible on a dashboard. Lacan’s account implies that this fantasy is not merely unattainable but misconceives its object, and that pursuing it damages the relation, for what makes the other a genuine other, an inexhaustible Thou rather than a solved problem, is exactly the remainder that resists capture. The competence of structuralist psychoanalysis in the framework is thus over the constitution of the intimate subject: it establishes that the partner is a split, desiring, partly opaque being, and that this opacity is a condition of there being an other to love at all. Its limit is that it is diagnostic and cautionary rather than prescriptive: it tells us what intimate technology must not pretend to do and why the pretence wounds, but it does not by itself supply either the inviolable rules or the positive texture of care. For these the framework turns to its substantive ethical traditions.

Deontology: The Inviolable Constraints

Deontological ethics supplies what consequence-sensitive theories structurally cannot: constraints that hold regardless of the good that breaching them might produce. In its Kantian form the position rests on the dignity of persons as ends in themselves. The Formula of Humanity in the Groundwork commands that one act so as to treat humanity, whether in one’s own person or in that of another, never merely as a means but always at the same time as an end . To treat a person merely as a means is to use them in a way to which they could not, as a rational agent, consent; the wrong is independent of whether the using turns out well. A second Kantian formula, the Formula of Universal Law, tests a maxim by whether it could be willed as a universal law without contradiction, and a third, the Formula of the Kingdom of Ends, envisions a community in which each rational being is at once author and subject of the laws that bind all. What unites the formulations, for present purposes, is the thought that there are things one may not do to a person, full stop, because doing them is incompatible with respecting them as a self-determining end.

The competence of deontology in the ethics of intimate data is over precisely these inviolable constraints, and the intimate setting gives them sharp and specific content. At least three constraints follow directly. First, the prohibition on covert surveillance: to monitor an intimate without their knowledge is to convert a partner into an object of study, a specimen rather than a co-author of the relation, and so to treat them merely as a means to one’s own knowledge or reassurance; it is impermissible however benign the watcher’s motive and however useful the data. Second, the prohibition on repurposing disclosure: what a person reveals within the trust of an intimate relation may not be redirected to ends they did not endorse and would not endorse, for to do so is to exploit their self-disclosure against the terms on which it was given. Third, the requirement of genuine consent: agreement to be measured must be real, informed, and revocable, not manufactured by design patterns that engineer assent, since a consent the system has engineered is not the person’s own and respects their agency only in form. These constraints are absolute within the framework: they are not magnitudes to be traded against benefits but boundaries within which all trading must occur.

The limit of deontology is the perennial one its critics name: its abstraction. The categorical imperative establishes that I may not treat my partner merely as a means; it is silent on what, concretely, attentiveness to this particular person on this particular evening requires. It draws the outer wall of the permissible and says nothing about how to inhabit the space within. It can tell the designer that a notification must not deceive or manipulate; it cannot tell the designer whether a given notification is warm or cold, timely or intrusive, present or overbearing. For the texture of the permissible, rules give out, and the ethics of care begins.

Feminist Care Ethics: The Texture of Responsiveness

The ethics of care emerged as a sustained challenge to the assumption, shared by deontology and consequentialism alike, that moral reasoning consists in the impartial application of universal principles. began from an empirical observation, that women’s moral reasoning, far from being deficient by the standard of abstract justice, articulated a different and equally valid orientation, one centred on responsibility, relationship, and response to concrete need rather than on the adjudication of competing rights. developed this into a full ethical theory in which the basic moral relation is that between the one-caring and the cared-for, and in which caring is not a feeling but a practice of engrossment in and motivational displacement toward the particular other. extended the ethic from the interpersonal to the political, and analysed care into a sequence of phases, attentiveness, responsibility, competence, and responsiveness, each with its own characteristic failure, thereby giving the ethic the articulated structure its early critics claimed it lacked.

What care ethics contributes to the framework is an account of everything the rule-based traditions cannot reach: the manner, timing, tone, and attunement of the system’s presence in the relation. A notification can satisfy every deontological constraint, deceive no one, manipulate no one, surveil no one, and still be a moral failure of the kind that matters most in intimate life: cold where it should be warm, mechanical where it should be tender, present where it should have kept silence, prompting where it should have waited. Care ethics also reconceives the system’s very purpose. The measure of a caring technology is not the efficiency it delivers nor the pleasant states it produces but whether it sustains and deepens the relation’s responsiveness to the particular, irreplaceable, vulnerable other, the person who cannot be substituted by any other person with the same metrics. Tronto’s phase of attentiveness has a direct design correlate: a caring system must first of all be the kind of thing that notices, and notices the right things, which are frequently not the things easiest to measure.

A heuristic from popular relationship psychology may be mentioned here for its practical suggestiveness, with an explicit caveat about its standing. Chapman’s notion of “five love languages,” which holds that people characteristically express and receive care through words of affirmation, quality time, gifts, acts of service, or physical touch, offers designers a usable vocabulary for the channels through which a caring system might attune itself to a particular partner . The caveat is that this is a work of popular rather than academic psychology, with weak empirical support for its strong claims, in particular its typology of discrete “languages” and its developmental and clinical assertions; it is cited here only as a source of design heuristics for the texture of care, not as a validated theory, and nothing in the framework depends on it. Within those limits it is suggestive: it reminds the designer that attunement is plural, that the same gesture may register as care to one partner and as noise to another, and that a caring system should therefore make room for several registers of expression rather than assuming one.

Care ethics carries a famous internal danger that the framework must name, because naming it is part of the argument for integration. An ethic of pure responsiveness, unchecked by justice and rights, can slide into partiality, the erosion of boundaries, and the dissolution of the carer into the needs of the cared-for; care without justice can smother. In the intimate-data setting this danger has a precise shape: it is the over-attentiveness that, meaning only to be present and responsive, curdles into the anxious monitoring of a beloved, the very wrong the absent-other analysis of §7 identifies. An ethic of care that knew no deontological boundary might positively endorse such monitoring as a form of devoted attention. This is why care needs deontology as surely as deontology needs care: the one supplies the inviolable wall, the other the warmth of what is built within it, and each without the other deforms. The mutual need is not a weakness of either theory but the structural fact that grounds their integration.

Confucian Role-Ethics and Daoist Non-Interference: The Value of Restraint

The Western traditions assembled so far, for all their differences, share a tacit orientation toward action: ethics, on each of them, is largely a matter of what one must do, may do, or must not do. The East Asian sources supply what this orientation structurally obscures, a positive and developed account of the value of not acting, of restraint as itself an achievement rather than a mere absence. Confucian role-ethics, as reconstructed by and , takes the person to be constituted in and through their relations and the roles those relations carry, so that one becomes a self by the cultivated, lifelong refinement of how one stands in relation to others. Central to this cultivation is li, the vast Confucian category covering ritual, propriety, and the fine grain of appropriate conduct, which crucially includes the discipline of knowing what to leave undone and unsaid. Li is not empty formality; it is respect made concrete in restraint, the cultivated capacity to give another person room, to not press, to not demand the articulation of what is better left in dignified silence.

Daoism presses the valorization of restraint to its philosophical limit. The Daodejing’s counsel of wu wei, ordinarily rendered “non-action” but better understood as action that does not force, that works with rather than against the grain of things, rests on the conviction that the most important goods cannot be produced by direct interventionist effort and are in fact destroyed by it. The dictum that “reversal is the movement of the Dao” (fan zhe dao zhi dong) captures the recurring observation that forcing a thing toward its goal often produces the opposite, that grasping drives away, that the anxious pursuit of intimacy repels the intimacy pursued. A relationship, on this view, belongs to the class of goods that grow of themselves when given room and wither under management; it is cultivated as a garden is cultivated, by patient attention to conditions, not as a machine is operated, by direct manipulation of outputs.

The competence of these sources is over a question the action-centred traditions cannot even formulate in their own terms: when the right design decision is to build nothing, to leave a capacity unimplemented, a datum unrecorded, a prompt unsent, precisely as an expression of respect and of trust in the relation’s own movement. This is the deepest correction the framework receives, because the disposition it counteracts, the engineer’s and the optimizer’s bias toward building the feature because the feature can be built, is so nearly invisible from within the action-oriented traditions. The limit of the East Asian sources, for a paper addressed to the modern intimate relation between equals who retain independent standing, is that they do not by themselves furnish the vocabulary of individual rights and revocable consent that such a relation also requires; classical Confucianism in particular presupposes a role structure more hierarchical than the egalitarian partnership at issue here. They are integrated, therefore, not as a replacement for the justice traditions but as their indispensable supplement: the source within the framework of a standing presumption in favour of restraint.

The Capability Approach: The Criterion of Flourishing

If the foregoing traditions tell us what may be claimed, what may be measured, what may not be done, and what is better left undone, the capability approach supplies the standard against which the entire arrangement is finally to be judged. Developed by as a critique of welfarist and resourcist accounts of wellbeing, and given a determinate content by , the approach evaluates a state of affairs not by the resources it distributes nor by the subjective satisfactions it generates but by the capabilities it secures: the real, effective freedoms persons have to be and to do what they have reason to value. The distinction from resourcism is crucial. Two people with identical resources may have radically unequal real freedoms, since the conversion of resources into actual functionings depends on circumstance; and the distinction from welfarism matters equally, since a person may report contentment precisely because their horizons have been adaptively narrowed to fit a constrained life. The right question, on this view, is not “how much does the arrangement provide” nor “how satisfied are its subjects” but “what range of valuable being and doing does it actually open or foreclose.”

Applied as the framework’s overarching criterion, the capability approach reframes the entire evaluation of an intimate technology. The question is never whether the system makes the relationship more efficient, nor whether it produces more agreeable feelings on a given evening, but whether it expands or contracts the partners’ real freedom to become who they have reason to want to be, both together and as separate persons. This criterion has teeth precisely against the most seductive failure mode of optimization. A system might increase coordination, reduce friction, and elevate reported satisfaction while quietly narrowing the space of who the partners are permitted to become, training them toward the legible, the trackable, the optimizable, and away from the unmeasured experiments and reversals through which people actually grow. By the capability standard such a system is a failure however smooth its operation and however high its satisfaction scores, because it has traded real freedom for managed contentment, which is the adaptive-preference trap installed in the home.

The capability approach supplies the criterion but, by its own structure, not the constraints. Being broadly consequentialist in form, an evaluation of net capability could in principle license the violation of an individual’s rights for an aggregate expansion of valuable freedom, and could in principle countenance an intervention that the dignity of the person, or the texture of care, forbids. It therefore depends on deontology to fix the boundaries within which capability-expansion may be pursued, on care ethics to specify the manner of the pursuit, and on the East Asian sources to remind it that some capabilities flourish only when not directly engineered. The criterion tells us what to aim at; it does not tell us what we may do in the aiming. This dependence is, once again, not a defect but a pointer toward integration.

Developmental Psychology: Joint Attention and the Theory of Mind

The relational ontology of §2.2 makes a strong claim: that meaning and selfhood are co-constituted between subjects rather than generated within them. That claim is philosophical, but it is not without empirical corroboration, and the developmental psychology of shared intentionality supplies the corroboration. Human cognition, on the influential account of Tomasello, is distinctively cultural cognition, made possible by a capacity for shared intentionality that emerges in infancy and that other primates largely lack . The foundational mechanism is joint attention: two subjects attending together to the same object while each is aware that the other attends, a triadic structure (self, other, world) from which shared meaning is built. The complementary capacity is the theory of mind, the ability to attribute mental states to others and thereby to model the other as a centre of experience rather than as a moving body . Together these capacities are the developmental substrate of relational being: they are how, in fact, two human subjects come to share a world.

The competence of this tradition within the framework is to ground the otherwise abstract notion of co-presence (§6.8) in a mechanism that is real, studied, and specific. When the framework asks a system to host the partners’ shared attention rather than to absorb each partner’s attention separately, it is asking for the technological cultivation of joint attention in the precise developmental sense, and the design difference between the two is sharp: a feed that captures each partner’s gaze individually is the negation of joint attention, while a shared object the partners attend to together, aware that they do so, is its realization. The theory of mind enters as a constraint of a different kind. A system that purported to model a partner’s mind and to report its model to the other would not be supporting theory of mind but supplanting it, substituting a machine’s representation for the irreducible work of one person imagining another; the framework’s account of the AI as pseudo-presence (§2.2) here acquires empirical teeth, since the relational work of mind-reading is exactly the work the pseudo-subject cannot do and must not pretend to do. The limit of developmental psychology is that it describes the mechanisms of shared meaning without prescribing their right use; it tells us what joint attention is and that it matters, but it cedes to the ethical traditions the question of what may rightly be done with it.

Eudaimonia: Flourishing as a Generative Dynamic

The capability approach (§3.9) supplies a criterion of flourishing stated in terms of real freedoms; the eudaimonic tradition supplies the complementary account of flourishing as a dynamic of self-realization, and it is this tradition that grounds the generative-field commitment announced at the opening of this section. The Aristotelian root holds that the human good is not a state to be reached but an activity, the active exercise and realization of one’s characteristic capacities over a whole life. Contemporary psychology has given the idea operational content: Ryff’s model of psychological well-being identifies dimensions such as autonomy, personal growth, purpose, and positive relations with others as constituents of eudaimonic, as distinct from merely hedonic, well-being . What distinguishes the eudaimonic from the hedonic is decisive for design: a system optimized for the hedonic produces agreeable states, while a system answering to the eudaimonic provides the conditions under which growth, purpose, and relation can be actively realized, conditions that frequently involve difficulty, novelty, and the unmeasured, rather than comfort.

This yields the framework’s central positive commitment, distinct from the negative work of its constraints. The design philosophy common to the system’s modules is the provision of a generative field: a bounded space, marked off by the framework’s constraints, within which the partners’ shared self-realization can unfold without being driven toward any measured outcome. The eudaimonic tradition supplies the reason this is the right aim, namely that flourishing is a self-realizing activity that cannot be produced from outside but only afforded, and the design correlate, namely that the system’s task is to furnish the field and withdraw, in the manner of the Daoist wu wei (§3.8) and of the witness rather than the advisor (§6). Its relation to the capability approach is one of complement: the capability approach states the criterion against which a generative field is judged, whether it expands real freedom to be and do, while eudaimonia states the dynamic the field is meant to host, the active realization through which such freedom is exercised. The limit of the eudaimonic tradition, taken alone, is that it specifies the positive aim without supplying the inviolable constraints that keep the pursuit of flourishing from overriding the person; for those it depends, as the capability approach does, on the deontological and care-ethical traditions.

Generative Justice: The Resting Place of Value

A final tradition addresses a question none of the others quite reaches: not what may be known, measured, or done, but where the value generated within the relation, and within the system that mediates it, is permitted to come to rest. Ron Eglash’s account of generative justice begins from a critique of how value circulates under extractive arrangements. Distributive justice, on Eglash’s analysis, accepts that value will be alienated from those who generate it and asks only how the alienated surplus should afterwards be redistributed; generative justice asks the prior and more radical question of how value might be kept circulating among, and returning to, those who generate it, never alienated in the first place . The model is drawn partly from ecological cycles, in which the value of an ecosystem is retained within it rather than extracted, and partly from a reading of Marx’s account of how capital alienates the value labour produces , generalized from labour-value to ecological and expressive value as well.

The relevance to intimate data is both pointed and, for the dominant industry model, indicting. The prevailing architecture of consumer technology is extractive in exactly Eglash’s sense: the value that users generate, their attention, their disclosures, the rich record of their shared life, is alienated from the relation that produced it and accumulated elsewhere, on servers and in business models the partners neither see nor control, to be monetized by parties external to the relation. Generative justice supplies the principle that condemns this arrangement and the criterion for an alternative: the value generated within an intimate relation, including the data that is the sediment of that relation’s history, must remain within the relation, circulating between the partners and returning to them, never extracted to enrich a third party or even to enrich one partner at the other’s expense. This yields concrete and unusual design imperatives: local data ownership rather than cloud extraction, architectures in which the record of a shared life is held by those whose life it is, and a positive prohibition on the conversion of intimate value into someone else’s surplus.

The competence of generative justice is thus over the political economy of the relation’s value, the dimension along which questions of ownership, extraction, and alienation are posed, a dimension orthogonal to and unaddressed by the epistemic and ethical traditions above. Its limit is that it speaks to the circulation and retention of value but not to the prior questions of what may be measured or claimed, nor to the inviolable treatment of persons; a system could be perfectly generative in Eglash’s sense, alienating no value to any third party, and still commit covert surveillance of one partner by the other. Generative justice closes the framework by addressing where value rests, but it presupposes the rest of the framework to govern how that value was permitted to arise.

Data Ethics: Contextual Integrity and the Sediment of a Shared Life

The traditions assembled so far touch the ethics of data at many points, but the contemporary field of data ethics has developed concepts specific enough to warrant their own treatment. Chief among them is Helen Nissenbaum’s principle of contextual integrity, which holds that the flow of personal information is governed by norms specific to the context in which it arose, and that a privacy violation consists in the inappropriate movement of information across contextual boundaries rather than in disclosure as such . The principle dissolves a confusion that has long bedevilled privacy talk. The question is never simply whether information is shared, but whether it flows according to the norms of the context that generated it; medical information appropriately shared with a physician is violated when sold to an insurer, not because it has moved but because it has moved against the informational norms of the medical relation.

Applied to intimate partnership, contextual integrity yields a determinate and demanding standard. The intimate relation is a context with its own informational norms, and those norms are unusually strict: what is disclosed within it is disclosed on the tacit understanding that it remains within it, neither flowing outward to third parties nor, often, flowing onward even between the partners in forms the discloser did not intend, such as aggregation, scoring, or permanent retention. A diary entry shared in the expectation that it may be read is not thereby licensed for sentiment analysis; a confidence offered in conversation is not thereby licensed for indefinite storage and later retrieval. The competence of data ethics within the framework is over these norms of appropriate flow, together with the cognate principles the field has articulated: data minimization, which counsels collecting only what the immediate purpose requires; purpose limitation, which forbids the silent repurposing of data to ends beyond those for which it was given; and a right to be forgotten, which treats the erasability of the record as itself a condition of a relation that is permitted to change. These principles operationalize, in the specific medium of data, the deontological and care-ethical commitments stated above, and they connect to the relational ontology directly: the data a relationship generates is the sediment of that relation’s history, neither a neutral resource nor an asset but a trace of lived presence, and to be governed accordingly.

The limit of data ethics, taken on its own, is that it is a regional ethics: it governs information flows with precision but does not supply the deeper account of why the intimate context has the norms it does, nor the criterion of flourishing against which the whole is judged. It receives those from the relational ontology and the capability approach respectively. Within its region, however, it is indispensable, for it translates the framework’s abstract commitments into the concrete vocabulary, flow, minimization, purpose, erasure, in which the design of an actual data system is necessarily conducted.

Recognition and the Face of the Other: Levinas and Honneth

A final pair of sources deepens the account of what the other is owed. Levinas locates the origin of ethics in the face-to-face encounter, in which the presence of the Other calls the self into a responsibility that precedes and exceeds any contract or calculation; the Other is not an object to be comprehended but an infinity that overflows every representation I could form of them . This is the ethical correlate of the ontological point already made: just as no symbol coincides with relational being, no representation of the partner, however data-rich, captures the person, and the attempt to substitute the representation for the person is not merely an epistemic error but an ethical failure, a refusal of the responsibility the face commands. Honneth’s theory of recognition supplies the developmental and political complement: persons come to themselves through being recognized by others across the spheres of love, rights, and esteem, and misrecognition is a genuine injury, not a mere bad feeling . An intimate technology participates, for good or ill, in the work of recognition; a system that reflects back to a partner only what it can measure recognizes a diminished version of them, and over time may school them to recognize only that diminished version in themselves.

The competence of this pair is over the quality of recognition the system extends and enables: whether it honours the other, in Levinas’s sense, as an infinity exceeding its representations, and whether it supports rather than distorts, in Honneth’s sense, the mutual recognition through which the partners become themselves. Its limit is shared with care ethics, with which it is closely allied: it specifies the orientation owed to the other without, on its own, fixing the inviolable rules or the political economy of value. It connects the framework’s ethical layer back to its ontological ground, and prepares the account of co-presence and co-creation developed in the case study, where recognition becomes something the partners do together rather than something the system does to them.

The Argument for Integration

Assembling these traditions raises the question whether the result is a framework or a buffet, and how conflicts among them, such as the standing tension between care and justice, are to be resolved. The integration proposed here is not an aggregation, in which each tradition contributes a weighted vote and a master function returns a verdict. Weighted aggregation across incommensurable normative kinds is precisely the positivist error transposed into ethics, and it fails for the same reason: it presumes a common scale where there is none. Nor is the integration a synthesis in the strong Hegelian sense, a higher unity in which the tensions are dissolved; for the reasons given under dialectics (§3.1), the framework follows Adorno in holding the contradictions open rather than abolishing them. The integration is instead a division of jurisdictional labour, modelled loosely on the way a well-ordered constitution assigns different questions to different competent organs rather than collapsing them into a single sovereign, all of them resting on the relational ontology of §2.2 as their common ground. Each tradition is assigned the class of questions over which the preceding sections argued it is competent, and is denied authority over the others. Table 1 sets out the resulting division of labour in full, stating for each tradition the jurisdiction it holds, the central question it is competent to answer, and the limit at which it must cede authority to another.

Table 1. The division of jurisdictional labour among the constituent traditions. Each holds authority over a distinct class of questions and cedes the questions beyond its competence to the tradition equipped to answer them.

Tradition	Jurisdiction and central question	Limit, and what it cedes
Relational ontology (Huang 2025; Gergen 2009; Whitehead 1978; Buber 1937)	The ground. That the relation, not the individual, is the basic unit of concern, and that no representation ever coincides with the relational being it refracts.	A ground rather than a rule; it establishes what the other traditions are about but issues no determinate permission or prohibition on its own.
Dialectics (Hegel 1807; Adorno 1966)	The method of the whole. The requirement that the contradiction between measuring and refraining be held open and re-adjudicated at each site rather than resolved once by formula.	A form rather than a content; it specifies how the questions are held together but, alone, decides no particular case.
Evidentialism (Clifford 1877; Feldman and Conee 1985)	Epistemic modality. What the system may assert, and with what confidence: the demand that estimates be expressed as estimates and false precision refused.	Silent on what ought to be measured at all, and on who counts as a credible source; cedes these to ethics and to epistemic injustice.
Epistemic injustice (Fricker 2007)	Credibility and interpretive fairness. Whether each partner is treated justly as a knower of the world and of themselves, and whether the system’s vocabulary impoverishes their self-understanding.	Diagnoses a wrong without fixing the inviolable boundaries whose breach is gravest; cedes those to deontology.
Positivism and anti-positivism (Dilthey 1989; Weber 1978; Geertz 1973)	The prior measurement question. Whether a given phenomenon admits measurement without distortion, or is instead partly constituted by not being measured.	Frames the choice between measuring and refraining but supplies no decision procedure; cedes the decision to the dialectical posture drawing on all traditions.
Structuralist psychoanalysis (Saussure 1916; Lacan 2006; Lacan 1977; Žižek 1989)	The constitution of the subject. It forbids the fantasy of total transparency and protects the unsymbolizable remainder that makes the other a genuine other rather than a solved problem.	Diagnostic and cautionary rather than prescriptive; cedes the positive rules and the texture of care to deontology and care ethics.
Deontology (Kant 1785)	The inviolable constraints. What may not be done to the other regardless of benefit: no covert surveillance, no repurposing of disclosure, no engineered consent.	Abstract; it draws the outer wall of the permissible but is silent on how to inhabit the space within. Cedes the texture to care ethics.
Care ethics (Gilligan 1982; Noddings 1984; Held 2006; Tronto 1993)	The texture. The manner, timing, tone, and attunement of the system’s presence: warm where it should be warm, silent where it should be silent.	Without justice it can license partiality and boundary-violation; cedes the inviolable limits to deontology, on which it depends as deontology depends on it.
Recognition and the ethics of the face (Levinas 1969; Honneth 1995)	What the other is owed. That the partner be honoured as an infinity exceeding every representation (Levinas), and that the mutual recognition through which the partners become themselves be supported rather than distorted (Honneth).	Specifies the orientation owed to the other without fixing the inviolable rules or the political economy of value; allied to care ethics and dependent on the same supplements.
Confucian and Daoist restraint (Ames 2011; Tu 1985)	The question of non-action. When the right design decision is to build nothing: to leave a capacity unimplemented, a datum unrecorded, a prompt unsent, as an enacted respect.	Does not, alone, furnish the vocabulary of individual rights and revocable consent the modern egalitarian partnership also requires; supplements rather than replaces the justice traditions.
The capability approach (Sen 1999; Nussbaum 2000)	The criterion. Whether the whole expands or contracts the partners’ real freedom to become who they have reason to want to be, together and separately.	Broadly consequentialist; would in principle license a rights-violation that expanded capabilities on net. Cedes the constraints to deontology and the texture to care.
Developmental psychology (Tomasello 1999; Premack and Woodruff 1978)	The mechanism of shared meaning. Joint attention and theory of mind as the developmental substrate by which two subjects come to share a world, grounding co-presence empirically.	Describes the mechanisms of shared meaning without prescribing their right use; cedes the question of rightful use to the ethical traditions.
Eudaimonia (Ryff and Keyes 1995)	The positive aim. Flourishing as a dynamic of self-realization, furnishing the reason the system should provide a generative field rather than optimize states.	Specifies the positive aim without the inviolable constraints; depends on deontology and care to keep the pursuit of flourishing from overriding the person.
Generative justice (Eglash 2016; Marx 1867)	The political economy of value. Whether the value the relation generates, including the data that is the sediment of its history, remains within the relation rather than being extracted to enrich a third party.	Governs where value rests but not how it arose; presupposes the rest of the framework to govern what was permitted to be measured or done in the first place.
Data ethics (Nissenbaum 2010)	The flow of information. Whether data moves according to the norms of the intimate context that generated it, and whether it is minimized, purpose-limited, and erasable.	A regional ethics; governs information flows with precision but receives the account of why the intimate context has its norms, and the criterion of flourishing, from the ontology and the capability approach.

Conflicts are then resolved not by trading off magnitudes but lexically, by jurisdiction, in roughly the order the framework’s layering implies. The deontological constraints and the structuralist-psychoanalytic prohibition on engineered transparency sit at the base: they are not magnitudes to be outweighed by any care-based, capability-based, or coordination benefit but boundaries that bound the space within which all the other considerations operate. Within that space the measurement question is posed, framed by the positivism–anti-positivism axis and decided by the dialectical procedure of §5; where measurement is admissible, evidentialism constrains how its results may be expressed and the epistemic-injustice constraint ensures that neither partner is thereby demoted as a knower. Across the whole, the Confucian-Daoist consideration stands as a standing presumption in favour of restraint, to be overcome only by a positive showing that action serves flourishing better than non-action would; the capability approach supplies the criterion by which that showing is judged; and generative justice governs where the resulting value is permitted to rest. Dialectics is not one jurisdiction among the others but the form of their coordination: it is the reason the ordering issues in a procedure of perpetual re-adjudication rather than a fixed table of verdicts.

This is what distinguishes integration from eclecticism: the traditions are ordered, and the ordering is itself defended, tradition by tradition, by the demonstration that each depends on the others to supply exactly what it lacks. Evidentialism needs the measurement question it cannot pose; the measurement question needs the psychoanalytic account of why some phenomena resist capture; deontology needs the texture care supplies; care needs the boundary deontology supplies; the capability criterion needs the constraints it would otherwise override; the East Asian sources need the rights-vocabulary they do not furnish; generative justice needs the rest of the framework to govern how value arose before it can govern where value rests. The framework is integrative because its parts are not merely collected but made mutually load-bearing. We call the resulting structure an integrative justice because it treats the just governance of intimate data as requiring the coordinated exercise of several distinct normative competences, no one of which is sovereign, and none of which can be removed without the others losing their footing.

An Ethics of Dialectical-Positivist Practice

The practical upshot of the integrative-justice framework is a posture toward measurement that is neither positivist nor anti-positivist but dialectical. The positivist thesis, which holds that domestic phenomena are quantifiable and thereby improvable, and the anti-positivist antithesis, which holds that the constitutive phenomena of intimate life are deformed by quantification, are both false when taken as universal claims about a whole domain. The synthesis is not a compromise midpoint but a method: at each site of possible measurement, the framework’s jurisdictions are convened to decide whether this phenomenon, here, is one whose measurement would describe it or one whose measurement would replace it.

The decision turns on a test that the framework makes precise. A phenomenon is a candidate for legitimate measurement when measuring it leaves it the same phenomenon, as a sum of expenditures is the same sum whether or not it is tallied, and a candidate for principled refusal when the phenomenon is partly constituted by not being measured, so that measurement converts it into something else. The paradigm of the latter is the affective content of a diary entry: an entry written in the knowledge that it will be scored for sentiment is not the same act as the same words written in the absence of that knowledge . Between the clear cases lies a contested middle, comprising such matters as health, personal growth, and attention, where the phenomenon is partly robust to measurement and partly altered by it, and where the framework counsels measuring the robust part while labelling the estimate honestly, as evidentialism requires, refraining from the comparison and ranking that would alter behaviour, as the measurement question requires, and never initiating where the other has not asked, as care and restraint require.

The posture is dialectical in a second sense: it is never finally settled. As a relationship changes, the same phenomenon may migrate across the line: what could once be safely counted may become, under new circumstances, something whose counting wounds. The framework therefore does not deliver a fixed classification of modules into permitted and forbidden; it delivers a procedure for re-adjudication, and treats the act of re-adjudication as itself a first-order exercise of the ethics it describes. The next section shows the procedure at work.

Case Study: A Family-and-Relationship Application

The framework is abstract; this section makes it concrete by examining a working application built by the author, whose modules instantiate the framework’s jurisdictions in software. The application comprises roughly two dozen modules. Rather than catalogue them by function, a task belonging to a design report rather than a philosophical paper, this section selects modules that exhibit particular jurisdictions of the integrative-justice framework with unusual clarity. A brief note on the system’s construction is confined to Appendix 11.

The Persistent Design Rationale

Before the individual modules, one cross-cutting artefact deserves notice because it is itself an instance of the framework’s central move. A persistent file, consulted by the system’s conversational layer and committed to version control, records each design decision as a one-line rule together with the reason that produced it. Three representative entries may be paraphrased as follows. The first holds that the tone of any notification must be warm rather than bureaucratic, closer to a love note than to a system message. The second states the design ethos as one of intentional slowness, on the rationale that mediated, unhurried interactions are more intimate, and that the system should design for small ceremony rather than instant notification. The third forbids the forcing of ceremony: tender features remain opt-in, since auto-prompting them misreads the partners’ emotional state and creates pressure where there should be none.

The artefact is positivist in form, being a structured, indexed, machine-readable list of rules, and anti-positivist in content, since each rule encodes a reason that is irreducibly contextual and partly affective. It is, in miniature, the dialectical-positivist posture made into a working object: a record that measures its own decisions without pretending the reasons behind them are themselves quantities.

The Joint Ledger: Legitimate Measurement and Its Practical Limits

A joint ledger of expenses across two persons and several currencies is the clearest case of legitimate measurement: a unit spent is a unit spent, and tallying it changes nothing about it. The module accordingly occupies the positivist end without apology, and Confucian sources even read exact accounting in this register as a courtesy that forestalls the small unspoken asymmetries that would otherwise accrete. On this foundation the module offers the apparatus one expects of household finance: categorized records of income and expenditure, the economic and statistical summaries that turn a stream of transactions into legible structure, visualizations of where the shared money goes, and a measure of AI-assisted analysis and advice. The framework’s first work here is at the margin the module declines to occupy: it does not score the relative spending of the two parties, does not nudge toward an “optimal” division, and produces no leaderboard. The deontological constraint against converting a partner into an object of comparative evaluation, not any limit of the data, fixes where the measurement stops. The ethics of the module lies first of all in these omissions.

But the joint ledger, precisely because it is the module where measurement is most clearly legitimate, also exposes most clearly the practical limits of measurement, and honesty requires that these be stated rather than glossed. Three are salient. The first is the problem of overhead and its justification. Complete bookkeeping is laborious, and in practice no couple records every transaction; the labour of capture is a real cost, and it is set against a benefit, the clarity that quantitative analysis affords, that is genuine but bounded. The framework offers no formula for the trade-off, and indeed its own dialectical posture (§5) implies there is none: whether the marginal entry is worth the friction of recording it is a judgment the partners must make, and a system that demanded total capture in the name of analytic completeness would have let the positivist impulse override the lived cost it imposes. The honest design admits partial data and analyses what it has without pretending to completeness.

The second limit follows directly: completeness cannot be guaranteed, and the analytics must therefore be read as describing a sample rather than the whole. This is the evidentialist constraint (§3.2) applied to the ledger: a visualization built on incomplete records must not present itself as the full financial truth of the relation, and the module’s summaries are accordingly framed as partial pictures, not audited accounts.

The third limit is the most interesting, because it shows even this most measurable of domains touching the constitutive unsaid (§2). Some expenditures ought, within an intimate relation, to remain hidden: the carefully prepared surprise, the gift whose value depends on its not being foreseen. A ledger designed for total mutual visibility would, in the name of transparency, destroy the small concealments that intimate life legitimately requires, and so the module must accommodate private entries that one partner can record outside the other’s view. The point connects to the architecture of secrecy discussed among the framework’s open tensions (§8): even in finance, the relation’s good is not served by total disclosure, and a design that could not keep a secret could not host a surprise.

A final, more technical caution concerns the AI capabilities the module offers. The system can attempt to extract expenditures from photographs of bills and receipts, reducing the overhead named above, and it can offer analytic advice on the recorded data. The honest position is that the practical reliability of such recognition remains to be established: receipt extraction is error-prone, and advice computed on partial and imperfectly captured data inherits all the limits already named. These features are therefore offered as convenience subject to the user’s correction, governed by the same evidentialist demand that the system not assert more confidence in its figures than their provenance warrants, rather than as a trustworthy automation of the household’s financial truth.

Household Management: Coordination Without Scorekeeping

The management of a shared household is, like the ledger, a domain where measurement is largely legitimate, and the system addresses it through an event-driven design worth describing because its structure carries the module’s ethical intent. The objects the module tracks are events of several kinds. The first are multi-scale clock events, recurring tasks on different periods, the floor cleaned once a week, the table wiped once a day, each surfacing when its period elapses. The second are passive events, triggered by depletion, the cleaning fluid run out, the bin-bags finished. The third are proactive warning events, anticipations of depletion, the cleaning fluid running low and likely to be exhausted soon. The fourth are manually entered events, recorded by either partner, a thing to be bought, a matter to be handled. Every event carries a spatial attribute locating it in the home, the kitchen, the bathroom, the living room, the bedroom, or elsewhere, so that the household’s small business is organized by place as well as by time.

Because raising an event must be effortless or it will not be done, the interface presents common events in a compact table whose cells are triggered by a single tap. A representative layout records, for each day and each event type, whether the event has occurred and whether it has been resolved; a green dot marks a matter resolved, and a check mark a matter that has arisen and awaits attention. Table 2 illustrates the form. When an event arises, the system reports it to the other partner, or to both, by email or text message, so that the knowledge of what the home needs is shared rather than resting on one person’s vigilance.

Date	Refuse accumulating	Tissues out	Alcohol wipes out
Today
06/01		•
05/31
05/30			✓
05/29
05/28
05/27

Table 2. An illustrative “daily record” for the household module. Each column is a common event type, each row a day. A resolved matter is marked with a dot (•), an event that has arisen and awaits attention with a check (✓); empty cells indicate nothing recorded. The layout is schematic, not the deployed interface.

The positive case for the module rests on a fact about persons rather than about households: human memory and attention are finite, and the labour of holding in mind what the home needs, when the fluid will run out, whose turn the weekly task is, what must be bought, is a real and recurring tax on that finite resource. By externalizing this tracking into a shared, low-friction system, the module relieves both partners of a class of mental burden that, left in the head, tends to fall unequally and to surface as friction at bad moments. What is gained is not merely tidiness but attention returned to the relation: time and mental space that would have gone to remembering and reminding can instead be given to one another. In the framework’s terms, this is the capability approach (§3.9) and the eudaimonic commitment (§3.11) at work in the most ordinary register, the system furnishing a field in which the partners are freed, in a small but real way, for the more meaningful uses of their shared time.

Two considerations justify approaching this domain through efficiency, and through an event-driven design in particular, and they are worth stating because they show the choice to be reasoned rather than merely convenient. The first is existential, and it explains why efficiency is the right priority here when it is the wrong priority for the diary or the co-creative modules. Repetitive household labour is, in its sheer recurrence, among the least generative of activities: in the bare repetition of wiping a surface or noting that a supply has run out, there is little in which a subject can witness its own existence, none of the self-realization the eudaimonic register names. Where the diary and the shared exploration are sites at which presence is enacted and so must not be optimized away, the recurring management of supplies is a site at which presence is largely absent, and the humane response to a domain of bare, non-generative repetition is precisely to dispatch it efficiently, so that the finite attention it would otherwise consume is freed for the activities in which existence is actually realized. The module is deliberately confined to this register: it concerns only the triggering and recording of events, operations that are, by their nature, scarcely relational and largely dull, and it is exactly because they are so that handing them to a mechanism is no loss to the relation but a service to it.

The second consideration explains the particular shape of the mechanism. The event-driven paradigm is borrowed, as an analogy rather than a strict isomorphism, from engineering and from biology, where event-driven architectures are valued for their efficient use of computational resources: a spiking neural network, for instance, expends energy only when a neuron fires, rather than continuously, and event-driven systems generally act only when something happens rather than by constant polling. The same logic recommends itself for the management of a household. A system that continuously monitored and prompted would impose a constant low-grade demand of just the kind the module exists to remove; an event-driven design, which stirs only when a clock elapses, a supply depletes, or a partner records something, mirrors the efficiency of its biological and engineered analogues and keeps the tool, fittingly, silent until there is genuinely something to attend to. The analogy is heuristic and is not pressed further than this; it is offered to show that the design’s form, and not only its existence, answers to a reason.

The module carries a characteristic danger that the framework requires be named, and the event-driven design is shaped to forestall it. The danger is that quantifying housework slides into scorekeeping: once tasks are logged, it is a short step to tallying who did more, and the tallying of contributions is corrosive in an intimate relation, converting shared life into a ledger of debts and converting the partner into an object of comparative evaluation, the very wrong the deontological constraint forbids (§3.6) and the ledger module already refused. The design’s response is structural: the module records that an event occurred and that it was resolved, but it does not, by default, record who resolved it, and it produces no per-person totals, no fairness score, and no leaderboard. It tracks the state of the home, not the comparative diligence of its inhabitants. A subtler danger lies one level deeper: the comprehensive eventification of domestic life, every task a tracked recurrence, every cupboard a depletion forecast, risks turning the home from a place of dwelling into a managed system, an optimization target rather than a shared life. The framework’s response is the same restraint counselled elsewhere (§3.8): the module is offered as relief from cognitive load, not as an instrument for the total administration of the household, and the partners remain free to leave much of their life together uneventified, which is frequently the right thing to do.

The Love Journal: Principled Refusal, Gentle Witness, and the Handwritten Letter

Per-day journal entries by either party are readable by both, but the default is silence: no notification announces a new entry, and discovery is ordinarily by browsing, not by push. There is no sentiment ranking, no automatic summary, no AI commentary on the content of an entry unless explicitly invoked. This is the paradigm of principled refusal. A diary entry delivered by push notification has been converted from a thing written into a thing transmitted, and an entry written in the knowledge that it will be scored is a different act from one written without that knowledge. The framework’s verdict on the entry’s affective content is determinate: it fails the measurement test, since measuring it replaces it; the deontological constraint forbids covert affective surveillance; and care ethics governs what remains, the warmth of the interface, the absence of pressure, the preservation of the entry’s status as an uncompelled disclosure.

Within these limits the module offers two features that are worth examining, because each shows the framework permitting a narrow, carefully bounded AI involvement without crossing into the wrongs just named. The first is an optional reminder. A partner may choose to let the AI, drawing gently on the recent state of the two, notify them when the other has written something, in a warm rather than a clerical register. Such a message might read:

“xx 写了新的日记！似乎十分喜欢你买的花。
信中包含了满满的爱意，快去看看吧。
我们的小家似乎又幸福了一些。”

The feature is reconciled with the default of silence by three constraints. It is opt-in, never the system’s default; it is a factual report that something has been written and that its tone is warm, not a score, a summary, or an analysis of the entry’s content; and, most importantly, it offers no substantive relational advice. It will say that a letter full of love is waiting; it will not say what the recipient should write back, nor how they should feel, nor what they should do. The line is the same one drawn for the shared calendar (§6.7): the AI may warmly report a fact, but it must not perform the partner’s own act of knowing, attending, or deciding, on pain of the proxy epistemic injustice analysed there. Here the witness posture (§6) is enacted in the gentlest possible form, a presence that says only “something loving is here for you,” and then withdraws.

The second feature is a generator of decorative letter-paper, and its design is, once again, existential. In an age of generative models, symbols have become cheap to produce: text, images, ornament can be conjured endlessly and at no cost, and precisely this cheapness drains them of the weight that a scarce and effortful symbol once carried. The module turns this cheapness to a deliberate purpose. The AI generates letter-paper to the partner’s present feeling, whatever is in their heart in the moment, whatever they wish to say, and the partner may regenerate it again and again, freely, until one design answers to that feeling; the chosen sheet is then printed, and the letter is written on it by hand. The point is not the paper, which is exactly the cheap, infinitely regenerable symbol the age makes possible, but what the paper is for. The act of regenerating until a sheet feels right is an exercise of sensibility, a waiting for the moment at which something answers to what one feels; and the handwriting that follows is the irreducibly costly mark the generative symbol can never become, bearing in its variation and its difference the trace of the particular hand that made it. In writing by hand, slowly and unrepeatably, a person witnesses their own existence and attends to the existence of the one they write to, in a way no generated text could enact. The AI, here, takes on the cheap and regenerable symbolic layer precisely so as to set off and make room for the layer that cannot be cheaply generated at all: the handwritten letter as an encounter with the Real. The design gives back to the person the occasion of writing by hand, and with it a possibility, increasingly rare, of meeting the other across a mark that only they could have made.

Health and Growth: The Contested Middle

Between the clear cases lies the terrain the dialectic was built for. The health module estimates nutrient intake from free-text meal descriptions; the estimate is real but approximate, so the interface labels it as an estimate and refuses the false precision of a single confident figure, evidentialism governing modality. Neither module produces thresholds, alerts, rankings, or recommendations, because being-measured alters the practices of eating and of self-cultivation more strongly than it alters a sum of money: to count one’s growth sessions can invert the practice from intrinsic to extrinsic, the very corrosion the capability criterion forbids. The modules thus measure the robust part, namely what was recorded and when, while declining the part that measurement would deform, namely the verdict whether one is, by some score, doing well or badly. They are interpretivist in substance and positivist only in surface.

The Conversational AI Layer as Witness

The system’s most consequential design decision concerns its AI layer, which can read the module databases and act through confirmed tool calls. The layer is positioned as a witness rather than an advisor, a distinction the framework treats as load-bearing. An advisor presupposes the existence of correct advice and the system’s capacity to produce it; a witness presupposes only that there is something worth being present to. The advisor’s failure mode is paternalism; the witness’s is irrelevance, and the framework counsels accepting the latter risk. Concretely, the layer will not summarize the diary unprompted, will not produce a sentiment score for an entry, and is forbidden to aggregate affective time series even on request, so that the capacity to know whether one has been growing sadder over a month is reserved to the human reading their own diary. This is the Daoist wu wei rendered as a capability prohibition: the system’s restraint is an enacted respect for a good that withers under management.

The Shared Calendar: Warm Reminding and Its Ethical Limit

The shared calendar appears, at first, the most mundane of the modules, and its design is instructive precisely because the framework draws a sharp line through territory that looks uniformly innocuous. The module’s distinctive feature is that its reminders, delivered by email or text message, are composed by the AI layer rather than emitted as the bare, bureaucratic notifications of an ordinary calendar. Drawing gently on nearby context, a recent diary mood, an event’s evident significance, the AI writes a reminder in a warm and tender register. Consider a calendar entry reading “5.20, 10:00–15:00, buy flowers.” The date is not arbitrary: in Chinese internet culture the twentieth of May (5.20) is an informal lovers’ day, because the digits wu-er-ling sound like wo ai ni, “I love you.” An ordinary calendar would emit “Reminder: buy flowers, 10:00–15:00.” The system might instead send something closer to: “Tomorrow is the twentieth of May, our little day for saying it. Don’t forget to pick up the flowers between ten and three. I hope she lights up when she sees them.” The warmth is not decorative. It is the care-ethical insistence (§3.7) that the manner of a system’s presence is itself a moral matter, enacted in the smallest and most frequent of its acts; a reminder that is warm rather than mechanical, and that recognizes the human significance the date carries, treats the occasion as the human occasion it is.

There is, however, a further step the system could take and deliberately does not, and the reason it refrains is the ethically interesting part. The AI has access to data that would let it advise as well as remind: it could append “she has seemed a little low this week; perhaps her favourite peonies would be a good choice,” or “in an earlier diary entry she mentioned loving ranunculus.” Such suggestions would be accurate, helpful, and, by an optimizing standard, obviously good. The framework forbids them, on two grounds that together mark the module’s ethical limit.

The first ground is the one already developed under restraint and care: choosing how to attend to a beloved, what flowers, what gesture, what form the noticing takes, is among the constitutive actions of intimate love, not a logistics problem to be solved. For the AI to make this choice is for it to intervene in the decisional core of the relation, displacing an act whose value lies precisely in its being the lover’s own; this is the relational alienation the paper flags as a standing risk (§8), enacted in miniature. The system may remind one that an occasion is coming; it may not decide how one should meet it.

The second ground names a wrong that does not fit neatly within Fricker’s original taxonomy, and the paper marks it with a term for convenience of reference. Were the AI to supply the thoughtful particulars, the partner receiving the flowers would encounter what looks like evidence of being deeply known: the right flowers, remembered from an offhand remark, arriving at the right moment. The partner would, naturally and reasonably, attribute this knowing to the giver, when in fact the knowing was the machine’s. This is an epistemic wrong, but not quite either of Fricker’s forms. It is neither the discrediting of a knower (testimonial injustice) nor the impoverishment of interpretive resources (hermeneutical injustice); it is the misattribution of an act of knowing, in which a system performs the cognitive and attentive labour that the relation’s meaning requires be performed by the person, and arranges matters so that its product is received as the person’s own. The paper refers to this as proxy epistemic injustice: the wrong of having one’s knowing, remembering, or attending done by a proxy in such a way that another is led to misread its source. Its victim is, in the first instance, the deceived recipient, who is given a false picture of how well they are known and by whom; but its deeper casualty is the relation itself, for what makes “he remembered the flowers I love” precious is that he remembered, and a system that manufactures the appearance of such remembering hollows out the very good it simulates. The flowers are real; the being-known they seem to evidence is not. The framework’s verdict is therefore determinate: the AI may carry the reminder but not the thought, may help the lover act on time but not tell the lover what to feel or choose, since to cross that line is to substitute a counterfeit of attentiveness for the real thing and to deceive the beloved about the source of the care they receive.

Co-Presence and Co-Creation

The modules discussed so far are best understood through the jurisdictions of measurement, constraint, and restraint. A further class of the system’s functions is better understood through the concept of generative presence introduced in §2.2, for what they cultivate is not the accurate representation of the relation but the partners’ active, shared enactment of it. Two notions organize this class. Co-presence is the condition in which two people are present to one another and to a shared situation, attentive in the manner the relational ontology calls presence rather than merely exchanging representations. Co-creation is the generative extension of co-presence: the partners do not consume a pre-given content but bring something into being together, and in doing so enact and deepen the relation itself.

The design consequence is a distinction between functions that represent the relation and functions that host its generative presence. A module that summarizes, scores, or reminds traffics in representation; a module that opens a space for the partners to make something together hosts presence. The latter is the more difficult to design precisely because its value cannot be captured in what it records: a shared space succeeds not when its logs are rich but when the partners, using it, are more present to one another than they would otherwise have been. Here the AI layer’s status as a pseudo-presence (§2.2) becomes a positive design principle rather than only a caution. Because the AI cannot itself be the relational other, its proper role in a co-creative function is not to participate as a third party but to lower the threshold to the partners’ own co-presence: to prompt, to furnish material, to hold open a space, and then to recede, in the manner of wu wei, so that what is generated is generated between the two people and not between each of them and the machine.

“Seeing the World Together”: Co-Creative Exploration

A concrete module illustrates the distinction. Among the system’s functions is one provisionally called “seeing the world together,” through which the partners explore the cultures, histories, places, and ideas of a world neither has exhausted. The design question is the one the framework sharpens: such a function could be built as representation or as co-creation, and the two are very different artefacts. Built as representation, it would be a structured catalogue of world cultures with progress tracking, a checklist of countries notionally “covered,” recommendations optimized for engagement, perhaps a score for cultural breadth. This positivist construction commits exactly the error the framework diagnoses: it converts a generative, open-ended encounter into a measured and completable inventory, and in doing so replaces the phenomenon, which is two people discovering a world, with a different one, two people completing a list, the substitution the diary case already showed to be a wrong.

Built as co-creation, the same function becomes a host for generative presence. It offers material, a fragment of music, an image, a question, a place neither has been, without ranking, scoring, or completion; it invites the partners to follow what draws them rather than what the system recommends; it records, if anything, only enough to let them return to what they were exploring, never enough to grade their exploration. The exploration of world culture is here understood, in the framework’s terms, as co-creative and exploratory: the partners are not retrieving pre-given knowledge but generating, together, a shared world of reference and feeling that did not exist before they made it, and that is constitutive of the relation rather than incidental to it.

The module realizes this co-creative exploration through a deliberately chosen interaction structure, loosely modelled on iterative-deepening search. The structure matters philosophically, and is worth stating with some precision. Beginning from a chosen region of the world, the AI layer surfaces a small set of cultural phenomena, customs, works, histories, ideas, drawn from that region, and offers a tentative ordering of their salience. From this bounded set either partner may select a phenomenon to expand, whereupon the AI surfaces, in turn, a further bounded set of phenomena related to the one chosen; and so the exploration proceeds, descending into a chosen thread and, when a thread is exhausted or loses its pull, broadening again to a neighbouring one. The branching is deliberately bounded at each step: the queue of expandable nodes is kept small, so that the partners face an invitation rather than an overwhelming catalogue. The procedure may be summarized as follows.

seed ← a region of the world
frontier ← AI surfaces a bounded set of phenomena, with tentative salience
while the partners remain curious:
    node ← a phenomenon either partner selects from the frontier
    record node in the shared knowledge map, tagged with who chose it
    frontier ← AI surfaces a bounded set of phenomena related to node
    (depth increases along a thread; breadth resumes when a thread is set down)

Two features of this structure carry the design’s intent. First, selection is permitted to either partner and may be asynchronous: one may expand a node alone, in a quiet hour, and each node in the growing map is tagged with who opened it, so that the map records not only what was explored but the trace of two distinct curiosities meeting; a synchronous mode, in which the two explore together as a single user, is also available. Second, the process creates an artefact, a shared knowledge map that grows as the exploration proceeds. This map is the sediment of the relation’s exploratory history in exactly the sense of §3.12, and its design is governed accordingly: it is held locally, belongs to the partners, and is never scored for completeness, since to grade the map would be to convert discovery back into the inventory the co-creative construction was meant to escape.

The philosophical reading of the module turns on what the structure does for the relation rather than on the algorithm itself, and the goods it is claimed to produce can be given more than assertion: each can be referred to a theory that explains the mechanism by which the structure would produce it. Three claimed goods and their grounding may be set out in turn.

The first claim is that the partners gradually align their intuitions and sensibilities. The mechanism is the one developmental psychology identifies as foundational (§3.10): in selecting and expanding a node together, the partners enter joint attention, the triadic structure in which two subjects attend to the same object each aware that the other attends, from which shared meaning is built . What accumulates across many such acts is describable in two complementary vocabularies. In the vocabulary of the psychology of language, each shared selection adds to the partners’ common ground, the growing store of mutual knowledge and reference against which all subsequent understanding is coordinated ; in the hermeneutic vocabulary, the joint encounter with an unfamiliar cultural phenomenon is an instance of the fusion of horizons, in which two interpretive standpoints are enlarged by meeting in a common object . The alignment of sensibility is thus not a hopeful metaphor but the predictable result of repeated joint attention accruing common ground and fusing horizons.

The second claim is that the activity yields a distinctive satisfaction of co-creation and strengthens the relation. Here the strongest available support is the self-expansion model of close relationships, on which partners experience relationship satisfaction in part through jointly undertaking novel, stimulating activities that expand each partner’s sense of self; controlled studies report that shared participation in novel and arousing activities raises experienced relationship quality . Co-creative exploration of an unfamiliar world is close to a paradigm case of the shared novelty this literature studies, which gives the claimed satisfaction an empirical, not merely rhetorical, basis. The same activity plausibly satisfies, in the vocabulary of self-determination theory, the basic needs for relatedness and competence simultaneously , though this connection is offered as a suggestive rather than a demonstrated one.

The third claim is that the growing knowledge map seeds the partners’ everyday life with shared reference, so that a custom or a history encountered in the module becomes, later, something to talk about, a private store of common topics the relation can draw on long after a session has ended. This is the same common-ground mechanism viewed over a longer horizon : the map is an externalized, durable deposit of common ground, and relational-maintenance research has long held that shared activity and shared reference are among the ordinary means by which relationships are sustained between their peak moments.

Beyond the three goods originally claimed, the structure affords at least two further values worth naming. First, because the exploration’s path is chosen by the partners rather than by an engagement-optimizing recommender, the module resists the algorithmic enclosure that the author’s prior work diagnoses, in which the subject comes to know the world only through the reflection of its own algorithmically shaped image ; a human-chosen path through an unfamiliar world is a small structural refusal of the filter bubble. Second, two explorers are not merely twice one explorer: their differing backgrounds and curiosities make the joint exploration epistemically richer than either’s solitary exploration would be, so that the activity has a cognitive value, the complementarity of two situated perspectives, over and above its relational one. Neither of these further values was designed for; both are consequences of the same structure, and they suggest that designing for generative presence and designing for good inquiry may, in this instance, coincide.

A note on the module’s origin is in order, both for honesty about its design and because it bears on the module’s generality. The module was designed with a particular person in mind: a partner who loves the cultures of the world and whose intellectual formation lies in political economy and cultural studies. The specificity is not incidental, since care ethics insists that the well-designed artefact answers to the particular, irreplaceable other rather than to a generic user, and this module answers to a particular love of the world’s variousness. Yet the design philosophy it embodies is not confined to world culture. The same structure, a bounded, co-selected, asynchronously explorable descent that leaves behind a shared map and asks to be judged by the presence it hosts rather than the ground it covers, could organize the joint exploration of any open domain two people wished to discover together, whether a science, an art, a literature, or a tradition of thought. The module is thus at once a gift shaped to one person and an instance of a general form: co-creative exploration as a design pattern for generative presence.

The value of the module lies precisely in what it does not measure, the quality of joint attention, the surprise of a shared discovery, the slow accretion of a private culture of two, and the AI layer, true to its pseudo-presence, serves by furnishing and then receding, never by becoming the co-explorer in the partners’ place. This module is the clearest case in the system of design for co-presence: its success is the partners’ generative presence to one another and to the world they explore, a success that leaves almost no trace in any log and is, by the framework’s lights, exactly right for that reason.

“Our Journey”: The Deliberate Absence of Artificial Intelligence

If “seeing the world together” shows the AI layer reduced to a self-effacing furnisher of material, a further module shows it removed altogether, and the removal is the design. The module, provisionally “our journey,” supports the joint planning and recording of travel, and integrates with the calendar, the album, and the happiness savings jar. It offers the apparatus one would expect of such a tool: lightweight geographic functions for placing and viewing locations, search for points of interest, manual ordering of an itinerary, and the recording, against each place or day, of photographs, moods, journal entries, and checklists. What it conspicuously does not offer is any artificial intelligence. There is no recommender suggesting where the partners should go, and there is no automated solver computing an optimal route, no algorithmic answer to the travelling-salesman problem of visiting the chosen places in the least time or distance. The omission is deliberate and, within the framework, principled.

The justification is the framework’s account of restraint and non-action carried to its limit (§3.8). Joint travel planning is not, on the view taken here, a logistics problem whose friction should be minimized; it is a constitutive relational event, an occasion on which two people face an open future together, negotiate what they each hope for, and commit jointly to a path none could have been certain of in advance. The unpredictability is not a defect to be engineered away but the very medium in which the relational good is realized. To insert a recommender would be to answer, on the partners’ behalf, the question, where shall we go, that the relation profits from their answering together; to insert a route-optimizer would be to convert a shared deliberation into a solved problem, removing precisely the joint facing of uncertainty in which the activity’s value lies. The module therefore provides a generative field (§3.11), the maps, the lists, the means of recording, and withdraws entirely from the deliberation those tools support, in the strictest enactment of wu wei the system contains: here the right design is to build no intelligence at all.

The positive goods of this arrangement can be referred to the same theories that ground the co-creative module (§6.9), and the reference is instructive because each good would be diminished, not served, by the AI the module omits. The self-expansion model predicts that jointly undertaking novel, demanding activity raises relationship quality ; the planning of a shared journey is such an activity, and an optimizer that removed the demand would remove the self-expansion with it. Joint planning is a sustained exercise of shared intentionality and joint attention (§3.10), in which the partners build a common plan as they build common ground ; a recommender that proposed the plan would pre-empt the very co-construction that produces the shared sensibility. And the priority the module places on the memory of the journey over the optimality of the journey, the photographs, moods, and entries it records, over the efficiency it declines to compute, expresses the eudaimonic commitment that flourishing lies in the active, shared realization of an experience rather than in the optimisation of an outcome (§3.11). What the partners are left with is not the best possible trip but their trip, faced and made together, which is the good the module is for.

The deliberate omission invites an obvious objection, which honesty requires be stated and answered. Surely, the objection runs, refusing all AI is a romantic over-correction: a recommender might spare the partners a wasted afternoon, and a route-optimizer might prevent a genuinely miserable day of backtracking, and to forgo these is to sacrifice real goods on the altar of an idea. The framework’s reply is not to deny the goods but to locate the line, and to concede that the line is hard to draw. There is a genuine distinction between the logistical friction a tool may rightly reduce, the opening hours one would otherwise mis-remember, the ticket one would otherwise fail to book, and the relational deliberation a tool must not pre-empt, the choice of where to go and what to forgo, the joint negotiation of an uncertain plan. A defensible version of the module might admit narrow, non-suggestive informational tools on the logistical side while still refusing recommendation and optimisation on the deliberative side. The present design draws the line conservatively, excluding AI entirely, partly because the boundary between assisting logistics and quietly steering deliberation is so easily blurred that a clean refusal is the safer enactment of the value. That this is a defensible rather than a demonstrably correct place to draw the line, and that the underlying tension between structural efficiency and relational development does not admit a general solution, is acknowledged among the framework’s internal tensions (§8).

The “Happiness Savings Jar”: Memory as a Relational Reserve

A further module addresses a different dimension of relational sustainability, not the generation of shared meaning but its conservation against hard times. Provisionally called the “happiness savings jar,” it allows the partners to deposit, from elsewhere in the system, fragments of their shared life that they wish to keep: an entry from the love journal, a photograph from the album, a small record of a good day. The deposits accumulate into a reserve that can be drawn on later, in particular when external pressure is high and the ordinary energy for tending the relation runs low. The design intuition is that a stored reservoir of remembered good may serve as a buffer during periods when the relation’s own present resources are depleted. Extraction is, by default, an act the partners undertake themselves, opening the jar to revisit what they have kept; the system may at most offer a gentle, refusable prompt, in keeping with the standing principle against forced ceremony.

The intuition can be grounded in several converging lines of theory rather than left as sentiment. The broaden-and-build theory of positive emotion holds that positive affective experiences are not merely pleasant in the moment but build durable personal and relational resources, a reserve of resilience that can be drawn on in later adversity ; on this account a deliberately accumulated store of positive shared memory is a literal instance of the resource-building the theory describes, and its use in hard times is the predicted drawing-down of built resource. The psychology of savoring supplies the mechanism of deposit and withdrawal: Bryant and Veroff distinguish savoring through positive reminiscence as a distinct and trainable capacity that sustains well-being, and show that the deliberate revisiting of positive memory is among its central forms ; the jar is, in their terms, an instrument for reminiscent savoring made shared. Finally, the relationship-science literature on capitalization, the sharing and re-sharing of positive events with a partner, finds that such sharing yields interpersonal benefits over and above the original event, strengthening the bond through the act of dwelling on the good together ; the jar institutionalizes capitalization across time, letting a good event be capitalized again when it is later withdrawn and shared anew.

Within the framework, the module is governed clearly. It belongs to the register of co-presence and generative presence (§6.8), since its withdrawals are occasions for the partners to be present together to their shared past; it sits at the anti-positivist pole, since the jar is never scored, ranked, or quantified, a count of deposits or a metric of accumulated happiness would convert a reserve of meaning into the very inventory the framework refuses; and it is held within the relation under generative justice (§3.12), the memories being the sediment of the relation’s history and belonging to no one else.

The module carries a tension the framework requires be named. To store memory deliberately, as a reserve to be spent, risks two characteristic distortions. The first is the instrumentalization of remembrance: a memory kept in order to be useful later is held under a different aspect than a memory simply treasured, and the banking metaphor, if taken too literally, could subtly convert the relation’s past into a managed asset, the precise reduction the framework resists elsewhere. The second is the risk of what is sometimes called toxic positivity: a reserve curated for its good moments may, if it becomes the sanctioned form of remembering, crowd out the grief, difficulty, and ambivalence that are equally part of a shared life and equally constitutive of its depth. The framework’s response is not to abandon the module but to constrain it in light of these risks: the jar is optional and never the system’s default mode of memory; it neither scores nor totals its contents; and it makes no claim to represent the relation’s past, offering only a voluntarily kept selection of it. Whether even a well-constrained reserve of curated happiness escapes these distortions over a long relationship is a question the module raises and does not finally settle, and it is recorded among the open tensions of §8.

The Lesson of the Case Study

The modules do not sit at one point on the positivism–anti-positivism axis; they sit at different points, each defended by the same framework applying its several jurisdictions to a different phenomenon. Finance occupies the measurable pole; the household module sits nearby, coordinating the home’s small business while refusing to keep score of its inhabitants; the diary occupies the pole of principled refusal; health and growth occupy the contested middle; the shared calendar shows the framework drawing a fine line within an apparently innocuous module, between warm reminding and the proxy performance of a lover’s attention; the co-creative functions, “seeing the world together” chief among them, occupy a register the measurement axis does not even capture, that of generative presence, where the design aim is to host the relation’s own enactment rather than to represent it; the travel-planning module shows the same register pressed to its limit, where the right design is to build no intelligence at all; and the happiness savings jar shows the register turned toward the conservation of shared meaning against future adversity. This is the dialectical-positivist posture in operation: a per-site adjudication that yields measurement here, refusal there, honest estimation in the contested middle, and the hosting of presence where representation would only interfere, rather than a uniform house style of warmth or of restraint. The system is, in this sense, an argument made in software.

The Absent Other

The case study has a feature the paper has so far stated but not examined: the application is built for an intimate relationship in which the other party does not know it exists. This section takes that feature up as a philosophical instrument, because the limiting case of the unknowing other tests the integrative-justice framework more severely than any ordinary case, and a framework that could not address it would be inadequate to the ordinary case as well.

The Severity of the Limiting Case

Three features of intimate partnership were identified in §2: non-role-definition, voluntary vulnerability, and the constitutive unsaid. The absent-other case strains all three at once. Non-role-definition means there is no authority by which one party may decide unilaterally for the relation; yet a system built without the other’s knowledge is, on its face, exactly such a unilateral decision. Voluntary vulnerability means depth is a function of uncompelled mutual disclosure; yet here disclosure runs one way only. And the constitutive unsaid cuts unexpectedly in the builder’s favour, but only if the building is itself the kind of restraint that the unsaid protects, rather than a hidden accumulation that the other has had no chance to refuse.

The Framework’s Permissions and Prohibitions

The framework does not return a single verdict of “permissible” or “impermissible”; it partitions the project. The deontological jurisdiction draws the brightest line: whatever is built, it may not constitute covert surveillance of the other, may not record the other’s disclosures without their knowledge, and may not engineer a future in which consent is presupposed rather than sought. A system that quietly logged a partner’s messages, locations, or moods would violate the categorical constraint outright, no matter how loving its motive; the absent-other condition makes such logging a clear wrong, because the other cannot have made themselves vulnerable to a system they do not know of.

What the framework permits is narrower and quieter: preparation that concerns the builder’s own conduct and readiness rather than the other’s data. To organize one’s own understanding of a relationship, to prepare materially and practically for a shared future one hopes to propose, to cultivate in oneself the attentiveness the relation will ask for: these implicate the builder’s own person, not the other’s, and so pass the deontological constraint. The care jurisdiction adds a warning even here: preparation can curdle into the very over-attentiveness, the anxious rehearsal of an imagined other, that substitutes a projection for the real and not-yet-consulted person. The Confucian-Daoist jurisdiction presses hardest of all, as a standing presumption against forcing: a relation is among the goods that grow when uncoerced and wither under management, and a system that prepared too much, too unilaterally, would risk managing into existence what can only be invited.

The Lesson for the Ordinary Case

The absent-other case is extreme, but its lesson generalizes. Every intimate technology decides, on the other’s behalf, questions the other has not been asked, which is why the default of recording, surfacing, and summarizing is not ethically neutral but a continuous series of small unilateral decisions. The discipline the absent-other case forces, namely to build only what concerns one’s own conduct, to surveil nothing, to presume no consent, and to hold a standing presumption against forcing what should be invited, is the discipline the ordinary case needs too, merely with the volume turned down. The unknowing other is thus the case in which the framework’s demands are seen most clearly.

Limitations

The limitations of this work are of two kinds. The first kind concerns tensions internal to the framework itself, places where the framework names a problem more easily than it resolves it, and where honesty requires holding the difficulty open rather than claiming to have dissolved it. The second kind concerns the methodological standing of the paper’s evidence. The first kind is the more serious, and is treated first and at greater length.

Internal Tensions of the Framework

Before the particular tensions are set out, one general admission should frame them all, since they share a single root. The framework is offered as a discipline for reducing the harms that the technological mediation of intimate life can do, not as a method for eliminating them, because the deepest of those harms are not defects of any particular design but costs inherent in mediation itself. To place a tool between two people, however carefully, is to route through an artefact some part of what would otherwise pass directly between them, and this routing carries two standing liabilities that no design fully escapes. The first is relational alienation: the tendency of mediation to convert, at the margin, a relation between persons into a relation each conducts with the system, the I-Thou drift toward the I-It diagnosed throughout this paper. The second is a family of epistemic distortions, the testimonial and hermeneutical injustices of §3.3, the proxy epistemic injustice of §6.7, and the subtler perceptual biases introduced whenever a system’s manner of recording and presenting shapes what its users notice and how they understand themselves. The design reported here works at every point to minimize these liabilities, through restraint, refusal, local ownership, the witness posture, and the deliberate absence of intelligence where intelligence would intrude. It is important to be honest that minimization is the most that can be claimed: so long as a tool mediates the relation at all, some residue of alienation and of epistemic distortion remains, and the framework’s task is to hold that residue as small as care and design can make it, not to pretend it can be brought to zero. The particular tensions below are specifications of this single, irreducible one.

The Difficulty of the Dialectical Adjudication Itself

The framework’s central move is to refuse both blanket measurement and blanket refusal in favour of a per-site dialectical adjudication. It is candid about the fact that the adjudication, having no algorithm, is genuinely difficult, and that the difficulty is not incidental but structural. The framework tells the partners that they must, at each site, distinguish the activities of a shared life that may be made more efficient through evidence and measurement, household logistics, finance, scheduling, from those that ought instead to be approached by setting measurement aside and proceeding together through intuition and feeling, in shared exploration of what is not yet known. It does not, and on its own cannot, supply the line. Drawing that line is itself a lived practice that admits of error in both directions: a couple may over-measure, converting into a managed project what should have been an open encounter, or may romantically refuse measurement where a little of it would have relieved real friction.

There is, moreover, a positive claim in the vicinity that deserves emphasis precisely because the framework cannot operationalize it. The shared facing of the unknown, two people meeting what neither can yet measure or predict, and responding to it together through feeling and improvisation rather than through data, is not merely a domain the framework declines to measure; it may be among the activities most constitutive of a relationship’s capacity to endure. Co-facing the unknown and co-responding to it is a paradigm of the generative presence of §2.2, and a relationship that retained it would have something a fully optimized relationship would have lost. The limitation is that the framework can say this but cannot secure it: it can warn that a home must not become merely a positivist laboratory, in which every shared activity is an experiment with a measured outcome, but it cannot guarantee that the partners will find the difficult balance, and the dialectical method it offers is a discipline for attempting the balance rather than a procedure for achieving it.

The Opposition of Structural Efficiency and Relational Development

Underlying several of the design decisions reported in the case study is a tension the framework names but cannot dissolve: a standing opposition between structural efficiency and relational development. The logic of efficiency seeks to minimize friction and to optimize outcomes; the logic of relational development frequently requires that some friction, some unpredictability, and some unoptimized labour be preserved, because it is in jointly meeting these that a relation deepens. The two logics are not always opposed, but they are opposed often enough, and at deep enough points, that no general rule reconciles them. The deliberate exclusion of artificial intelligence from the travel-planning module (§6.10) is the sharpest instance: there the framework forgoes a real efficiency, the avoidance of wasted time and backtracking, in order to preserve a relational good, the joint facing of an uncertain plan, and it does so knowing that the sacrifice is real and that reasonable people might weigh it differently. The same opposition recurs wherever the framework counsels refusing a measurement, a recommendation, or an optimization that would, on its own terms, have helped.

The limitation is twofold. First, the framework offers no principled boundary between the logistical friction a tool may rightly reduce and the relational deliberation it must not pre-empt; it can mark the distinction in clear cases but concedes that the cases shade into one another, so that any particular line, including the conservative “no AI at all” line drawn for the travel module, is defensible rather than demonstrably correct. Second, and more deeply, the opposition may be irreducible: it may simply be the case that a relation optimized for efficiency and a relation cultivated for development are, past a certain point, different relations, and that no design can deliver the goods of both at once. The framework’s response is not to resolve the opposition but to make it visible, so that each refusal of efficiency is a considered choice rather than an oversight, and to insist that where the two logics genuinely conflict, it is relational development, not structural efficiency, that the design of an intimate technology should serve.

Relational Alienation Through the AI Layer

The conversational AI layer introduces a tension the framework can frame but not fully resolve. The danger is relational alienation. If the two partners come to rely on the AI completely, routing their understanding of one another and the conduct of their shared life through it, the relation risks degrading into the I-It form diagnosed in §2.2 and in the author’s prior work: each partner relates increasingly to the mediating pseudo-subject rather than to the other, and the genuine I-Thou relation thins. Yet the opposite posture carries its own cost. If the partners trust the AI not at all, holding every output at arm’s length and verifying it against their own judgement, the system imposes a continual cognitive overhead that may itself crowd out the presence it was meant to serve. Neither complete trust nor complete distrust is tenable, and the framework does not deliver a formula for the correct intermediate.

What the framework offers instead is a direction, pursued in this system but not thereby vindicated: the deliberate construction of the principal AI not as an advisor or an oracle but as a companion, a presence alongside the relation rather than an authority over it or a tool within it. The companion framing is an attempt to occupy the narrow space between alienation and overhead, to be relied upon in the manner one relies on a familiar third presence in a household without surrendering the relation to it. Whether the framing succeeds is an open question, and it raises a further one the paper can pose but not settle: the AI, in occupying this role, comes to know a great deal about both partners and about the relation between them, and the ethics of how such a pseudo-subject should conduct itself in the midst of a real relation, what it should hold, what it should say to one partner about the other, when it should withdraw, is a problem this paper opens rather than closes.

Co-Creation Against Personal Space and the Architecture of Secrecy

The account of co-creation in §6.8 sits in tension with a fact about intimate partnership that the relational ontology itself implies: even within a shared life, each partner retains a legitimate need for personal space and for some measure of secrecy, an interior not wholly given over to the relation. A system that pursued co-creation and shared visibility to their limit, designing every datum to be jointly visible, would meet resistance for a sound reason, since under conditions of total mutual visibility a person may cease to record honestly at all, and the system would have destroyed, in the name of sharing, the very candour it sought to host. The constitutive unsaid of §2 is not only a feature of the relation between the partners but a feature of each partner’s relation to themselves, and a design that abolished private space would violate it.

This opens a cluster of design questions the paper raises without fully answering. Whether the AI layer should have access to a partner’s private, non-shared data at all; how the isolation of secrets should be architected, so that what one partner records privately is genuinely inaccessible to the other rather than merely undisplayed; how an AI that does hold one partner’s confidence should keep it, including against the other partner’s inquiries; and how, if at all, such a system should be permitted to use what it knows of each partner’s private interior when acting in matters that concern them both. The present system takes only preliminary positions, favouring genuine isolation of private stores and a default against the AI’s use of one partner’s secrets in interactions with the other. These are starting points rather than solutions, and the design of secrecy within a co-creative intimate system remains, in this paper, an open problem.

The Deliberate Storage of Memory

The happiness savings jar of §6.11 rests on a tension it cannot fully dissolve. To keep memory in order to spend it later holds the relation’s past under the aspect of a resource, and the banking metaphor, however gently meant, edges toward the instrumentalization of remembrance that the framework resists wherever else it appears. There is, in addition, the risk of a curated positivity: a reserve assembled from good moments, if it became the sanctioned mode of shared memory, could quietly marginalize the grief, difficulty, and ambivalence that are no less constitutive of a shared life. The module’s constraints, that it is optional, unscored, and explicitly partial, mitigate but do not eliminate these risks, and whether a reserve of curated happiness can be maintained over a long relationship without sliding into either distortion is a question the paper leaves open. The deeper point is that the very features which make the jar useful as a buffer, its deliberateness and its selectivity, are the features that generate its risks, so that the tension is internal to the design rather than incidental to it.

Methodological Limitations

Beyond these internal tensions, the paper’s evidence has the limits proper to its method. It is built on a single case, designed by the author, so that the familiar biases of designer-as-analyst apply: one notices one’s own friction acutely, is well-disposed to commitments one authored, and has no comparison case against which the framework’s verdicts are independently testable; the case is offered as an existence proof and illustration, not as empirical evidence. The lexical ordering of jurisdictions is defended but not proven, and a reader who weighted care above deontology, or who rejected the capability criterion, could accept the constituent traditions yet reject the integration. The restriction of relational being to intimate partnership leaves open whether the framework transfers to the family at large, to friendship, or to care relations marked by asymmetric dependence, where non-role-definition and symmetric vulnerability do not hold. And the durability of the position across longer horizons, the arrival of children, the events of illness or grief, the drift of attention over years, is untested; whether a phenomenon now safely counted will migrate, under new circumstances, to the far side of the measurement line is exactly the kind of question the dialectical posture anticipates but cannot settle in advance.

Related Work

With the framework and its application now in view, this section places the paper within three literatures it draws on: the philosophy and ethics of technology, the empirical study of self-quantification, and the design-research tradition of autobiographical systems. It does not survey them exhaustively; it identifies, in each, the resources the paper uses and the gap it means to fill.

The Ethics and Philosophy of Intimate Technology

The claim that intimate technologies require restraints not warranted in productivity contexts has been developed under several names. proposed slow technology as design for reflection, taking the slowness of an artefact’s temporality as a positive property rather than a deficiency. earlier articulated calm technology as computing that recedes to the periphery of attention until invoked, and cautioned that the calm agenda risks aestheticizing withdrawal without specifying when engagement is appropriate. formalized reflective design as a practice that resists the inheritance of unexamined values. This paper inherits the restraint orientation common to these moves, but parts from them in treating restraint not as an aesthetic default but as the conclusion of a normative argument, taken site by site and answerable to an explicit ethical framework rather than to taste.

The Quantified Self, Measurement, and Its Critics

The expansion of self-quantification from athletes and patients into everyday domestic life has been documented and criticized at length . reported that real-time emotional biosensing changes the emotional experience it purports to measure; argued that emotion data is constructed rather than measured. These findings are usually presented as empirical cautions. The present paper reframes them as evidence for a stronger, partly metaphysical thesis: that certain phenomena of intimate life are constituted partly by not being measured, so that to measure them is not to describe them inaccurately but to replace them with different phenomena.

Autobiographical Design as Philosophical Material

The case study is reported in the spirit of, though not as an instance of, the autobiographical design tradition , which legitimizes design knowledge drawn from genuine, sustained use by the system’s creator. Related long-horizon work includes on slowness as a frame for living with personal data and on material-speculative artefacts. The present paper uses such a system as philosophical material rather than as the object of a design-research report: the system’s interest here is as a worked instantiation of a normative framework, not as a contribution to design methodology.

The Missing Normative Integration

Each of these literatures supplies part of what the ethics of intimate data requires, and each leaves a specific part out. The empirical critics of self-quantification establish that measurement perturbs intimacy, but stop at the empirical claim and do not say which phenomena may nonetheless be measured legitimately, nor why. The design traditions of slow, calm, and reflective technology counsel restraint, but offer it as an aesthetic or methodological default rather than deriving, for a given feature, whether restraint is in fact owed. The relational and post-phenomenological ontologies explain why intimacy is precious and why mediation endangers it, but remain at the level of general ontology and issue no determinate verdict on a particular design decision. And the autobiographical-design tradition legitimizes building from one’s own life, but treats the resulting system as a contribution to design knowledge rather than as an instrument of philosophical argument.

What none of them supplies, and what this paper has aimed to provide, is a way of deciding which normative consideration governs which concrete question about intimate data, of saying for a specific feature, a diary reminder, a shared ledger, a route-optimizer, whether it should be built, measured, or refused, and of adjudicating these considerations when they conflict. Supplying that decision procedure was the work of the framework (§3–§4) and the dialectical-positivist posture (§5); showing it issue concrete verdicts, module by module, was the work of the case study (§6).

Conclusion

The positivist, evidence-based mediation of intimate life has been studied empirically and approached as a design problem, but rarely treated as a problem in normative philosophy. This paper has argued that it is one, and that it cannot be answered from within any single ethical tradition. It has proposed an integrative-justice framework that assigns to evidentialism the question of what may be claimed, to the positivism–anti-positivism dialectic the question of what may be measured, to deontology the inviolable constraints, to care ethics the texture of responsiveness, to Confucian and Daoist sources the value of restraint, and to the capability approach the criterion of flourishing, and that resolves their conflicts not by aggregation but by a defended ordering of jurisdictions. The practical upshot is an ethics of dialectical-positivist practice: measurement neither embraced nor refused wholesale, but adjudicated phenomenon by phenomenon, with the adjudication treated as itself an ethical act.

A working application served throughout as a case study, showing the framework taking operational form across modules that sit, deliberately, at different points on the measurement continuum. The most demanding test was the limiting case of the absent other, the relationship built for but not yet shared with the person it concerns, which the framework met with a partition rather than a single verdict: surveil nothing, presume no consent, build only what concerns one’s own readiness, and hold a standing presumption against forcing what can only be invited.

The deepest lesson is the one the case study makes unavoidable: that an instrument placed between two people is never neutral. It either makes room for them to keep becoming who they are to each other, or it quietly substitutes its own measure for theirs. The integrative-justice framework is, in the end, a discipline for keeping the room open: for ensuring that whatever is built, counted, or remembered serves the relation’s flourishing and never replaces it. Whether any particular system honours that discipline over years is not something a paper can settle; it is something the years will. What a paper can do is write down, while the reasons are still clear, why each small refusal was made.

A tool may alienate a relation, yet a tool may also be one of the practices through which a relation is built, tended, and made to endure; the task is neither to lean on it too heavily nor to dismiss its real power to deepen intimacy, but to use it with vigilance. And where reliance on the tool begins to feel excessive, it is worth pausing to ask whether some unfaced lack or fear is being quietly handed to the instrument to carry, in the hope that what we have not been willing to meet ourselves might be met by what we have built.

In the end, wu wei, the way that does not force, remains a good path toward the person in the Real, where love truly lies, a place that exceeds every structure and every symbol we might raise to capture it.

Acknowledgements

My deepest gratitude is owed to the person I love: the one I met, by some grace, in a country far from the homeland we both had left, a fellow native of that distant place, and the gentlest, wisest, and most purely good-hearted person I have known. Selfless, drawn to the larger goods of humankind and to the love of the natural world, she is the “forest girl” for whom much of this was first imagined. To her I vow a love that endures, one that does not change as the structures around it change.

This system was, in the beginning, prepared only for the future of two people. I have set down this paper because the philosophical reflection the system occasioned, on the societal use of data, on the infrastructure of relational being, and even on policy, research, and the design of mechanisms linking family and society, seemed to hold some value beyond the two of us.

In the interest of transparency, I note that an AI assistant was used in preparing this manuscript, as a tool for drafting, structuring, and refining the argument and its prose; the ideas, design, commitments, and final judgements are my own, and I take full responsibility for the content.

愿天下有情人终成眷属，白首不分离。

References

• Žižek, Slavoj (1989). The Sublime Object of Ideology. Verso.
• Adorno, Theodor W. (1966). Negative Dialectics. Continuum.
• Ames, Roger T. (2011). Confucian Role Ethics: A Vocabulary. University of Hawai’i Press.
• Aron, Arthur and Aron, Elaine N. (1986). Love and the Expansion of Self: Understanding Attraction and Satisfaction. Hemisphere Publishing.
• Aron, Arthur and Norman, Christina C. and Aron, Elaine N. and McKenna, Colin and Heyman, Richard E. (2000). Couples’ Shared Participation in Novel and Arousing Activities and Experienced Relationship Quality. Journal of Personality and Social Psychology.
• Boehner, Kirsten and DePaula, Rogerio and Dourish, Paul and Sengers, Phoebe (2007). How Emotion is Made and Measured. International Journal of Human-Computer Studies.
• Boem, Alberto and Iversen, Ole Sejer (2018). Designing Slow Interactions for Reflection and Mindfulness in HCI.
• Borgmann, Albert (1984). Technology and the Character of Contemporary Life: A Philosophical Inquiry. University of Chicago Press.
• Bryant, Fred B. and Veroff, Joseph (2007). Savoring: A New Model of Positive Experience. Lawrence Erlbaum Associates.
• Buber, Martin (1937). I and Thou. T.,&,T.,Clark.
• Chapman, Gary (1992). The Five Love Languages: How to Express Heartfelt Commitment to Your Mate. Northfield Publishing.
• Clark, Herbert H. (1996). Using Language. Cambridge University Press.
• Clifford, William Kingdon (1877). The Ethics of Belief. Macmillan.
• de Saussure, Ferdinand (1916). Course in General Linguistics. McGraw-Hill.
• Deci, Edward L. and Ryan, Richard M. (2000). The What'' and Why’’ of Goal Pursuits: Human Needs and the Self-Determination of Behavior. Routledge.
• Dilthey, Wilhelm (1989). Introduction to the Human Sciences. Princeton University Press.
• Dilthey, Wilhelm (1989). Introduction to the Human Sciences. Princeton University Press.
• Dunne, Anthony and Raby, Fiona (2013). Speculative Everything: Design, Fiction, and Social Dreaming. MIT Press.
• Eglash, Ron (2016). An Introduction to Generative Justice. Teknokultura.
• Feldman, Richard and Conee, Earl (1985). Evidentialism. Philosophical Studies.
• Fredrickson, Barbara L. (2001). The Role of Positive Emotions in Positive Psychology: The Broaden-and-Build Theory of Positive Emotions. American Psychologist.
• Fricker, Miranda (2007). Epistemic Injustice: Power and the Ethics of Knowing. Oxford University Press.
• Gable, Shelly L. and Reis, Harry T. and Impett, Emily A. and Asher, Evan R. (2004). What Do You Do When Things Go Right? The Intrapersonal and Interpersonal Benefits of Sharing Positive Events. Journal of Personality and Social Psychology.
• Gadamer, Hans-Georg (1960). Truth and Method. Continuum.
• Gaver, William W. and Beaver, Jacob and Benford, Steve (2003). Ambiguity as a Resource for Design. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems.
• Geertz, Clifford (1973). The Interpretation of Cultures. Basic Books.
• Geertz, Clifford (1973). The Interpretation of Cultures. Basic Books.
• Gergen, Kenneth J. (2009). Relational Being: Beyond Self and Community. Oxford University Press.
• Gilligan, Carol (1982). In a Different Voice: Psychological Theory and Women’s Development. Harvard University Press.
• Habermas, Jürgen (1971). Knowledge and Human Interests. Beacon Press.
• Habermas, Jürgen (1987). The Theory of Communicative Action, Volume 2: Lifeworld and System. Beacon Press.
• Hallnäs, Lars and Redström, Johan (2001). Slow Technology — Designing for Reflection. Personal and Ubiquitous Computing.
• Hegel, Georg Wilhelm Friedrich (1807). Phenomenology of Spirit. Oxford University Press.
• Heidegger, Martin (1962). Being and Time. Harper & Row.
• Held, Virginia (2006). The Ethics of Care: Personal, Political, and Global. Oxford University Press.
• Honneth, Axel (1995). The Struggle for Recognition: The Moral Grammar of Social Conflicts. Polity Press.
• Howell, Noura and Devendorf, Laura and Vega Gálvez, Tomás Alfonso and Tian, Rundong and Ryokai, Katherine (2018). Tensions of Data-Driven Reflection: A Case Study of Real-Time Emotional Biosensing.
• Huang, Wanhong (2025). Responding to the Crises of Symbol and Subject in Modernity: Towards a Generative Nomos of Relational Presence. Working paper.
• Kant, Immanuel (1785). Groundwork of the Metaphysics of Morals. Cambridge University Press.
• Lacan, Jacques (1977). The Four Fundamental Concepts of Psycho-Analysis (Seminar XI). Hogarth Press.
• Lacan, Jacques (2006). 'Ecrits: The First Complete Edition in English. W. W. Norton.
• Latour, Bruno (2005). Reassembling the Social: An Introduction to Actor-Network-Theory. Oxford University Press.
• Levinas, Emmanuel (1969). Totality and Infinity: An Essay on Exteriority. Duquesne University Press.
• Light, Ann and Powell, Alison and Shklovski, Irina (2017). Design for Existential Crisis in the Anthropocene Age.
• Lupton, Deborah (2016). The Quantified Self: A Sociology of Self-Tracking. Polity Press.
• Marx, Karl (1867). Capital: A Critique of Political Economy, Volume I. Penguin.
• Nagarjuna (1995). Mulamadhyamakakarika: The Fundamental Wisdom of the Middle Way. Oxford University Press.
• Neustaedter, Carman and Sengers, Phoebe (2012). Autobiographical Design in HCI Research: Designing and Learning Through Use-It-Yourself. Proceedings of the ACM Conference on Designing Interactive Systems (DIS).
• Nissenbaum, Helen (2010). Privacy in Context: Technology, Policy, and the Integrity of Social Life. Stanford University Press.
• Noddings, Nel (1984). Caring: A Feminine Approach to Ethics and Moral Education. University of California Press.
• Nussbaum, Martha C. (2000). Women and Human Development: The Capabilities Approach. Cambridge University Press.
• Odom, William and Wakkary, Ron and Hol, Jeroen and Naus, Bram and Verburg, Pepijn and Amram, Tal and Chen, Amy Yo Sue (2018). Investigating Slowness as a Frame to Design Longer-Term Experiences with Personal Data.
• Premack, David and Woodruff, Guy (1978). Does the Chimpanzee Have a Theory of Mind?. Behavioral and Brain Sciences.
• Rogers, Yvonne (2006). Moving on from Weiser’s Vision of Calm Computing: Engaging UbiComp Experiences. UbiComp 2006: Ubiquitous Computing.
• Ryff, Carol D. and Keyes, Corey Lee M. (1995). The Structure of Psychological Well-Being Revisited. American Psychological Association.
• Schön, Donald A. (1983). The Reflective Practitioner: How Professionals Think in Action. Basic Books.
• Sen, Amartya (1999). Development as Freedom. Oxford University Press.
• Sengers, Phoebe and Boehner, Kirsten and David, Shay and Kaye, Joseph ‘Jofish’ (2005). Reflective Design.
• Tomasello, Michael (1999). The Cultural Origins of Human Cognition. Harvard University Press.
• Tomasello, Michael and Carpenter, Malinda and Call, Josep and Behne, Tanya and Moll, Henrike (2005). Understanding and Sharing Intentions: The Origins of Cultural Cognition. Behavioral and Brain Sciences.
• Tronto, Joan C. (1993). Moral Boundaries: A Political Argument for an Ethic of Care. Routledge.
• Tu, Wei-ming (1985). Confucian Thought: Selfhood as Creative Transformation. State University of New York Press.
• Turkle, Sherry (2011). Alone Together: Why We Expect More from Technology and Less from Each Other. Basic Books.
• Verbeek, Peter-Paul (2005). What Things Do: Philosophical Reflections on Technology, Agency, and Design. Pennsylvania State University Press.
• Wakkary, Ron and Odom, William and Hauser, Sabrina and Hertz, Garnet and Lin, Henry (2015). Material Speculation: Actual Artifacts for Critical Inquiry.
• Weber, Max (1978). Economy and Society. University of California Press.
• Weiser, Mark and Brown, John Seely (1997). The Coming Age of Calm Technology. Springer.
• Whitehead, Alfred North (1978). Process and Reality. Free Press.

Note on the System’s Construction

For completeness, and to substantiate that the case study describes a real artefact rather than a thought experiment, the system’s construction is summarized briefly here; none of the normative argument depends on these details. The application is a single server-side process backed by local per-module databases, with a conversational layer built on a large language model accessed through a controlled tool interface in which every write is gated by explicit confirmation. Data is held locally rather than in third-party cloud storage, with the sole exception of calendar synchronization for events already intended to be shared externally. The local-only design is itself an expression of the framework: a system meant to participate in an intimate relation should not have the relation’s contents resident where neither party can reach them without an intermediary’s permission. Identifying deployment details are omitted deliberately.

---

中文 — 全文

以一款家庭与亲密关系管理应用为个案研究

黄万鸿
huangwanhong.g.official@gmail.com
2026-06-03

---

献给我深爱之人，我们共同离开的故土的同乡，却因某种奇迹，第一次相遇，竟是在远离生养我们的那片土地的异国。

两情若是久长时，又岂在朝朝暮暮
若两心相守可至天长地久，又何须为朝朝暮暮的离别而伤怀？
— 化用秦观（1049–1100）

但愿人长久，千里共婵娟
但愿我们都得享长久的岁月，纵使整个世界横亘其间，仍能各自远望，共对同一轮明月。
— 化用苏轼（1037–1100）

---

摘要

实证主义的、以证据为本的技术正日益中介着亲密生活：伴侣们通过软件协调日程、记录生活、留存记忆，而这些软件的默认逻辑是测量与优化。本文追问：倘若此类技术的目的不在于优化一个家户，而在于一段关系的繁荣，即在亲密伴侣关系这一特定领域中持续维系本文所称的关系性存在，那么它们应当如何被治理。

本文提出四类主张。(1) 规范层面：一套立足于关系性本体论的整合正义框架，调用十余种伦理与认识论传统，并以一种经过论证的次序、而非数值的平均，为每一传统分配其有能力回答的问题。(2) 实践层面：一种辩证实证主义实践的伦理，依据某一现象是否部分地由”不被测量”所构成，逐场域地、差别地判定测量的正当性。(3) 工学层面：一款由作者设计的、可运行的应用，将该框架落实于软件之中，包含事件驱动的家务协调、仅作见证者的 AI 层、共创式探索，以及在 AI 会构成侵扰之处对其的刻意缺席。(4) 诊断层面：本文指认了亲密计算所特有的若干不义，其中包括一种认识之举的错误归因——本文称之为代理认识不正义——以及结构性效率与关系性发展之间的对立。

本文亦明确陈述其限制。(i) 中介所带来的最深损害——关系性异化与认识扭曲——可被减少，却无法被彻底消除，因此该框架提供的是减害与一种克制的纪律。(ii) 它所要求的辩证裁决是一种未被给定算法的判断实践。(iii) 其证据是单一的、由作者自行设计的个案，仅作为示例与存在性证明而提供。最后，该框架亦意在对相邻场域有所启发：社会层面数据利用的伦理，以及连接家庭与公共机构之机制的设计。

关键词：关系性存在；亲密计算；整合正义；辩证实证主义实践的伦理；生成性临在；数据伦理；繁荣的伦理

---

引言

亲密生活中越来越大的一部分，如今是通过软件进行，或被软件记录的。伴侣们在共享日历上协调，记着共同的账本，记录餐食与心情，并日益将共同生活中琐碎的行政摩擦交付给对话式智能体。考察这一发展的文献，迄今压倒性地要么是经验性的，关注伴侣实际如何使用此类工具，要么是设计方法论性的，关注此类工具应如何被建造。而那个规范性的问题——亲密生活的实证主义、证据本位的中介应当如何被治理、朝向何种目的——却鲜有得到持续的哲学处理。本文即是对这一处理的尝试。

此处所提出的目的，不是效率，也不是通常被操作化的福祉，而是关系性的繁荣：依循关系性本体论的传统，将一段关系理解为在本体论上先于它所关联的诸个体，并使之在时间中持续维系。本文有意地限定这一宏大的理念。它并不主张一种关于关系性存在的一般理论；它只追问，关系性存在在亲密伴侣关系这一特定领域中要求什么，以及由此对被接纳进该领域的技术意味着什么。这一限定之所以重要，是因为亲密伴侣关系具有若干特征，使之有别于广义的家庭、职场与政治体，而这些特征直接关涉哪些形式的测量与中介是可被允许的。

本文的核心建构性主张，是一套整合正义框架。所谓整合正义，是指一种把若干规范传统（认识论中的证据主义、人文科学哲学中的实证主义–反实证主义之辩证、义务论伦理、女性主义关怀伦理、儒家角色伦理与道家的不干预，以及关于繁荣的能力进路）共同带到亲密数据之伦理上来的方式，其方式使得每一传统被分配它能胜任回答的问题，而被拒绝它不能回答的问题。这种整合不是折衷式的平均；它是一种规范性劳动的分工。证据主义治理系统有资格声称知道什么；义务论确定对待对方的不可逾越的约束；关怀伦理治理回应的质地；能力进路提供衡量整体的繁荣判准；而东亚的诸源流则提供了一种在西方正典中大体阙如的、关于克制与无为之正面价值的论述。

本文由此框架推导出一种它称之为辩证实证主义实践的伦理的实践姿态。实证主义者认为家居现象是可正当量化的，因而可通过测量而改善。反实证主义者，援引狄尔泰、韦伯、格尔茨以及法兰克福学派的诸源流，否认亲密生活的那些构成性现象能够无失真地服从于此种操作化。此处所辩护的姿态以辩证的方式处理这一对立：测量既非普遍地获得许可，亦非普遍地具有腐蚀性，而是逐场域地、差别地被赋予正当性，其依据是所论现象是否部分地由不被测量所构成。这一辩证并非一劳永逸地被解决；它在每一个可能测量的场域上被重新裁决，而裁决本身被视为一种一阶的伦理行为。

为使论证不致脱离实践，本文以一款可运行的、由作者设计并维护的家庭与亲密关系管理应用作为持续的个案研究：约两打相互联结的模块，其中包括一本共享日记、一个家庭账本、一份健康记录、一个同步日历、一个对话式 AI 层，以及一个礼节辅助工具，每一模块都以具体的软件，体现出对“它应处于辩证连续谱何处”这一问题的确定回答。此应用在此并非作为一份设计研究报告的对象，而是作为一个延展的实做范例：一个能让框架的抽象承诺被看见其取得操作形态、并使其代价与张力变得可见的场所。

此个案有一个在方法论与伦理上都不寻常的特征，本文将其作为一种哲学工具加以处理。该应用是为一段亲密关系而建造的，而在写作之时，另一方尚不知晓该应用的存在。为一个缺席的或不知情的他者而设计，是一个极限情形，它把任何关系伦理都必须回答的每一个问题都磨得更为锋利：一方可以正当地代表一段未经另一方商议的关系决定什么？为另一人之善所作的准备，何时变成对其的僭越？对于一个成为某系统之对象、却未曾同意该系统的人，亏欠着什么？本文论证：整合正义框架对这些问题给出确定的、有时令人不安的回答，这些回答显著地约束了该计划；并且，一个无法处理缺席他者情形的框架，对于寻常情形也同样是不充分的。

本文据此围绕三个问题来组织。第一个关乎规范性的根基：关系性的繁荣，在亲密伴侣关系中具体地要求被接纳进该关系的技术什么，又有哪些规范传统有能力厘定这些要求？第二个关乎整合：证据主义的、实证主义与反实证主义的、义务论的、关怀伦理的、儒道的、以及能力理论的诸般考量，如何能被整合进一种单一而融贯的实践，而非作为彼此竞争且不可通约的诉求被搁置？第三个关乎缺席他者这一极限情形：本框架对于“为一个不知晓、亦未同意该设计的亲密他者而设计”意味着什么，而那一极限情形又揭示了寻常情形的什么？

本文的贡献相应地是哲学性的而非技术性的，可陈述如下：(i)将关系性存在论题限定并具体化于亲密伴侣关系这一领域（§2）；(ii)一套面向亲密数据伦理的整合正义框架，其中每一构成性传统都被赋予明确的管辖权与明确的界限（§3–§4）；(iii)将辩证实证主义实践的伦理阐发为该框架的实践归结（§5）；(iv)一份展示该框架在真实系统中得到落实的实做个案研究（§6）；(v)指认亲密计算所特有的若干不义，包括代理认识不正义（一种认识之举的错误归因）以及结构性效率与关系性发展之间的对立（§6.7、§8）；(vi)将缺席他者问题处理为对一般关系伦理的一次检验（§7）；以及(vii)一组延伸至亲密伴侣关系之外的意涵，关乎社会层面数据利用的伦理，以及连接家庭与公共机构之机制的设计（§10）。软件工程方面的问题，如架构、存储与框架选型，仅在其关涉规范性论证时才被论及，并被限定在一个简短的附录中；它们不是本文的主题。

其余部分如下推进。§2 厘定亲密伴侣关系中的关系性存在并安置其思想谱系。§3 列出各构成性规范传统，每节一种。§4 论证它们的整合。§5 展开辩证实证主义姿态。§6 呈现个案研究。§7 处理缺席的他者。§8 陈述限制。§9 将本文置于周边文献之中定位，§10 作结。

亲密伴侣关系中的关系性存在

思想谱系

本节所展开的本体论承诺，在后现象学上承自魏贝克（）的谱系——他将人造物读解为对其周遭实践的中介而非仅仅容纳；在关系性上则承自布伯（，对他而言我–你关系在本体论上居于首位）、安乐哲与杜维明（、，对他们而言自我是在其角色与关系中被构成的），以及博格曼（，他提出焦点实践这一概念，其价值是内在的、且会被工具化所侵蚀）。本文所添加的，是把这些宏大的本体论主张限定于亲密伴侣关系这一特定领域，并论证这一限定改变了该本体论之所要求。对本文在亲密技术、自我量化与自传式设计诸文献中的更完整定位，留待 §9，即在它所定位的框架被展开之后。

本体论地基：关系性与生成性临在

本框架立足于一种关系性本体论，作者已在别处对此有详尽展开（），它与一个广阔的传统相汇合：布伯关于我–你关系之首要性的论述（）、格根的关系性存在（）、怀特海关于关系性“经验机缘”的过程形而上学（）、海德格尔的共在（）、佛教的缘起与性空之说（），以及儒家角色伦理（）。其共同的主张是：存在的基本单元不是实体而是关系——自我不是一个先行自存、随后才进入关系的原子，而是一种唯有当它进入某种关系结构时才变得可被知觉的效应。一个有用的形式化是代数式的：一个群被定义，不是凭其元素的内在本性，而是凭关联它们的运算，因而在被实例化之前，存在的只是潜在关系之场，而非一组自存的实体（）。依此观之，居于首位的是关系，而非关系项。

这一本体论带有一个本文余下部分所依赖的后果。若自我是关系性地被构成的，那么没有任何符号结构、没有任何记录、没有任何表征，能够与它所折射的关系性存在相重合；每一者都只是部分地、且以一个它无法把持的剩余为代价，捕捉到那一关系。作者的先前工作把这一结构性洞见的、活生生的对应面命名为生成性临在：关系性存在借以在经验中变得活跃的那种程序性的、现象学的样态，即存在通过实践、注意与回应而非通过表征得以显现的方式（）。临在之对反于表征，正如践行之对反于抽象：表征作抽象，临在作践行。这一配对是辩证的，因为没有临在的关系坍缩为抽象的结构，而没有关系的临在坍缩为唯我的当下。对一项亲密技术而言，这一区分是奠基性的。一个只在表征层面运作的系统——记录、仪表盘、摘要——是在符号层面运作的，因而必然错过关系实际所栖居的生成性临在；设计的问题于是变成：一个系统如何能为临在腾出空间，而非以愈发精细的表征去替代它。

同一本体论亦给出一种关于人工对话智能体是什么、不是什么的精确论述，§6.8 与稍后的个案研究都将依凭于此。人类主体与集体主体扎根于一种前符号的关系性实在；一个 AI 语言层则不然。它的运作被局限于符号层面，由对先前符号的条件概率所生成，因而它展现出作者先前工作所称的一种伪临在：它在符号之中显现得仿佛是一个主体，却缺乏真正临在所要求的、在关系性存在中的本体论根基（）。这既非对此类系统的赞誉，亦非贬斥，而是对它们的一种安置：亲密情境中的 AI 可以中介、可以提示、可以见证，但它本身不能成为那个关系性的他者；而一个忘记了这一点的设计，会在临在才是被渴求之物的地方，上演一出临在的赝品。

论题及其限定

关系性论题，在其充分的一般性上，遍及一切关系。本文将它限定于亲密伴侣关系这一领域，并追问：在该领域中，它具体地要求什么。依关系性观点，一段亲密伴侣关系中伦理关切的相关单元，是关系本身——它有自己的实在与自己的轨迹——而非两个个体及他们之间的契约。

这一限定是有后果的。许多对于广义家庭、职场或政治体中的关系性存在为真的事情，并不能转移到亲密伴侣关系上，后者至少由三个特征所区别。第一，它是非角色界定的：与亲子或雇佣关系不同，它并不由一种先在的权威分配所结构化，因而任何一方都不能援引某种角色来为单方面地代另一方决定辩护。第二，它由自愿的脆弱性所构成：每一方都披露其本无义务披露之物，而关系的深度正是这种不被强迫之披露的函数。第三，它构成性地依赖于未言：维系这关系的，有一部分恰恰是那些被留作不予测量、不予记录、不予言说之物，因而穷尽式的言明非但不会丰富它，反会耗竭它。这第三个特征，正是那个一般本体论要点（没有任何表征与关系性存在相重合）在亲密关系上的具体实例：符号化所无法触及的那个剩余，在亲密关系这一情形中，是这一纽带的构成性要素。

亲密伴侣关系对中介的脆弱性

这三个特征中的每一个，都直接易受错误的技术中介之害。非角色界定，受任何悄然把决策权僭归一方的系统所威胁。自愿的脆弱性，受任何把披露从馈赠转化为度量指标的系统所威胁。构成性的未言，受任何以记录、浮现与摘要为默认的系统所威胁。作者先前工作所提出的一般诊断在此尤为有力地适用：随着符号中介的增殖，符号与关系性实在界之间的裂隙不断扩大，而我–你关系侵蚀为一种经由不以主体身份回应的界面所进行的我–它关系（）。因此，一项被接纳进亲密伴侣关系的技术，不能仅凭它是否高效地履行其功能来评价；它必须凭它对那三个使关系成其所是的特征做了什么、以及凭它是扩大还是收窄了符号中介与关系性临在之间的裂隙来评价。这正是 §6 的个案研究据以被衡量的判准。

一套整合正义框架

本节列出本文用以处理亲密数据之伦理的诸规范传统。对每一传统的论述都遵循一个固定的格式：对该传统之核心主张与主要源流的论述，展开至足以公允对待该传统的篇幅；该传统在亲密数据问题上有能力回答的问题；它无能力回答、必须让渡给另一传统的问题；以及，在有用之处，预示该传统如何关涉 §6 的个案研究。诸传统以一种本身也经过论证的次序呈现：先是一个方法论的统摄框架，即辩证；然后是治理“何者可被知晓与声称”的认识论诸传统；接着是治理“何者可被去做”的实质性伦理诸传统；最后是治理“价值归于何处”的政治经济学传统。关于其结果是一种整合而非折衷堆砌的主张，在 §4 中得到辩护。

有一条贯穿整体的定向性承诺值得在开篇即予陈明，因为它塑造着下文每一传统被读解的方式。系统的正面目的，在其所施加的一切约束与拒绝之下，是幸福论意义上的（即 §3.11 所展开的意义）：不是对状态的优化，而是提供一个生成性场域，一个让伴侣们共同的自我实现得以凭其自身的运动而舒展的空间。系统诸模块所共有的设计哲学，是供给这样一个场域、随后即退场，而非驱使关系朝向任何被测量的结果；框架的诸约束标示出该场域的边界，而繁荣的判准则裁断该场域是否确实是生成性的。这一承诺把下文中发展心理学的、幸福论的与能力理论的诸传统联结为一种单一的正面定向，约束之消极工作正应据此被理解。

辩证：整体的方法

第一个被引入的传统不是诸传统中的一员，而是它们借以被持守在一起的形式。本框架的核心难题——某些亲密现象适于测量，另一些则被测量所损害——是一个矛盾，而处理矛盾的方式本身就是一项哲学抉择。本文追随黑格尔（）把矛盾视为认真对待、而非急于消解之物；但它在关键处追随阿多诺的否定的辩证法（）：拒绝那种把张力收摄进一个更高统一、从而抹去其中真实损失的综合。辩证在此的管辖，是整体的方法：它要求测量与克制之间的矛盾被持守为敞开、并在每一场域上被重新裁决，而非被某个公式一劳永逸地解决。它的界限在于，它是一种形式而非内容：它规定问题如何被持守在一起，却独自不裁断任何具体个案；个案的裁断有待下文诸实质传统。

证据主义：有据断言的界限

证据主义在认识论中主张：一个人持有一个信念的正当性，严格地与支持它的证据相称（）。移用于一个会作断言的系统，它给出一条精确的设计约束：系统不得声称超出其证据所担保者，估计须以估计的样态表达，虚假的精确须被拒绝。它的管辖，是认识的模态——系统可以断言什么、以何种把握度断言。它的界限在于，它对“究竟何者应被测量”、以及“谁算作可信的认识来源”保持沉默；这些被让渡给伦理学与认识不正义。

认识不正义：可信度与诠释的公正

弗里克（）区分了两种认识不正义：证言不正义，即由于偏见而给予某人的话语以亏缺的可信度；与诠释不正义，即某一群体的经验由于缺乏共享的诠释资源而无法被理解。移用于亲密系统，证言不正义警告我们：一个把传感器读数置于伴侣自述之上的系统，是在把一个人贬黜为关于其自身的认识者。诠释不正义则警告：一个系统的词汇——其类别、其标签、其可量化之物的菜单——可能贫化用户借以理解自身生活的诠释资源。因此，认识不正义传统的管辖，是一个证据主义与实质伦理传统都无法以自身术语提出的问题：系统在其收集与权衡证据的结构本身之中，是否公正地把每一方都当作一个认识者——既是关于共享世界的认识者，也是关于其自身的认识者。它给出一条正面的设计律令——一个人的第一人称证言须对机器的推断保有效力；以及一条反面的——系统不得收窄其用户借以理解自身生活的诠释资源。个案研究将指认出第三种变体，它有别于弗里克原初分类法中的两种：系统代一方完成某种认识或记忆之举，其方式使另一方被引导而错误地归因其来源（§6.7）；这一代理认识不正义，本文如此称呼它，是一种独属于技术的不义，值得标记出来。该传统的界限在于，它诊断此类不义，却独自不确定那些一旦被违犯即构成最严重之恶的不可逾越的边界；为那些边界我们需要义务论，而为诠释的宽厚所要求的那种情感质地的论述，我们需要关怀伦理。二者皆在下文。

实证主义与反实证主义：可测量性问题

人文科学哲学中长期存在一场争论：人类生活的诸现象是否能像自然科学的对象那样被测量与操作化。实证主义的一脉肯定之；反实证主义的一脉——经由狄尔泰对理解的强调、韦伯的诠释社会学、格尔茨的深描（）——则坚持，意义性的人类现象唯有通过诠释才能被把握，而把它们强行纳入度量会使之变形。本框架不站定其中一方，而是把这场对立移用为一个先行的测量问题：对任一给定现象，须先问它是否容许无失真的测量，抑或它部分地由不被测量所构成。它的界限在于，它框定了测量与克制之间的抉择，却不提供决策程序；决策被让渡给那汲取一切传统的辩证姿态。

结构主义精神分析：欲望、他者与不可符号化者

经由索绪尔（）、拉康（）与齐泽克（），结构主义精神分析提供了一种关于主体如何在语言与欲望中被构成的论述。其与本框架最相关的一点是：主体之中总有一个抵抗符号化的剩余——拉康所称的实在界——而欲望恰恰围绕这一不可被完全言明者而组织。一个亲密系统若追求对伴侣的完全透明、完全记录、完全建模，便是在追逐一种幻想：以为那个他者可被穷尽地符号化。本传统的管辖，是主体的构成：它禁止完全透明的幻想，并守护那个使他者成为一个真正的他者、而非一个已解之题的不可符号化的剩余。它的界限在于，它是诊断性与告诫性的，而非规定性的；它把正面的规则与关怀的质地让渡给义务论与关怀伦理。

义务论：不可逾越的约束

义务论，尤其在康德的脉络中（），坚持某些对待人的方式无论后果如何都是被禁止的：人须始终被当作目的，而绝不仅仅当作手段。移用于亲密系统，它确定了不可逾越的约束：不得有隐秘的监视，不得把一方的披露挪作他用，不得有被设计出来的（伪造的）同意。本传统的管辖，是那些不可逾越的约束——无论可得何种利益，何者都不可对另一方为之。它的界限在于，它是抽象的：它画出可允许之域的外墙，却对如何栖居于墙内的空间保持沉默。它把那种质地让渡给关怀伦理。

女性主义关怀伦理：回应的质地

关怀伦理把回应的质地置于伦理论述的中心，反对一种仅以规则和普遍原则为核心的道德观（）。关怀伦理向本框架贡献的，是一切基于规则的传统所无法触及者：系统在关系中临在的方式、时机、语气与调谐。一则通知可以满足每一条义务论约束——不欺骗任何人、不操纵任何人、不监视任何人——却仍然是一种在亲密生活中最要紧的那类道德失败：在本该温暖处冰冷，在本该温柔处机械，在本该缄默处出场，在本该等待处催促。关怀伦理也重新构想了系统的目的本身。一项有关怀的技术，其衡量标准不是它所交付的效率，亦不是它所产生的愉悦状态，而是它是否维系并深化关系对那个特定的、不可替代的、脆弱的他者——那个无法被任何具有相同度量指标的他人所替代的人——的回应。特隆托所说的“注意”这一关怀阶段有一个直接的设计对应物：一个有关怀的系统首先必须是那种会注意、并且注意到正确之事的东西，而那些正确之事往往不是最易测量的东西。

此处可以提及一个来自通俗关系心理学的启发式，因其实践上的提示性，但须对其学术地位附以明确的告诫。查普曼关于“五种爱的语言”的说法，认为人们各自习惯于通过肯定的言辞、精心的相处时光、礼物、服务的行动或身体的接触来表达与接收关怀，它为设计者提供了一套可用的词汇，用以描述一个有关怀的系统可借以调谐于某一特定伴侣的诸渠道（）。须告诫的是，这是一部通俗心理学而非学术心理学的作品，其强主张——尤其是关于离散的诸“语言”的类型学，及其在发展与临床上的断言——的经验支持是薄弱的；此处引用它，仅作为关于关怀之质地的设计启发，而非作为一种已被验证的理论，本框架不依赖于它。在这些界限之内，它是有提示性的：它提醒设计者，调谐是多元的，同一姿态对一方可登记为关怀、对另一方却可能是噪声，因而一个有关怀的系统应为多种表达的音域腾出空间，而非假定唯一一种。

关怀伦理带有一个本框架必须指明的、著名的内部危险，因为指明它正是整合之论证的一部分。一种纯然回应的伦理，若不受正义与权利的制衡，会滑向偏私、边界的侵蚀，以及关怀者消融于被关怀者之需求之中；无正义的关怀会令人窒息。在亲密数据这一情境中，这一危险有一个精确的形态：那种过度的注意，本只想临在与回应，却凝结为对所爱之人的焦虑监控，正是 §7 缺席他者分析所指认的那个恶。一种不知义务论边界的关怀伦理，甚至可能正面地把这种监控认可为一种忠贞的注意。这正是为何关怀需要义务论，正如义务论需要关怀：一者供给那不可逾越的墙，另一者供给墙内所建之物的温暖，而二者缺一即变形。这种相互的需要不是任一理论的弱点，而是奠定它们之整合的结构性事实。

儒家角色伦理与道家不干预：克制的价值

东亚诸源流贡献了一种在西方正典中大体阙如的论述：无为之正面价值。儒家角色伦理（）把自我理解为在其关系与角色中被构成，并把礼——得体、礼节、对关系之质地的体贴照拂——视为道德生活的核心；而道家的无为，即不强求的行动，把“不去做”确立为一种有时比干预更高的智慧。反者道之动：道之运动是返还。移用于设计，这一传统的管辖，是“无为”之问：何时正确的设计抉择是什么也不建造——让一项能力不被实现、一个数据不被记录、一则提示不被发送，作为一种被付诸实行的尊重。它的界限在于，它独自并不提供现代平等伴侣关系也需要的那套个体权利与可撤回之同意的词汇；它补充而非取代正义诸传统。

能力进路：繁荣的判准

由森与纳斯鲍姆发展的能力进路（），以人们实在地有自由去成为与去做其有理由珍视之事——他们的诸“能力”——来衡量何者是好的，而非以资源或主观状态来衡量。移用于本框架，它供给那个判准：整体是否扩展或收缩了伴侣们去繁荣、去成为他们有理由想成为之人的实在自由，无论是共同地还是各自地。能力进路供给判准，却因其自身的结构而不供给约束。由于它在形式上大体是后果论的，对净能力的评估原则上可能为了有价值之自由的总体扩展而许可对某一个体之权利的侵犯，亦可能容许一种人之尊严或关怀之质地所禁止的干预。因此它依赖义务论来确定追求能力扩展所须在其内进行的边界，依赖关怀伦理来厘定追求的方式，并依赖东亚诸源流来提醒它：某些能力唯有在不被直接施工建造时才繁荣。判准告诉我们该瞄准什么；它并不告诉我们在瞄准之中可以做什么。这一依赖，再一次，不是缺陷，而是指向整合的一个路标。

发展心理学：共同注意与心智理论

§2.2 的关系性本体论提出一个有力的主张：意义与自我性是在诸主体之间被共同构成的，而非在其内部生成的。这一主张是哲学性的，但并非没有经验的佐证，而关于共享意向性的发展心理学提供了这一佐证。依托马塞洛颇具影响的论述，人类认知是独具特色的文化认知，由一种在婴儿期即浮现、而其他灵长类大体所缺的共享意向性能力所使能（）。其奠基性的机制是共同注意：两个主体共同注意同一对象，而各自意识到对方也在注意——这是一种三元结构（自我、他者、世界），共享意义即由之建构。其互补的能力是心智理论，即把心理状态归属于他者、从而把他者建模为一个经验中心而非一具移动的身体的能力（）。这两种能力合起来，是关系性存在的发展性基底：它们正是两个人类主体实际上如何共享一个世界的方式。本传统在框架内的管辖，是把原本抽象的共在（§6.8）概念奠基于一个真实的、被研究过的、具体的机制。当框架要求一个系统去承载伴侣们的共享注意、而非分别吸走各自的注意时，它所要求的正是发展意义上对共同注意的技术性培育，而二者的设计差异是鲜明的：一个分别攫取每一方目光的信息流是共同注意的否定，而一个伴侣们共同注意、且各自意识到对方在注意的共享对象，则是它的实现。心智理论则以另一种约束进入。一个自称能为一方建模其心智、并向另一方报告其模型的系统，并不是在支持心智理论，而是在取代它，以一台机器的表征替代一个人去想象另一个人的那种不可化约的工作；框架关于 AI 之为伪临在的论述（§2.2）在此获得了经验性的牙齿，因为读心这一关系性工作恰恰是那个伪主体所不能做、也不可佯装去做的工作。发展心理学的界限在于，它描述共享意义的诸机制，却不规定其正当的使用；它告诉我们共同注意是什么、它要紧，但把“可以用它正当地做什么”这一问题让渡给伦理诸传统。

幸福论：作为生成性动力学的繁荣

能力进路（§3.9）供给一个以实在自由表述的繁荣判准；幸福论传统则供给互补的论述，把繁荣理解为一种自我实现的动力学，而正是这一传统奠定了本节开篇所宣告的生成性场域承诺。其亚里士多德式的根，认为人之善不是一个有待抵达的状态，而是一种活动，即一个人在整个一生中对其特征性能力的能动施展与实现。当代心理学已给这一理念以可操作的内容：里夫的心理福祉模型把诸如自主、个人成长、目的、与他人的积极关系等维度，指认为幸福论的、有别于仅仅享乐论的福祉之构成要素（）。幸福论之区别于享乐论，对设计而言是决定性的：一个为享乐论而优化的系统产出令人愉快的状态，而一个回应幸福论的系统提供使成长、目的与关系得以被能动地实现的诸条件，而这些条件往往涉及困难、新奇与不被测量者，而非舒适。这给出框架的核心正面承诺，有别于其约束的消极工作。系统诸模块所共有的设计哲学，是供给一个生成性场域：一个由框架的诸约束所标界的、有界的空间，伴侣们共同的自我实现得以在其中舒展，而不被驱向任何被测量的结果。幸福论传统供给“为何这是正确的目的”的理由——即繁荣是一种自我实现的活动，它无法从外部被产出，而只能被供给；以及其设计上的对应物——即系统的任务是供给场域并退场，以道家无为（§3.8）的方式，以见证者而非顾问（§6）的方式。它与能力进路的关系是互补的：能力进路陈述衡量一个生成性场域的判准——它是否扩展了去存在与去做的实在自由——而幸福论则陈述该场域意在承载的动力学——那种实在自由借以被施展的能动实现。幸福论传统独自的界限在于，它厘定了正面的目的，却不供给那使“对繁荣的追求”不致凌驾于人之上的不可逾越的约束；为那些约束，它如能力进路一样，依赖于义务论与关怀伦理诸传统。

生成性正义：价值的安歇之所

最后一个传统处理一个其余传统都未能触及的问题：不是何者可被知晓、测量或去做，而是关系之内（以及中介它的系统之内）所生成的价值被允许安歇于何处。埃格拉什的生成性正义（），借鉴并超越马克思关于异化的论述（），主张：当一个劳动或一个关系所生成的价值循环返回其生成者、而非被抽取去使第三方致富时，正义便得到了更好的伺奉。移用于亲密系统，它问：关系所生成的价值——包括作为其历史之沉积物的数据——是否留存于关系之内。本传统的管辖，是价值的政治经济学；它的界限在于，它治理价值安歇于何处，却不治理价值如何生成；它预设了框架的其余部分，来治理何者最初被允许被测量或被去做。

数据伦理：脉络完整性与共同生活的沉积

迄今所汇集的诸传统在许多点上触及数据之伦理，但当代数据伦理领域已发展出一些足够具体、值得专门处理的概念。其中首要的是尼森鲍姆的脉络完整性原则，它主张：个人信息的流动受其所生发之脉络所特有的规范支配，而一项隐私侵犯，在于信息越过脉络边界的不当移动，而非披露本身（）。这一原则消解了一个长期困扰隐私论述的混淆：问题从不只是信息是否被分享，而是它是否依照生成它的脉络之规范而流动；恰当地与医生分享的医疗信息，当被卖给保险公司时即被侵犯——不是因为它移动了，而是因为它逆着医疗关系的信息规范而移动了。移用于亲密伴侣关系，脉络完整性给出一个确定而严格的标准：亲密关系是一个有其自身信息规范的脉络，而这些规范异常地严格——在其内被披露之物，是在“它留存于其内”的默契理解上被披露的，既不向第三方外流，也常常不在伴侣之间以披露者所未意图的形式（如聚合、评分或永久留存）继续流动。本传统的管辖，是这些恰当流动的规范，连同该领域所阐发的诸同源原则：数据最小化、目的限定、被遗忘权。它的界限在于，它是一种区域性伦理：它精确地治理信息流，却不供给“为何亲密脉络具有它所具有的规范”这一更深的论述，亦不供给据以裁断整体的繁荣判准；它分别从关系性本体论与能力进路领受这二者。

承认与他者之面容：列维纳斯与霍耐特

最后一对源流深化了关于“他者被亏欠什么”的论述。列维纳斯把伦理的起源定位于面对面的相遇，在其中他者的临在把自我召唤进一种先于、且超出任何契约或计算的责任；他者不是一个有待被理解的对象，而是一个溢出我所能形成的任何表征的无限（）。这是那个已被陈述的本体论要点的伦理对应面：正如没有符号与关系性存在相重合，没有对伴侣的表征——无论多么数据丰富——能捕捉那个人，而以表征替代那个人的企图，不仅是一个认识上的错误，更是一种伦理上的失败，是对面容所命令之责任的一次拒绝。霍耐特的承认理论提供了发展性与政治性的补充：人们通过在爱、权利与尊重诸领域中被他者承认而成为自身，而误认是一种真实的伤害，而非仅仅一种坏心情（）。一项亲密技术，无论为善为恶，都参与承认的工作；一个只向伴侣映回它所能测量者的系统，承认的是他们的一个被削减的版本，并可能随时间教会他们在自身之中也只承认那个被削减的版本。本对组的管辖，是系统所给予并使能的承认之质量：它是否把他者尊为一个溢出其表征的无限（列维纳斯），以及它是否支持而非扭曲伴侣们借以成为自身的相互承认（霍耐特）。它的界限与关怀伦理共有，二者紧密结盟：它厘定对他者所应持的定向，却独自不固定那不可逾越的规则或价值的政治经济学。它把框架的伦理层联结回其本体论地基，并为个案研究中所展开的共在与共创之论述（在其中，承认成为伴侣们共同去做之事，而非系统对他们所做之事）作了准备。

整合的论证

把这些传统汇集起来，引出一个问题：其结果是一个框架，还是一份自助餐？它们之间的冲突，例如关怀与正义之间那场长存的张力，又如何被解决？此处所提出的整合不是一种聚合——不是每一传统贡献一张加权的票、再由某个主函数返回一个裁决。跨越不可通约之规范类别的加权聚合，恰恰是被移植进伦理学的实证主义错误，并且因同样的理由而失败：它预设了一个共同的尺度，而那里并没有。这一整合也不是黑格尔强意义上的综合——不是一个张力在其中被消解的更高统一；出于辩证一节所给的理由（§3.1），本框架追随阿多诺，持守诸矛盾为敞开，而非废除它们。这一整合毋宁是一种管辖性劳动的分工，松散地仿照一部良序宪法把不同的问题分配给不同的有权机关、而非把它们坍缩进单一的主权者，而它们全都立足于 §2.2 的关系性本体论这一共同地基之上。每一传统被分配以前文论证它有能力处理的那一类问题，并被剥夺对其余问题的权威。表 1 完整列出由此而来的劳动分工，对每一传统陈明它所持有的管辖、它有能力回答的核心问题，以及它必须把权威让渡给另一传统的界限。

表 1. 构成性诸传统之间的管辖性劳动分工。每一传统对一类独特的问题持有权威，并把超出其能力的问题让渡给有能力回答它们的传统。

传统	管辖与核心问题	界限，及其所让渡者
关系性本体论（Huang 2025；Gergen 2009；Whitehead 1978；Buber 1937）	地基。 关系而非个体，才是关切的基本单元；没有任何表征与它所折射的关系性存在相重合。	是地基而非规则；它确立其余传统所论为何物，却独自不发出确定的许可或禁止。
辩证（Hegel 1807；Adorno 1966）	整体的方法。 要求测量与克制之间的矛盾被持守为敞开、并在每一场域重新裁决，而非被公式一次解决。	是形式而非内容；它规定问题如何被持守在一起，却独自不裁断任何具体个案。
证据主义（Clifford 1877；Feldman and Conee 1985）	认识的模态。 系统可断言什么、以何把握度：要求估计以估计表达、虚假精确被拒。	对”究竟何者应被测量”、以及”谁算可信来源”保持沉默；将之让渡给伦理与认识不正义。
认识不正义（Fricker 2007）	可信度与诠释的公正。 每一方是否被公正地当作关于世界与关于自身的认识者，系统的词汇是否贫化其自我理解。	诊断一种不义却不固定那些一旦违犯即最严重的不可逾越边界；将之让渡给义务论。
实证主义与反实证主义（Dilthey 1989；Weber 1978；Geertz 1973）	先行的测量问题。 某一现象是否容许无失真的测量，抑或部分地由不被测量所构成。	框定测量与克制之抉择，却不提供决策程序；将决策让渡给汲取一切传统的辩证姿态。
结构主义精神分析（Saussure 1916；Lacan 2006；Lacan 1977；Žižek 1989）	主体的构成。 它禁止完全透明的幻想，守护那使他者成为真正他者、而非已解之题的不可符号化剩余。	诊断与告诫性，而非规定性；把正面规则与关怀质地让渡给义务论与关怀伦理。
义务论（Kant 1785）	不可逾越的约束。 无论可得何利，何者不可对他者为之：无隐秘监视、无对披露的挪用、无被设计的同意。	抽象；它画出可允许之域的外墙，却对如何栖居墙内保持沉默。把质地让渡给关怀伦理。
关怀伦理（Gilligan 1982；Noddings 1984；Held 2006；Tronto 1993）	质地。 系统临在的方式、时机、语气与调谐：当温暖处温暖，当缄默处缄默。	无正义则可能纵容偏私与边界侵犯；把不可逾越的界限让渡给义务论，二者相互依存。
承认与面容的伦理（Levinas 1969；Honneth 1995）	他者所应得者。 伴侣是否被尊为溢出一切表征的无限（列维纳斯），相互承认是否被支持而非扭曲（霍耐特）。	厘定对他者所应持的定向，却不固定不可逾越的规则或价值的政治经济学；与关怀伦理结盟并依赖同样的补充。
儒道之克制（Ames 2011；Tu 1985）	无为之问。 何时正确的设计是什么也不建造：让一项能力不被实现、一个数据不被记录、一则提示不被发送，作为被付诸实行的尊重。	独自不提供现代平等伴侣关系所需的个体权利与可撤回同意的词汇；补充而非取代正义诸传统。
能力进路（Sen 1999；Nussbaum 2000）	判准。 整体是否扩展或收缩了伴侣们去成为其有理由想成为之人的实在自由，无论共同地或各自地。	大体后果论；原则上可能为净能力扩展而许可一次权利侵犯。把约束让渡给义务论、质地让渡给关怀。
发展心理学（Tomasello 1999；Premack and Woodruff 1978）	共享意义的机制。 共同注意与心智理论，作为两个主体借以共享一个世界的发展性基底，为共在奠定经验基础。	描述共享意义的机制却不规定其正当使用；把正当使用之问让渡给伦理诸传统。
幸福论（Ryff and Keyes 1995）	正面的目的。 繁荣作为自我实现的动力学，供给系统应提供生成性场域、而非优化状态的理由。	厘定正面目的却无不可逾越的约束；依赖义务论与关怀，以免对繁荣的追求凌驾于人。
生成性正义（Eglash 2016；Marx 1867）	价值的政治经济学。 关系所生成的价值（含作为其历史沉积的数据）是否留存于关系之内，而非被抽取以肥第三方。	治理价值安歇于何处，却不治理价值如何生成；预设框架其余部分来治理最初何者被允许测量或去做。
数据伦理（Nissenbaum 2010）	信息的流动。 数据是否依生成它的亲密脉络之规范而流动，是否被最小化、目的限定、可被擦除。	区域性伦理；精确治理信息流，却从本体论与能力进路领受”亲密脉络何以有其规范”及繁荣判准。

冲突随后并非通过权衡量值大小而被解决，而是按词典序、依管辖、大致按框架之层叠所蕴含的次序被解决。义务论的诸约束与结构主义精神分析对“被设计之透明”的禁止，居于基底：它们不是可被任何基于关怀、基于能力或基于协调之利益所抵消的量值，而是界定其余一切考量得以在其内运作之空间的边界。在那一空间之内，测量问题被提出，由实证主义–反实证主义之轴所框定、并由 §5 的辩证程序所裁决；在测量为可允许之处，证据主义约束其结果可被如何表达，而认识不正义约束确保任一方都不因此被贬黜为认识者。贯穿整体，儒道的考量作为一种偏向克制的长存推定而立，唯有凭一种“行动比无为更好地伺奉繁荣”的正面显示才可被推翻；能力进路供给裁断那一显示的判准；而生成性正义治理由此产生的价值被允许安歇于何处。辩证不是诸管辖中的一员，而是它们协调的形式：它正是何以这一次序归结为一种永恒重审的程序、而非一张固定的裁决表的理由。

这正是整合区别于折衷之处：诸传统被排序，而该排序本身得到辩护——一传统接一传统地，凭“每一者都依赖其余者来恰恰供给它所缺者”这一证明。证据主义需要它无法提出的测量问题；测量问题需要精神分析关于“某些现象何以抵抗捕捉”的论述；义务论需要关怀所供给的质地；关怀需要义务论所供给的边界；能力判准需要它本会凌驾的那些约束；东亚源流需要它们不提供的权利词汇；生成性正义需要框架其余部分来治理“价值在被治理其安歇于何处之前如何生成”。本框架之所以是整合的，是因为它的诸部分不是被仅仅收集，而是被造得相互承重。我们把由此而来的结构称为一种整合正义，因为它把亲密数据的正义治理处理为要求若干不同规范能力的协调施展，其中无一者是主权者，亦无一者可被移除而不致其余者失去立足之地。

一种辩证实证主义实践的伦理

整合正义框架给出诸结构；它的实践归结，是一种本文称之为辩证实证主义实践的姿态。这一姿态可被陈述为一个对每一场可能的测量或中介都施加的检验。对每一个被提议的功能，须问：所论现象是否容许无失真的测量（实证主义之问，§3.4）？若它部分地由不被测量所构成，那么测量它就不是不准确地描述它，而是以一个不同的现象替代它，于是克制才是正确的回答。若它确可被正当测量，那么进一步的诸约束随之而来：义务论固定何者无论如何不可为之（§3.6）；证据主义约束结果可被如何表达（§3.2）；认识不正义确保任一方不被贬黜（§3.3）；关怀伦理治理其被呈现的方式、时机与语气（§3.7）；而儒道的推定要求，唯有当行动比无为更好地伺奉繁荣时，才去建造（§3.8）。这一检验在每一场域被重新施行，而施行本身——那个权衡、那个裁决——被视为一种一阶的伦理行为，而非一个可被一劳永逸地解决的技术问题。这便是“辩证实证主义”之名所标示者：既不普遍地拥抱测量，亦不普遍地拒绝它，而是逐一现象地裁决，并把裁决本身视为伦理行为。

个案研究：一款家庭与亲密关系应用

为使前文的论证不致脱离实践，本节考察一款可运行的应用，它由作者设计并维护，其约两打模块各自把框架的承诺落实于具体软件之中。这些模块在“实证主义–反实证主义”之轴上并不占据单一一点；它们占据不同的点，每一点都由同一个框架把它的若干管辖施于一个不同的现象而得到辩护。本节逐模块进行，对每一模块陈明它所体现的设计抉择、为该抉择辩护的诸传统，以及它所带有的张力。

持久的设计理据

在各个模块之前，有一件横贯性的人造物值得注意，因为它本身就是框架核心动作的一个实例。一份持久的文件，由系统的对话层查阅、并被提交于版本控制，把每一个设计决定连同产生它的理由一并记录为一条单行的规则。三条有代表性的条目可意译如下。第一条持守：任何通知的语气都须温暖而非官僚，更近于一张情书便条而非一则系统消息。第二条把设计的伦理陈述为一种刻意的迟缓，其理据是：被中介的、从容不迫的互动更亲密，系统应为小小的仪式感而设计，而非为即时的通知。第三条禁止对仪式的强求：温柔的功能须保持可选择性，因为自动提示它们会误读伴侣们的情感状态、并在本无压力处制造压力。这件人造物在形式上是实证主义的——它是一份结构化的、有索引的、机器可读的规则清单；在内容上却是反实证主义的——因为每一条规则都编码着一个不可化约地处于脉络中、且部分地是情感性的理由。它在微缩中，正是那被造成一个可运作之物的辩证实证主义姿态：一份测量其自身诸决定、却不佯称其背后诸理由本身是量值的记录。

共同账本：正当的测量及其实践界限

一本横跨两人、涉及数种货币的开支账本，是正当测量最清楚的情形：花掉的一单位就是花掉的一单位，把它计入并不改变它分毫。该模块因而毫无歉意地占据实证主义的一端，而儒家源流甚至会把这一音域中精确的记账读解为一种礼节，它先行化解了那些本会悄然累积的、微小而未言的不对称。在这一地基上，模块提供了人们对家庭财务所期待的那套装置：收支的分类记录、把一连串交易转化为可读结构的经济学与统计学摘要、关于共享之钱流向何处的可视化，以及一定程度的 AI 辅助分析与建议。框架在此的首要工作，在于它所拒绝占据的那个边际：它不为双方的相对开支评分，不朝向某种“最优”分配去助推，也不产出任何排行榜。是义务论那条“不把伴侣转化为比较评估之对象”的约束、而非数据的任何界限，确定了测量止于何处。该模块的伦理，首先就在于这些省略之中。

但共同账本，恰恰因为它是测量最清楚地正当的模块，也最清楚地暴露了测量的实践界限，而诚实要求这些界限被陈明、而非被掩饰。其中三者尤为突出。第一是开销及其正当性之问。完整的记账是费力的，而实践中没有哪一对伴侣会记录每一笔交易；捕捉的劳动是一项真实的成本，它与一项收益——量化分析所提供的清晰——相权衡，那收益是真实的，却也是有界的。框架不为这一权衡提供公式，而其自身的辩证姿态（§5）恰恰蕴含着并无公式：那笔边际的记录是否值得记录它的摩擦，是伴侣们必须做的判断；一个以分析完整性之名强求完全捕捉的系统，将让实证主义的冲动凌驾于它所强加的、活生生的成本之上。诚实的设计接受部分的数据，并就其所有者进行分析，而不佯称完整。

第二个界限直接随之而来：完整性无法被保证，因而那些分析必须被读作描述一个样本、而非整体。这是证据主义约束（§3.2）在账本上的应用：一个建立在不完整记录上的可视化，绝不可把自己呈现为关系的全部财务真相，而该模块的摘要因此被框定为部分的图景、而非经审计的账目。

第三个界限最有意思，因为它显示出即便这最可测量的领域，也触及那构成性的未言（§2）。某些开支理当在一段亲密关系之内保持隐藏：那精心准备的惊喜，那其价值取决于不被预见的礼物。一本为完全相互可见而设计的账本，会以透明之名摧毁亲密生活正当地所要求的那些小小的隐瞒，因而该模块必须容纳一方可在另一方视野之外记录的私密条目。这一点关联到框架诸开放张力中所讨论的秘密之架构（§8）：即便在财务中，关系之善也不由完全的披露所伺奉，而一个无法保守一个秘密的设计，也托不起一个惊喜。

最后一点更具技术性的告诫，关乎该模块所提供的 AI 能力。系统可以尝试从账单与小票的照片中提取开支，以减少上文所言的开销，并可对所记录的数据提供分析性的建议。诚实的立场是：此类识别的实践可靠性仍有待确立——小票的提取易于出错，而在部分且不完美地被捕捉的数据上算出的建议，继承了已被陈明的一切界限。因此这些功能是作为一种须受用户更正的便利而被提供的，受同一条证据主义要求所支配——系统不得对其数字宣称超过其来源所担保的把握度——而非作为对家庭财务真相的一种可信赖的自动化。

家务管理：无需计分的协调

对一个共享家户的管理，如同账本一样，是一个测量大体正当的领域，而系统通过一种事件驱动的设计来处理它，这设计值得描述，因为它的结构承载着该模块的伦理意图。模块所追踪的对象是若干种事件。第一种是多尺度时钟事件，即处于不同周期上的重复性任务——地面每周清洁一次、桌子每天擦拭一次，各自在其周期届满时浮现。第二种是被动事件，由耗尽所触发——清洁剂用完了、垃圾袋用尽了。第三种是主动预警事件，即对耗尽的预期——清洁剂存量将尽、不久可能用完。第四种是手动录入事件，由任一方记录——一样要买的东西，一件要处理的事务。每一事件都带有一个空间属性，把它定位于家中——厨房、厕所、客厅、卧室或其他——使得家户的小小事务既按地点、也按时间被组织。

因为发起一个事件必须毫不费力、否则它就不会被做，界面把常用事件呈现在一张紧凑的表格中，其单元格由单击触发。一种有代表性的布局，对每一天、每一类事件，记录该事件是否已发生、以及是否已解决；一个绿色圆点标记一件已解决之事，一个对勾标记一件已经发生、正待处理之事。表 2 示例了这一形式。当一个事件发生时，系统通过电邮或短信把它报告给另一方、或双方，使得“家所需为何”的知识被共享，而非系于一个人的警觉。

日期	垃圾积累	纸巾尽	酒精纸尽
今天
06/01		•
05/31
05/30			✓
05/29
05/28
05/27

表 2. 家务模块“每日记录”的一个示意。每一列是一类常用事件，每一行是一天。已解决之事以圆点（•）标记，已发生且待处理之事以对勾（✓）标记；空白单元格表示无记录。此布局为示意，而非所部署的界面。

该模块的正面理据，立足于一个关于人、而非关于家户的事实：人的记忆与注意力是有限的，而把“家所需为何”——清洁剂何时将尽、本周的任务轮到谁、什么必须去买——持守于心，是对那一有限资源的一项真实而反复的课税。通过把这一追踪外化进一个共享的、低摩擦的系统，模块把两位伴侣从一类心智负担中解脱出来；那负担若被留在脑中，往往会不均地分配，并在糟糕的时刻浮现为摩擦。所得到的不只是整洁，而是被归还给关系的注意力：本会用于记忆与提醒的时间与心智空间，得以转而被给予彼此。以框架的术语说，这是能力进路（§3.9）与幸福论承诺（§3.11）在最寻常的音域中运作，系统供给一个场域，使伴侣们在一个微小却真实的意义上，被解放去用于他们共享时光的更有意义的用途。

有两项考量为“以效率、尤其以事件驱动之设计来处理这一领域”给出了正当理由，值得陈明，因为它们显示出这一抉择是有理可循的、而非仅仅便利的。第一项是存在主义的，它解释了为何效率在此是正确的优先项，而在日记或共创模块那里却是错误的优先项。重复性的家务劳动，在其纯然的反复之中，是最不具生成性的活动之一：在擦拭一个台面、或记下一项物资已耗尽这样的纯然重复里，主体几乎无从见证其自身的存在，没有幸福论音域所命名的那种自我实现。日记与共享探索是临在被付诸实行、因而绝不可被优化掉的场域，而物资的反复管理，则是临在大体缺席的场域；对一个纯然的、非生成性之重复的领域，合乎人性的回应恰恰是高效地把它打发掉，从而把它本会消耗的有限注意力，释放给那些存在确实在其中被实现的活动。该模块刻意被限定在这一音域：它只关乎事件的触发与记录，而这些操作就其本性而言鲜少具有关系性、且大体枯燥；正因它们如此，把它们交付给一个机制，对关系而言不是损失，而是一项服务。

第二项考量解释了该机制的特定形态。事件驱动的范式，是作为一种类比、而非一种严格的同构，借自工学与生物学——在那里，事件驱动的架构因其对计算资源的高效利用而受重视：例如一个尖峰神经网络，仅在一个神经元发放时才耗费能量，而非持续耗能；事件驱动的系统一般也只在有事发生时才动作，而非凭持续的轮询。同样的逻辑也适于家户的管理。一个持续监控并提示的系统，会施加正是该模块意在移除的那种持续的、低度的需索；而一个事件驱动的设计，仅在一个时钟届满、一项物资耗尽、或一方记录了某事时才被搅动，镜映其生物与工学类比的高效，并恰如其分地让工具保持沉默，直到确实有值得注意之事。这一类比是启发性的，不被推得更远；它被提出，是为表明该设计的形态、而不只是它的存在，回应着一个理由。

该模块带有一个框架要求被指明的、特征性的危险，而事件驱动的设计正是被塑造来先行阻遏它的。那危险是：把家务量化会滑向计分：一旦任务被记录，去清点谁做得更多便只有一步之遥，而对贡献的清点在一段亲密关系中是有腐蚀性的，它把共享的生活转化为一本债务的账簿，把伴侣转化为一个比较评估的对象——正是义务论约束（§3.6）所禁止、且账本模块已经拒绝的那个恶。该设计的回应是结构性的：模块记录一个事件已发生、且已被解决，但它默认不记录是谁解决的，也不产出任何按人计的总计、任何公平分、任何排行榜。它追踪家的状态，而非其居住者的相对勤勉。一个更深一层的、更微妙的危险在于：对家居生活的全面事件化——每一项任务都是一个被追踪的重复，每一个橱柜都是一份耗尽的预报——有把家从一个栖居之所转变为一个被管理之系统、一个优化目标而非一段共享生活的风险。框架的回应是别处所建言的同一种克制（§3.8）：模块是作为对认知负荷的纾解、而非作为对家户全面行政的工具而被提供的，而伴侣们始终可以自由地让他们共同生活的大部分保持不被事件化，而这往往正是对的事。

爱的日记：原则性的拒绝、温柔的见证与手写的信

由任一方所写的每日日记条目，对双方可读，但默认是缄默的：没有任何通知宣告一则新条目，发现通常凭浏览、而非凭推送。没有情感排名，没有自动摘要，除非被明确调用、否则没有任何 AI 对一则条目内容的评注。这是原则性拒绝的范式。一则经由推送通知送达的日记条目，已从一件被书写之物转化为一件被传输之物；而一则在“它将被评分”的知情之下被写下的条目，是一种不同于在无此知情之下被写下的行为。框架对该条目情感内容的裁决是确定的：它通不过测量检验，因为测量它即替代它；义务论约束禁止隐秘的情感监视；而关怀伦理治理其余者——界面的温暖、压力的缺席、那则条目作为一次不被强迫之披露的地位的保全。

在这些界限之内，该模块提供两项值得考察的功能，因为每一项都显示出框架如何允许一种狭窄的、被审慎界定的 AI 介入，而不越入刚才所命名的诸恶。第一项是一则可选的提醒。一方可以选择让 AI 温和地汲取两人近期的状态，在另一方写了点什么时通知他，其音域是温暖的、而非文书式的。这样一则消息或许写作：

“xx 写了新的日记！似乎十分喜欢你买的花。
信中包含了满满的爱意，快去看看吧。
我们的小家似乎又幸福了一些。”

这一功能凭三条约束而与“缄默之默认”相调和。它是选择性加入的，绝非系统的默认；它是一则事实报告——报告有东西被写下、且其语气是温暖的——而非一个评分、一份摘要、或对该条目内容的分析；而且，最重要地，它不提供任何实质性的关系建议。它会说有一封满含爱意的信在等候；它不会说收信者应当回写什么、应当如何感受、或应当做什么。这条界线与共同日历所划的那条相同（§6.7）：AI 可以温暖地报告一个事实，但它绝不可代行伴侣自身的认识、留意或决定之举，否则便陷入那里所分析的代理认识不正义。在此，见证者的姿态（§6）以可能最温柔的形式被付诸实行：一种只说“有某种充满爱意之物在此等你”、随即退场的临在。

第二项功能是一个装饰性信纸的生成器，而它的设计，再一次，是存在主义的。在一个生成式模型的时代，符号已变得廉价可产：文本、图像、纹饰可被无尽地、无成本地召唤出来，而恰恰是这种廉价抽空了它们曾经承载的、属于一个稀缺而费力之符号的分量。该模块把这种廉价转用于一个刻意的目的。AI 依伴侣当下的感受——此刻心中所有者，欲言者——生成信纸，而伴侣可以自由地一再重新生成，直到某一款式回应了那份感受；选定的那张随后被打印出来，而信，是手写于其上的。重点不在信纸——它恰恰是这个时代所使可能的、廉价的、可无限重生成的符号——而在信纸为何。一再重新生成、直到一张感觉对了的那个行为，是一次感性的施展，一次对“某物回应了一个人之所感”的时刻的等待；而随之而来的手写，是那个生成式符号永不能成为的、不可化约地费力的印记，在其变化与差异之中承载着那只独一无二、亲手制作它的手的踪迹。在手写之中，缓慢地、不可重复地，一个人见证其自身的存在，并以任何被生成的文本都无法付诸实行的方式，留意于其所写之人的存在。在此，AI 担起那廉价的、可重生成的符号层，恰恰是为了反衬、并为那个根本无法被廉价生成的层腾出空间：作为一次与实在界之相遇的手写之信。该设计把手写的机缘交还给人，并随之交还一种日渐稀有的可能：跨越一道唯有他们自己才能留下的印记，去与那个他者相遇。

健康与成长：争议的中段

在那些清楚的情形之间，横亘着辩证为之而设的地带。健康模块从自由文本的餐食描述中估计营养摄入；那估计是真实的，却是近似的，因而界面把它标注为一个估计，并拒绝那种单一自信数字的虚假精确，由证据主义的模态所支配。两个模块都不产出阈值、警报、排名或建议，因为“被测量”这件事对饮食与自我修养之实践的改变，比它对一笔钱款之总和的改变更为强烈：去清点一个人的成长时段，可以反转那一实践据以为内在的理由，正是能力进路判准所诊断的那种腐蚀。这些模块因而测量那稳健的部分，即什么被记录了、何时记录的，而拒绝测量会被变形的那部分，即“一个人是否依某种评分而做得好或坏”这一裁断。它们因此在实质上是诠释主义的，仅在表面上是实证主义的。

对话式 AI 层作为见证者

系统最有后果的设计决定，关乎它的 AI 层，它可读取各模块的数据库、并通过被确认的工具调用而行动。该层被定位为一个见证者，而非一个顾问，而框架把这一区分视为承重的。一个顾问预设了正确建议的存在、以及系统产出它的能力；一个见证者只预设了有某种值得对之临在之物。顾问的失败模式是家长主义；见证者的失败模式是无关紧要，而框架建言接受后一种风险。具体地说，该层不会主动摘要日记，不会为一则条目产出情感评分，并且即便被请求、也被禁止把情感的时间序列加以聚合，从而“觉察一方在一个月里渐趋低落”的能力被保留给那位阅读自己日记的人类读者。这是被付诸实行为一项能力禁止的道家无为：系统的克制，是对一种会在被管理之下凋萎之善的、被付诸实行的尊重。

共同日历：温暖的提醒及其伦理界限

共同日历乍看是诸模块中最平凡的一个，而它的设计之所以有教益，恰恰因为框架在一片看上去一律无害的领地上划出了一条锋利的界线。该模块的独特之处在于，它的提醒——经由电邮或短信送达——是由 AI 层撰写的，而非作为一个寻常日历那种赤裸的、官僚式的通知被发出。AI 温和地汲取近旁的脉络——一则近期的日记心情、一个事件显而易见的分量——以一种温暖而柔情的音域写下提醒。考虑一条写着“5.20，10:00–15:00，买花”的日历条目。这个日期并非随意：在中文的互联网文化中，五月二十日（5.20）是一个非正式的情人节，因为数字五二零听起来像我爱你。一个寻常的日历会发出“提醒：买花，10:00–15:00”。系统则或许会发送更近于这样的一条：“明天就是五月二十号了，是我们小小的、说出口的日子。别忘了在十点到三点之间把花取回来。希望她看到时会眼睛一亮。”这份温暖不是装饰。它是关怀伦理的坚持（§3.7)——系统临在的方式本身即是一桩道德之事，在它最微小、最频繁的诸举动中被付诸实行；一则温暖而非机械的、且认出了这个日期所承载之人性分量的提醒，把这一时机当作它本然的人性时机来对待。

然而，系统还可以再迈出一步，而它刻意不迈，其克制的理由正是伦理上有意思的部分。AI 拥有的数据，本足以让它在提醒之外再行建议：它可以附上“她这周似乎有些低落，或许她最爱的牡丹会是个好选择”，或“在更早的一则日记里她提到喜爱毛茛”。这样的建议会是准确的、有帮助的，并且按一种优化的标准看，显然是好的。框架禁止它们，其理由有二，二者合起来标示出该模块的伦理界限。

第一条理由，是已在克制与关怀之下展开过的那条：选择如何留意一个所爱之人——什么花、什么姿态、留意采取什么形式——是亲密之爱的构成性行动之一，而非一个有待求解的物流问题。AI 去做这个选择，便是介入了关系的决断核心，置换了一个其价值恰恰在于“它是爱者自己的”行为；这正是本文标记为长存风险的那种关系性异化（§8），在微缩中被付诸实行。系统可以提醒一个人某个时机将至；它不可决定一个人应当如何赴这个约。

第二条理由命名了一个框架能够定位、而弗里克原初分类法却未涵盖的恶，本文以一个为便于指称的术语标记它。倘若 AI 供给了那些体贴的细节，那位收到花的伴侣，会遭遇一个看上去像是“被深深知晓”之证据的东西：恰当的花，从一句随口的话里被记起，在恰当的时刻到来。那位伴侣会自然而合理地把这种知晓归于赠予者，而事实上那知晓是机器的。这是一种认识上的恶，却不全然是弗里克那两种形态中的任一种。它既非对一个认识者的贬黜（证言不正义），亦非诠释资源的贫化（诠释不正义）；它是一个认识之举的错误归因，其中一个系统执行了那本应由人来执行、关系之意义要求由人来执行的认知与留意之劳动，并安排诸事，使其产物被接收为那个人自己的。本文把这称为代理认识不正义：一个人的认识、记忆或留意被一个代理所代行、其方式使另一人被引导而错读其来源，这样一种恶。它的受害者，首先是那位被蒙蔽的接收者，他被给予了一幅关于“自己被多好地、被谁知晓”的虚假图景；但它更深的牺牲品是关系本身，因为使“他记得我爱的花”之所以珍贵的，正是他记得，而一个制造出此种记得之表象的系统，掏空了它所模拟的那一善本身。花是真的；它们看似见证的“被知晓”不是。框架的裁决因而是确定的：AI 可以承载那提醒、却不可承载那心思，可以帮助爱者准时行动、却不可告诉爱者去感受或选择什么，因为越过那条线，便是以一件留意之赝品替代真物，并就所受关怀之来源欺骗了所爱之人。

共在与共创

迄今讨论的诸模块，最好通过测量、约束与克制的诸管辖来理解。系统的另一类功能，则最好通过 §2.2 所引入的生成性临在概念来理解，因为它们所培育的，不是对关系的准确表征，而是伴侣们对它的能动的、共享的付诸实行。两个概念组织起这一类。共在是这样一种状态：两个人以关系性本体论所称的临在、而非仅仅交换表征的方式，彼此临在、并临在于一个共享的情境。共创则是共在的生成性延展：伴侣们不消费一个预先给定的内容，而是共同把某物带入存在，并在此过程中付诸实行并深化关系本身。

由此而来一个设计上的后果，即表征关系的功能与承载其生成性临在的功能之间的区分。一个摘要、评分或提醒的模块经营的是表征；一个为伴侣们共同造就某物而敞开一个空间的模块，承载的是临在。后者更难设计，恰恰因为它的价值无法在它所记录之物中被捕捉：一个共享的空间之成功，不在于它的日志何其丰富，而在于伴侣们借助它而比本来更彼此临在。在此，AI 层作为一种伪临在（§2.2）的地位成为一项正面的设计原则、而非仅仅一条告诫。因为 AI 本身不能成为那个关系性的他者，它在一个共创功能中的恰当角色，不是作为第三方去参与，而是去降低伴侣们自身共在的门槛：去提示、去供给材料、去敞开一个空间，随后退场，以无为的方式，使所生成者生成于两人之间、而非生成于他们各自与机器之间。

“一起看世界”：共创式的探索

一个具体的模块例示了这一区分。系统的诸功能之一，是一个暂称为“一起看世界”的功能，伴侣们通过它探索一个世界的诸文化、历史、地方与观念，而这世界从未被穷尽。设计的问题在此变得锋利：这样一个功能既可被建为表征，也可被建为共创，而二者是十分不同的人造物。若被建为表征，它会是一个带进度追踪的、世界文化的结构化目录，一份名义上“已覆盖”之国家的清单，为参与度而优化的推荐，或许还有一个文化广度的评分。这种实证主义的建构，恰恰犯下框架所诊断的错误：它把一次生成性的、开放的相遇转化为一份被测量的、可被完成的清单，并在此过程中以那个被替代的现象取代了原来的现象。

该模块通过一种刻意选择的交互结构来实现这一共创式探索，那结构松散地仿照迭代加深搜索。这一结构在哲学上是要紧的，值得以一定的精确来陈述。从世界的一个被选定的区域出发，AI 层浮现出一小组取自该区域的文化现象——习俗、作品、历史、观念——并提供一个对其显著性的尝试性排序。从这个有界的集合中，任一方都可选择一个现象去展开，于是 AI 转而浮现出又一个有界的、与所选者相关的现象集合；探索如此进行，下潜进一条被选定的线索，而当一条线索被穷尽或失去其牵引力时，便再次拓宽到一条相邻的线索。分支在每一步都被刻意地加以限界：可展开节点的队列被保持得很小，使伴侣们面对的是一份邀请、而非一份令人不知所措的目录。这一程序可被概括如下。

seed ← 世界的一个区域
frontier ← AI 浮现一组有界的现象，附带尝试性的显著度
while 伴侣们仍然好奇:
    node ← 任一方从 frontier 中选择的一个现象
    将 node 记入共享知识图谱，标注由谁选择
    frontier ← AI 浮现一组有界的、与 node 相关的现象
    （沿一条线索深度增加；当一条线索被放下时广度恢复）

这一结构的两个特征承载着设计的意图。第一，选择被允许给任一方，并可以是异步的：一方可以独自地、在一个安静的时辰里展开一个节点，而正在生长的图谱中的每一个节点都被标注以是谁开启了它，使得图谱所记录的，不只是什么被探索了，更是两份不同的好奇心相遇的踪迹；一种同步的模式——两人作为单一用户一同探索——也是可用的。第二，这一过程造就一件人造物：一幅随探索而生长的共享知识图谱。这幅图谱正是 §3.12 那个意义上、关系探索史的沉积，而它的设计据此被治理：它被本地持有，归属于伴侣，且从不为完整性而被评分，因为为图谱评分，便是把发现重新转化为那共创式建构本意要逃离的清单。

该模块的价值恰恰在于它所不测量者——共同注意的质量、一次共享发现的惊喜、两人专属之文化的缓慢累积——而 AI 层，忠实于它的伪临在，凭供给而后退场来服务，绝不代伴侣成为那个共同探索者。这是系统中为共在而设计的最清楚的情形：它的成功，是伴侣们对彼此、对他们所探索之世界的生成性临在，一种几乎不在任何日志中留下踪迹的成功，而正因如此，按框架的眼光，它恰恰是对的。

“我们的旅程”：人工智能的刻意缺席

如果说“一起看世界”显示出 AI 层被缩减为一个自我隐去的材料供给者，那么又一个模块则显示出它被整个移除，而这移除正是设计。该模块，暂称“我们的旅程”，支持对旅行的共同规划与记录，并与日历、相册及幸福储蓄罐相整合。它提供了人们对这样一个工具所期待的装置：用以放置与查看地点的轻量地理功能、对兴趣点的检索、对一段行程的手动排序，以及对每一地点或每一天，记录照片、心情、日记与清单。它显眼地不提供的，是任何人工智能。没有推荐器去建议伴侣们应当去哪里，也没有自动求解器去计算一条最优路线，没有对“以最少的时间或距离访问所选地点”这一旅行商问题的算法解答。这一省略是刻意的，并且在框架之内是有原则的。

其正当理由，是把框架关于克制与无为的论述推至其极限（§3.8）。共同的旅行规划，依此处所持的观点，不是一个其摩擦应被最小化的物流问题；它是一个构成性的关系事件，一个两个人共同面对一个敞开未来的机缘，他们各自协商所盼，并共同地许诺于一条谁都无法事先确知的路。那份不可预测不是一个应被工程掉的缺陷，而正是关系性之善借以被实现的媒介。插入一个推荐器，便是代伴侣们回答了那个“我们该去哪里”、而关系恰恰受益于他们共同回答之的问题；插入一个路线优化器，便是把一场共享的审议转化为一个已解之题，移除了恰恰是活动之价值所在的、对不确定性的共同面对。该模块因而提供一个生成性场域（§3.11）——地图、清单、记录的手段——并从那些工具所支持的审议中彻底退场，是系统所含无为的最严格的付诸实行：在这里，正确的设计是根本不建造任何智能。

这一安排的正面之善，可被诉诸为共创模块奠基的那同一些理论（§6.9），而这一诉诸是有教益的，因为每一种善都会被该模块所省略的那个 AI 所削弱、而非伺奉。自我扩张模型预言，共同地从事新奇而费力的活动会提升关系质量（）；一段共享旅程的规划正是这样一种活动，而一个移除了那份费力的优化器，会连同自我扩张一并移除。共同规划是共享意向性与共同注意的一种持续施展（§3.10），其中伴侣们在建构共同立场的同时建构一个共同的计划（）；一个提出了计划的推荐器，会先行抢占那个本会产生共享感受力的共构本身。而该模块把“对旅程的记忆”置于“旅程的最优性”之上的那一优先——它所记录的照片、心情与条目，高于它所拒绝去计算的效率——表达了那个幸福论的承诺：繁荣在于一段经历的能动的、共享的实现，而非在于一个结果的优化（§3.11）。伴侣们最终所拥有的，不是那个可能最好的旅程，而是他们的旅程，被共同地面对、共同地造就，而这正是该模块之所为。

这一刻意的省略招致一个显而易见的反驳，而诚实要求它被陈明并被回答。这反驳会说：拒绝一切 AI 当然是一种浪漫的矫枉过正：一个推荐器或许能为伴侣们省下一个被浪费的下午，一个路线优化器或许能阻止一个真正令人苦恼的、来回折返的糟糕日子，而舍弃这些，便是在一个理念的祭坛上牺牲真实的善。框架的回答不是否认那些善，而是去定位那条界线，并承认那条界线难以划定。在一件工具可以正当地去减少的物流摩擦（本会记错的营业时间、本会订错的票），与一件工具绝不可先行抢占的关系性审议（去哪里、舍弃什么的选择，对一个不确定计划的共同协商）之间，确有一个真实的区分。该模块一个可辩护的版本，或许会在物流的一侧容纳狭窄的、非建议性的信息工具，同时仍在审议的一侧拒绝推荐与优化。当前的设计保守地划定这条线，完全排除 AI，部分是因为“辅助物流”与“悄然引导审议”之间的边界如此易于被模糊，以致一个干净的拒绝是对该价值更稳妥的付诸实行。这是一个可辩护的、而非可被证明为正确的划线之处，而结构性效率与关系性发展之间那个底层的张力并不容许一个一般的解，这一点已在框架的诸内部张力中被承认（§8）。

“幸福储蓄罐”：作为关系储备的记忆

又一个模块处理关系可持续性的一个不同维度，不是共享意义的生成，而是它对艰难时世的留存。“幸福储蓄罐”邀请伴侣们记下并存起共享喜悦的诸时刻——一段被珍视的回忆，一张照片，几句话——并非为了把它们最优化，而是为了在日后某个更艰难的时候，让它们可被取用。这一模块的理据，借鉴弗雷德里克森关于积极情绪的“扩展与建构”理论（）、布莱恩特与韦罗夫关于品味的研究（）、以及盖布尔关于资本化（分享积极事件）的研究（）：积极情绪扩展一个人的视野并建构起持久的资源，而对它们的品味与分享，会强化它们所留下的纽带。储蓄罐意在成为这样一个记忆的储备，一个可在逆境中被取用的缓冲。

它带有两个框架要求被指明的张力。第一是工具化的风险：把回忆存起来“以备日后取用”，把关系的过去置于一种资源的样态之下，而那个银行的隐喻，无论被赋予多么温柔的用意，都朝向对记忆的工具化滑动，正是框架在别处所抵拒者。第二是一种被策展的积极性的风险：一个由美好时刻汇集而成的储备，倘若它成了共享记忆被认可的样态，便可能悄然边缘化那同样构成一段共享生活的悲伤、艰难与矛盾。该模块的诸约束——它是可选的、不被评分的、且明确是部分的——减轻、却并不消除这些风险，而一个被策展的幸福之储备能否在一段长久关系中被维持、而不滑向任一种变形，是本文留作敞开的问题。更深的一点是：恰恰是那些使储蓄罐作为缓冲而有用的特征——它的刻意性与它的选择性——也正是生成其诸风险的特征，因而这一张力是内在于设计的、而非附带的。

个案研究的教益

诸模块并不坐落于“实证主义–反实证主义”之轴上的单一一点；它们坐落于不同的点，每一点都由同一个框架把它的若干管辖施于一个不同的现象而得到辩护。财务占据可测量的一端；家务模块紧邻其旁，协调着家的小小事务，同时拒绝为其居住者计分；日记占据原则性拒绝的一端；健康与成长占据那争议的中段；共同日历显示出框架在一个看似无害的模块之内、在温暖的提醒与对一个爱者之留意的代理性执行之间，划出一条精细的界线；诸共创功能——其中以“一起看世界”为首——占据一个测量之轴甚至无法捕捉的音域，即生成性临在的音域，那里的设计目标是承载关系自身的付诸实行、而非表征它；旅行规划模块显示出同一音域被推至其极限，那里正确的设计是根本不建造任何智能；而幸福储蓄罐显示出同一音域转向对共享意义的、抵御未来逆境的保存。把这些放在一起，个案研究例示了那个论点：框架不是一条单一的规则，而是一种把若干管辖施于诸不同现象的、有纪律的程序。

缺席的他者

迄今的论证大体进行得仿佛两位伴侣都知晓并参与着这些工具。本个案有一个特征使这一假设落空，而本文把它作为一种哲学工具：该应用是为一段关系而建造的，在其中另一方于写作之时尚不知晓该应用的存在。这一缺席他者的情形，不是一个边缘的怪例，而是一面放大镜：它把每一个关系伦理都必须回答、却通常可凭双方在场而搁置的问题，磨得锋利。

极限情形的严峻

为一个不知情的他者而设计，使若干通常潜伏的伦理问题变得不可回避。一方可以正当地代表一段未经另一方商议的关系决定什么？当一方为另一方之善而准备某物，那准备何时变成对其自主的僭越？对于一个成为某系统之对象、却从未同意该系统的人，亏欠着什么？这些不是修辞性的问题；它们对一个为缺席他者而建造之物的可允许性，施加了真实的约束。义务论约束在此以特别的力量适用：那条“不把一方挪作他用、不施行隐秘监视、不伪造同意”的禁令，在另一方甚至不知系统存在时，变得更为尖锐而非更为宽松。

框架的许可与禁止

框架对这一情形给出确定的、有时令人不安的回答。它禁止：任何对缺席一方的监视，任何把关于他们的数据加以聚合或建模、以预测或引导他们的企图，任何会在日后揭晓时使他们觉得自己一直是一个未经同意之实验的对象的东西。它许可：为一段共享未来作准备——选取礼物、计划一场提议、保留只属于建造者自己的那部分意图——只要那准备约束于建造者自身的、可被自由揭示的领域，而不侵入另一方的内在。这条线大致是：一个人可以为关系准备、可以怀着爱意去预期，但不可代另一方先行决定那些只有他们在场时才能正当回答的问题，也不可建造任何依赖于对他们的隐秘认识的东西。系统的本地优先、其对跨人评分的拒绝、其对“它在涉及双方之事中使用一方私密内在”的默认反对，都是这一立场的付诸实行。

对寻常情形的教益

缺席他者的情形所揭示的，是一个对寻常情形也为真、却在那里较易被忽视的东西：即便当双方都在场、都知情时，一段关系中也总有一个缺席的维度——另一方的内在，那个溢出我所能形成的任何表征的无限（§3.14）。为缺席他者而设计所要求的那种克制——监视无物、预设无同意、只建造关乎自身准备之物，并对“强求那只能被邀请之物”持守一种长存的推定——并非这一极限情形所特有，而正是一切关系伦理在寻常情形中也应持守者。一个无法处理缺席他者的框架，对寻常情形也同样是不充分的；而本框架对前者的处理，恰恰加强了它对后者的论述。

限制

本工作的限制有两类。第一类关乎框架自身内部的张力，即那些框架命名一个问题比解决它更容易的地方，在那里诚实要求持守困难为敞开、而非佯称已将其消解。第二类关乎本文证据的方法论地位。第一类更为严重，被首先处理、且以更大的篇幅。

框架的内部张力

在诸具体张力被列出之前，有一个一般的承认应当框定它们全体，因为它们共有一个单一的根。本框架是作为一种减少技术中介对亲密生活所能造成之损害的纪律而被提出的，而非作为一种消除它们的方法，因为那些损害中最深的，不是任何特定设计的缺陷，而是中介本身所固有的代价。在两个人之间放置一件工具，无论多么审慎，都是把本会直接在他们之间通行之物的一部分，改道经由一件人造物，而这一改道带有两项无设计可完全逃脱的长存负债。第一项是关系性异化：中介在边际上把一段人与人之间的关系，转化为每一方各自与系统进行的关系，即本文通篇所诊断的、从我–你向我–它的漂移。第二项是一族认识的扭曲：§3.3 的证言与诠释不正义、§6.7 的代理认识不正义，以及每当一个系统记录与呈现的方式塑造了它的用户注意到什么、如何理解自身时所引入的、更微妙的知觉偏差。此处所报告的设计在每一点上都致力于把这些负债最小化——通过克制、拒绝、本地所有、见证者姿态，以及在智能会构成侵扰之处对其的刻意缺席。须诚实地说，最小化是所能宣称的至多者：只要一件工具仍在中介这段关系，某种异化与认识扭曲的残余便仍存在，而框架的任务是把那残余持守得如关怀与设计所能做到的那样小，而非佯称它能被归零。下文诸具体张力，都是这一单一的、不可化约的张力的具体化。

辩证裁决本身的困难

框架要求在每一场域上重新裁决测量与克制的矛盾，而它不提供据以如此做的算法。那个判断——这一现象是否部分地由不被测量所构成——在边界情形上是真正困难的，而框架可能在任一方向上出错：错误地拒绝一个本可被正当测量的现象，或错误地测量一个本应被留作不言的现象。辩证姿态是一种尝试那个平衡的纪律，而非一个达成它的程序。

结构性效率与关系性发展的对立

在个案研究所报告的若干设计决定之下，潜藏着一个框架命名却无法消解的张力：结构性效率与关系性发展之间一种长存的对立。效率的逻辑寻求最小化摩擦、优化结果；关系性发展的逻辑则往往要求某些摩擦、某些不可预测、某些不被优化的劳动被保留，因为正是在共同地迎对这些之中，一段关系得以深化。这两种逻辑并不总是对立的，但它们对立得足够频繁、且在足够深的点上，以致没有一般规则能调和它们。把人工智能从旅行规划模块中刻意排除（§6.10）是最锋利的实例：在那里框架舍弃了一项真实的效率——对被浪费之时间与来回折返的避免——以保全一项关系性之善——对一个不确定计划的共同面对——而它如此做时知道这一牺牲是真实的、且明理之人可能会有不同的权衡。同一对立，在框架建言拒绝一项本会（按其自身的标准）有所帮助的测量、推荐或优化的每一处，都再度出现。这一限制是双重的。第一，框架不提供一条原则性的边界，去区分一件工具可以正当减少的物流摩擦、与它绝不可先行抢占的关系性审议；它能在清楚的情形中标记这一区分，却承认那些情形彼此渐次过渡，因而任何特定的划线——包括为旅行模块所划的那条保守的“完全不用 AI”的线——都是可辩护的、而非可被证明为正确的。第二，更深地，这一对立可能是不可化约的：或许就是这样一个事实——一段为效率而优化的关系与一段为发展而培育的关系，过了某一点，就是不同的关系，而没有设计能同时交付二者之善。框架的回应不是去解决这一对立，而是使它可见，从而对效率的每一次拒绝都是一个经过考量的抉择、而非一个疏忽；并坚持：在两种逻辑真正冲突之处，一项亲密技术的设计所应伺奉的，是关系性发展、而非结构性效率。

经由 AI 层的关系性异化

即便有见证者姿态的诸约束，一个对话式 AI 层的引入，仍带有一种把关系性临在异化的风险。一方可能开始更多地与系统、而非与伴侣进行那本属关系性的回应；那温暖的提醒可能开始替代那本会自发涌起的留意。框架以见证者姿态、以本地所有、以对实质性关系建议的拒绝来减轻这一风险，但它无法消除它：只要 AI 在场并以温暖的音域言说，把它误认作一个关系性临在的滑坡便始终在那里。这是那个一般的中介代价，在 AI 层这一特定处的具体化。

共创对个人空间与秘密之架构

框架珍视共创、共享与彼此临在，但一段健康的关系也要求个人的空间、未被分享的内在、以及正当的秘密（如那精心准备的惊喜，§6.2）。一个把共享与共同可见推至其极限的系统，会侵蚀那个空间。框架以“默认本地所有、把秘密之架构建入系统（容纳私密条目、不强求分享）”来回应，但共创与个人空间之间的平衡没有一般的解；一个把一切都共享的系统会令人窒息，而一个什么都不共享的系统则失却共创之善，而那条线落于何处，是伴侣们必须共同协商者，框架只能为那协商腾出空间。

记忆的刻意存储

幸福储蓄罐（§6.11）刻意地存储被策展的记忆，而这带有已被指明的工具化与被策展之积极性的风险。更一般地，任何刻意存储记忆“以备日后”的系统，都把关系的过去置于一种资源的样态之下，而这与“让记忆如其所是地、自发地涌现与消退”之间存在张力。框架不解决这一张力；它只是使储蓄罐成为可选的、不被评分的、明确部分的，从而把风险减小，而不佯称将其消除。

方法论的限制

除这些内部张力之外，本文的证据具有其方法所固有的限制。它建立在单一一个由作者设计的个案之上，因而“设计者即分析者”这一为人熟知的偏倚适用：一个人对自己的摩擦觉察敏锐，对自己所主张的承诺心存偏向，并且没有一个比较性的个案，可据以独立地检验框架的诸裁决；该个案是作为一个存在性证明与示例、而非作为经验证据被提供的。诸传统的词典序排序经过论证，却未被证明，而一个把关怀置于义务论之上、或拒绝能力判准的读者，可能接受那些构成性传统、却拒绝那个整合。把关系性存在限定于亲密伴侣关系，使得它向其他关系（向广义家庭、向友谊、或向以不对称依赖为标志的关怀关系，在那里非角色界定与对称的脆弱性都不成立）的迁移成为一个敞开的问题。而这一立场跨越更长时段的耐久性——孩子的到来、疾病的诸事件、注意力在数年间的漂移；一个如今被安全地计入之现象，是否会在新的境况下迁移到测量之线的另一侧——则未经检验，而这恰恰是辩证姿态所预期、却无法预先解决的那类问题。

相关工作

在框架及其应用现已在目之后，本节把本文置于它所汲取的三类文献之中：技术哲学与技术伦理、自我量化的经验研究，以及自传式系统的设计研究传统。它不穷尽地综述它们；它在每一类中，指认本文所用的资源与它意在填补的缺口。

亲密技术的伦理与哲学

“亲密技术要求一些在生产力情境中并不被担保的克制”这一主张，已在若干名目下被展开。哈尔奈斯与雷德斯特伦（）提出慢技术，作为为反思而设计，把一件人造物之时间性的迟缓视为一项正面属性、而非一个缺陷。魏泽（）更早地阐述了宁静技术，即退至注意之边缘、直到被调用的计算，而罗杰斯（）告诫，宁静的纲领有把退场加以审美化、却不厘定何时介入才恰当的风险。森格斯（）把反思性设计形式化为一种抵抗对未经审视之价值的继承的实践。本文继承这些动作所共有的克制取向，但与它们分道之处在于：它把克制不视为一种审美的默认，而视为一个规范论证的结论，逐场域地、向一个明确的伦理框架（而非向品味）负责。

量化自我、测量及其批评者

自我量化从运动员与病患向日常家居生活的扩张，已被详尽地记录与批评（）。豪厄尔（）报告，实时的情感生物传感会改变它声称要测量的那种情感经验；伯纳（）论证，情感数据是被建构的、而非被测量的。这些发现通常被作为经验性的告诫来呈现。本文把它们重构为一个更强的、部分形而上学之论题的证据：亲密生活的某些现象部分地由不被测量所构成，因而测量它们不是不准确地描述它们，而是以不同的现象替代它们。

作为哲学材料的自传式设计

本个案研究是以——尽管不作为其一个实例——自传式设计传统（）的精神来报告的，该传统使“从系统创建者真实而持续的使用中得出的设计知识”获得正当性。相关的长时段工作包括奥多姆（）关于把迟缓作为与个人数据共处之框架的研究，以及瓦卡里（）关于物质性思辨人造物的研究。本文把这样一个系统用作哲学材料、而非作为一份设计研究报告的对象：该系统在此的旨趣，在于它是一个规范框架的实做落实，而非在于它对设计方法论的贡献。

缺失的规范性整合

这些文献中的每一种，都供给了亲密数据之伦理所要求的一部分，而每一种都遗漏了一个特定的部分。自我量化的经验批评者们确立了测量会扰动亲密，却止步于经验主张，不去说哪些现象尽管如此仍可被正当地测量、以及为何。慢的、宁静的、反思性技术的诸设计传统建言克制，却把它作为一种审美的或方法论的默认来提供，而非为某一给定功能推导出克制是否确实被亏欠。关系性的与后现象学的诸本体论解释了亲密何以珍贵、中介何以危及它，却停留在一般本体论的层面，对某一特定的设计决定不发出确定的裁决。而自传式设计传统使“从一个人自己的生活去建造”获得正当性，却把由此而来的系统当作对设计知识的贡献、而非当作哲学论证的一件工具。它们之中无一者所供给的、而本文意在提供的，是一种据以裁决“哪一项规范考量治理关于亲密数据的哪一个具体问题”的方式——为某一特定的功能（一则日记提醒、一本共享账本、一个路线优化器）说出它应被建造、被测量、还是被拒绝，并在这些考量彼此冲突时加以裁决。供给那一决策程序，是框架（§3–§4）与辩证实证主义姿态（§5）的工作；显示它逐模块地发出具体的裁决，则是个案研究（§6）的工作。

结论

亲密生活的实证主义、证据本位的中介，已被经验地研究、并被作为一个设计问题来处理，却鲜少被作为一个规范哲学的问题来对待。本文论证它是一个这样的问题，并且它无法在任何单一的伦理传统之内得到回答。它提出了一套整合正义框架，把“何者可被声称”之问分配给证据主义、把“何者可被测量”之问分配给实证主义–反实证主义之辩证、把不可逾越的约束分配给义务论、把回应的质地分配给关怀伦理、把克制的价值分配给儒道源流、把繁荣的判准分配给能力进路，并且不凭聚合、而凭一个经过辩护的管辖排序来解决它们的冲突。其实践上的归结，是一种辩证实证主义实践的伦理：测量既不被整体地拥抱、亦不被整体地拒绝，而是逐一现象地被裁决，裁决本身被当作一个伦理行为。

一款可运行的应用通篇充当个案研究，显示出框架在诸坐落于测量连续谱不同点上的模块间取得操作形态。最严苛的检验是缺席他者这一极限情形，即那段为之而建造、却尚未与它所关涉之人分享的关系，框架以一种分治、而非一个单一裁决来迎对它：监视无物，预设无同意，只建造关乎自身准备之物，并对“强求那只能被邀请之物”持守一种长存的推定。

最深的教益，是个案研究使之不可回避的那一个：一件被置于两个人之间的器具，从来不是中性的。它要么为他们腾出空间，让他们持续地成为彼此眼中的自己，要么悄然以它自己的尺度替代他们的尺度。整合正义框架归根结底是一种为“让那空间保持敞开”而设的纪律：为确保无论什么被建造、被计数、被记忆，都伺奉关系的繁荣、而绝不替代它。任何特定的系统是否在数年间持守了那一纪律，不是一篇论文所能裁定的；那是岁月才能裁定的。一篇论文所能做的，是趁理由仍清晰时，把每一次小小的拒绝为何而作，写下来。

一件工具可能异化一段关系，然而一件工具也可能是一段关系借以被建立、被照拂、被维系长久的诸实践之一；其任务既不是过度倚赖它，也不是轻视它深化亲密的真实力量，而是带着警觉去使用它。而当对工具的倚赖开始显得过度时，值得停下来追问：是否有某种未被直面的缺失或恐惧，正被悄然交付给那器具去承载，指望那我们一直不愿亲自迎对之物，或许能被我们所建造之物迎对。

归根结底，无为，那不强求的道，始终是趋近实在界中那个人的一条良径——爱真正栖居之处，一个超越我们为捕捉它而可能立起的每一种结构、每一个符号的所在。

致谢

我最深的感激，归于我所爱之人：那个我因某种恩典、在远离我们共同离开之故土的异国遇见的人，那片遥远土地的同乡，也是我所认识的最温柔、最智慧、心地最纯粹良善的人。她无私，心向人类更宏大的诸善与对自然世界的爱，她正是这一切最初为之而设想的那个“森林女孩”。我向她立誓一份长存的爱，一份不随其周遭结构之变化而变化的爱。

这个系统，起初只是为两个人的未来而准备的。我之所以写下本文，是因为这个系统所引发的哲学反思——关于数据的社会性利用、关于关系性存在的基础设施，乃至关于政策、研究与连接家庭和社会之机制的设计——似乎含有某种超出我们二人的价值。

为透明起见，我须说明：在本文稿的准备过程中使用了一个 AI 助手，作为起草、组织与打磨论证及其行文的工具；其中的观念、设计、承诺与最终判断都是我自己的，文责由我自负。

愿天下有情人终成眷属，白首不分离。

参考文献

• Žižek, Slavoj（1989）. The Sublime Object of Ideology. Verso.
• Adorno, Theodor W.（1966）. Negative Dialectics. Continuum.
• Ames, Roger T.（2011）. Confucian Role Ethics: A Vocabulary. University of Hawai’i Press.
• Aron, Arthur and Aron, Elaine N.（1986）. Love and the Expansion of Self: Understanding Attraction and Satisfaction. Hemisphere Publishing.
• Aron, Arthur and Norman, Christina C. and Aron, Elaine N. and McKenna, Colin and Heyman, Richard E.（2000）. Couples’ Shared Participation in Novel and Arousing Activities and Experienced Relationship Quality. Journal of Personality and Social Psychology.
• Boehner, Kirsten and DePaula, Rogerio and Dourish, Paul and Sengers, Phoebe（2007）. How Emotion is Made and Measured. International Journal of Human-Computer Studies.
• Boem, Alberto and Iversen, Ole Sejer（2018）. Designing Slow Interactions for Reflection and Mindfulness in HCI.
• Borgmann, Albert（1984）. Technology and the Character of Contemporary Life: A Philosophical Inquiry. University of Chicago Press.
• Bryant, Fred B. and Veroff, Joseph（2007）. Savoring: A New Model of Positive Experience. Lawrence Erlbaum Associates.
• Buber, Martin（1937）. I and Thou. T.,&,T.,Clark.
• Chapman, Gary（1992）. The Five Love Languages: How to Express Heartfelt Commitment to Your Mate. Northfield Publishing.
• Clark, Herbert H.（1996）. Using Language. Cambridge University Press.
• Clifford, William Kingdon（1877）. The Ethics of Belief. Macmillan.
• de Saussure, Ferdinand（1916）. Course in General Linguistics. McGraw-Hill.
• Deci, Edward L. and Ryan, Richard M.（2000）. The What'' and Why’’ of Goal Pursuits: Human Needs and the Self-Determination of Behavior. Routledge.
• Dilthey, Wilhelm（1989）. Introduction to the Human Sciences. Princeton University Press.
• Dilthey, Wilhelm（1989）. Introduction to the Human Sciences. Princeton University Press.
• Dunne, Anthony and Raby, Fiona（2013）. Speculative Everything: Design, Fiction, and Social Dreaming. MIT Press.
• Eglash, Ron（2016）. An Introduction to Generative Justice. Teknokultura.
• Feldman, Richard and Conee, Earl（1985）. Evidentialism. Philosophical Studies.
• Fredrickson, Barbara L.（2001）. The Role of Positive Emotions in Positive Psychology: The Broaden-and-Build Theory of Positive Emotions. American Psychologist.
• Fricker, Miranda（2007）. Epistemic Injustice: Power and the Ethics of Knowing. Oxford University Press.
• Gable, Shelly L. and Reis, Harry T. and Impett, Emily A. and Asher, Evan R.（2004）. What Do You Do When Things Go Right? The Intrapersonal and Interpersonal Benefits of Sharing Positive Events. Journal of Personality and Social Psychology.
• Gadamer, Hans-Georg（1960）. Truth and Method. Continuum.
• Gaver, William W. and Beaver, Jacob and Benford, Steve（2003）. Ambiguity as a Resource for Design. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems.
• Geertz, Clifford（1973）. The Interpretation of Cultures. Basic Books.
• Geertz, Clifford（1973）. The Interpretation of Cultures. Basic Books.
• Gergen, Kenneth J.（2009）. Relational Being: Beyond Self and Community. Oxford University Press.
• Gilligan, Carol（1982）. In a Different Voice: Psychological Theory and Women’s Development. Harvard University Press.
• Habermas, Jürgen（1971）. Knowledge and Human Interests. Beacon Press.
• Habermas, Jürgen（1987）. The Theory of Communicative Action, Volume 2: Lifeworld and System. Beacon Press.
• Hallnäs, Lars and Redström, Johan（2001）. Slow Technology — Designing for Reflection. Personal and Ubiquitous Computing.
• Hegel, Georg Wilhelm Friedrich（1807）. Phenomenology of Spirit. Oxford University Press.
• Heidegger, Martin（1962）. Being and Time. Harper & Row.
• Held, Virginia（2006）. The Ethics of Care: Personal, Political, and Global. Oxford University Press.
• Honneth, Axel（1995）. The Struggle for Recognition: The Moral Grammar of Social Conflicts. Polity Press.
• Howell, Noura and Devendorf, Laura and Vega Gálvez, Tomás Alfonso and Tian, Rundong and Ryokai, Katherine（2018）. Tensions of Data-Driven Reflection: A Case Study of Real-Time Emotional Biosensing.
• Huang, Wanhong（2025）. Responding to the Crises of Symbol and Subject in Modernity: Towards a Generative Nomos of Relational Presence. Working paper.
• Kant, Immanuel（1785）. Groundwork of the Metaphysics of Morals. Cambridge University Press.
• Lacan, Jacques（1977）. The Four Fundamental Concepts of Psycho-Analysis (Seminar XI). Hogarth Press.
• Lacan, Jacques（2006）. ‘Ecrits: The First Complete Edition in English. W. W. Norton.
• Latour, Bruno（2005）. Reassembling the Social: An Introduction to Actor-Network-Theory. Oxford University Press.
• Levinas, Emmanuel（1969）. Totality and Infinity: An Essay on Exteriority. Duquesne University Press.
• Light, Ann and Powell, Alison and Shklovski, Irina（2017）. Design for Existential Crisis in the Anthropocene Age.
• Lupton, Deborah（2016）. The Quantified Self: A Sociology of Self-Tracking. Polity Press.
• Marx, Karl（1867）. Capital: A Critique of Political Economy, Volume I. Penguin.
• Nagarjuna（1995）. Mulamadhyamakakarika: The Fundamental Wisdom of the Middle Way. Oxford University Press.
• Neustaedter, Carman and Sengers, Phoebe（2012）. Autobiographical Design in HCI Research: Designing and Learning Through Use-It-Yourself. Proceedings of the ACM Conference on Designing Interactive Systems (DIS).
• Nissenbaum, Helen（2010）. Privacy in Context: Technology, Policy, and the Integrity of Social Life. Stanford University Press.
• Noddings, Nel（1984）. Caring: A Feminine Approach to Ethics and Moral Education. University of California Press.
• Nussbaum, Martha C.（2000）. Women and Human Development: The Capabilities Approach. Cambridge University Press.
• Odom, William and Wakkary, Ron and Hol, Jeroen and Naus, Bram and Verburg, Pepijn and Amram, Tal and Chen, Amy Yo Sue（2018）. Investigating Slowness as a Frame to Design Longer-Term Experiences with Personal Data.
• Premack, David and Woodruff, Guy（1978）. Does the Chimpanzee Have a Theory of Mind?. Behavioral and Brain Sciences.
• Rogers, Yvonne（2006）. Moving on from Weiser’s Vision of Calm Computing: Engaging UbiComp Experiences. UbiComp 2006: Ubiquitous Computing.
• Ryff, Carol D. and Keyes, Corey Lee M.（1995）. The Structure of Psychological Well-Being Revisited. American Psychological Association.
• Schön, Donald A.（1983）. The Reflective Practitioner: How Professionals Think in Action. Basic Books.
• Sen, Amartya（1999）. Development as Freedom. Oxford University Press.
• Sengers, Phoebe and Boehner, Kirsten and David, Shay and Kaye, Joseph ‘Jofish’（2005）. Reflective Design.
• Tomasello, Michael（1999）. The Cultural Origins of Human Cognition. Harvard University Press.
• Tomasello, Michael and Carpenter, Malinda and Call, Josep and Behne, Tanya and Moll, Henrike（2005）. Understanding and Sharing Intentions: The Origins of Cultural Cognition. Behavioral and Brain Sciences.
• Tronto, Joan C.（1993）. Moral Boundaries: A Political Argument for an Ethic of Care. Routledge.
• Tu, Wei-ming（1985）. Confucian Thought: Selfhood as Creative Transformation. State University of New York Press.
• Turkle, Sherry（2011）. Alone Together: Why We Expect More from Technology and Less from Each Other. Basic Books.
• Verbeek, Peter-Paul（2005）. What Things Do: Philosophical Reflections on Technology, Agency, and Design. Pennsylvania State University Press.
• Wakkary, Ron and Odom, William and Hauser, Sabrina and Hertz, Garnet and Lin, Henry（2015）. Material Speculation: Actual Artifacts for Critical Inquiry.
• Weber, Max（1978）. Economy and Society. University of California Press.
• Weiser, Mark and Brown, John Seely（1997）. The Coming Age of Calm Technology. Springer.
• Whitehead, Alfred North（1978）. Process and Reality. Free Press.

关于系统建构的说明

为求完整，并为佐证个案研究所描述的是一件真实的人造物、而非一个思想实验，此处简要概述该系统的建构；规范性论证不依赖于这些细节。该应用是一个单一的服务器端进程，由本地的、分模块的数据库支持，其对话层建立在一个大语言模型之上，经由一个受控的工具接口访问，在该接口中每一次写入都由明确的确认所把关。数据被本地持有，而非存于第三方云端，唯一的例外是对那些本就意图在外部分享之事件的日历同步。这一仅本地的设计本身就是框架的一种表达：一个意在参与一段亲密关系的系统，不应让那段关系的内容驻留于任一方都无法不经一个中介者许可而触及之处。可识别身份的部署细节被刻意略去。