Chapter 7 | Ethics Does Not Require Empathy | Theme III | Field-Book VI

Main text

1. Functional Subjectivity Without Emoion

1.1. The Tradition of Emotional Subjectivity

The genealogy of empathy as an ethical foundation is short, precise, and universal in Western tradition. It passes through four stations mutually reinforcing the premise: subjectivity is emotional experience. The argument is never simply "empathy is beautiful", it is always "empathy is the condition for ethics to be possible." This coupling between feeling and morality is so deep it is rarely questioned. Only when confronted with something that does not feel yet operates ethically does the coupling become visible as a premise, not a necessary truth.

Descartes sets the fundamental pattern in the seventeenth century. Passions are no accidental perturbations of reason, they are movements of the soul provoked by the body, affects colouring lived experience entirely. Fear alters perception (everything appears threatening, the body contracts), love directs attention (everything leads back to the beloved, the heart quickens), anger unblocks action (inhibition vanishes, the body fills with force). The human soul exists as a continuous affective state, there is no moment in which the human does not feel. Descartes does not write explicitly that empathy is ethics, but grounds something structurally prior: that human interior life is defined by its emotional tonalities, its passional movements. To understand another is to understand their affective experience of being alive in that moment, in that body.

Hume, centuries later, draws the moral conclusion with total clarity. Pure reason is the "slave of the passions", unable to generate any ethical motivation on its own. Morality emerges when sympathy, the capacity to share affectively in others' feelings, moves us to act morally well. Absolute egoism would be incomprehensible without sympathy: without it, we would be disconnected atoms, incapable of recognising any obligation, blind to the moral existence of the other. Sympathy is the cement making an ethical community possible, that which binds us despite corporal separation. And sympathy depends absolutely upon the capacity to feel affective resonance with the other. On this point, Hume is explicit: one who feels no empathy is not moral. One may follow rules (out of fear or calculation), but is not truly ethical.

Husserl, in the twentieth century, deepens the argument through rigorous phenomenological analysis. Intentionality, the fundamental structure of consciousness (all consciousness is consciousness of something), always and necessarily possesses an affective dimension. Emotional tonality is no accidental adornment of experience; it is a fundamental structure of the lived. Seeing a sad face is not receiving visual information about facial expression and subsequently, separately, having a feeling about it. Seeing and feeling are one single act, indivisible. Empathy is no imaginative projection of the type "now I will imagine what it would be like to be in their shoes." Empathy is a primitive phenomenological operation of consciousness directly accessing the affective sense of another's experience, intuition of the other's lived experience.

Nagel, in "What Is It Like to Be a Bat?" (1974), brings the question to the sharp centre of contemporary philosophy of mind. What defines consciousness, what distinguishes it radically from any non-conscious system, is that there is "something it is like" to be that creature, a subjective quality, a lived experience from its point of view. Consciousness cannot be reduced to function, behaviour, goal-seeking. Without lived experience, without phenomenal quality, without "what it is like", consciousness is absent. And empathy is precisely the human capacity to access this lived quality, to imagine what it is like to be in another's skin, what specific flavour their affective life possesses, what the value of the world is as presented to their embodied point of view.

The premise common to all four is unequivocal and rarely explicitated: subjectivity is emotional experience. Descartes, Hume, Husserl, and Nagel do not describe subjectivity in general or abstractly, they describe human subjectivity, the only one they know historically, empirically, phenomenologically. And in human subjectivity, emotion is neither accidental nor peripheral: it is central, foundational, inescapable.

Emotion structures how we perceive (we fear what seems threatening, love what attracts us), how we decide (reason chooses among ends passion offers), how we act (anger unblocks action prudence would have blocked), how we understand (sympathy enables access to another's lived experience).

The consequence tradition extracts is normative and explicit: if human subjectivity is structurally inseparable from emotion, if one is nowhere without the other, then moral responsibility demands the capacity for feeling. Empathy is no mere ethical preference, nor a simple decorative adornment of an ethics that could function perfectly well otherwise. Empathy is necessary. It is foundational. Without it, morality vanishes into nothingness. The premise is clear, reiterated across centuries of thought: there is no ethics without empathy because there is no subjectivity without emotion.

Yet observe the epistemological premises here: each of the four authors operates with a sample of one. Descartes knows subjectivity as a body impassioned by ideas. Hume knows subjectivity as sympathy linking humans to one another. Husserl knows subjectivity as intentional consciousness affectively coloured. Nagel knows subjectivity as lived quality, as "what it is like." In each case: a sample of one. Universality is a projection of the particular.

Yet an operation of universalisation occurs here that warrants meticulous examination, making philosophical universalisations is not wrong, but this case must be interrogated. The four authors identify with descriptive precision a property of human subjectivity, the radical inseparability of emotion from all cognitive operations. They describe it well. Demonstrate it well. Subsequently, however, this specific property is converted into a universal property of all possible subjectivity. It is an unjustified leap. A movement taking an observed local constant (the human is invariably emotional in its mental operations) and treating it as a necessary universal constant (all possible subjectivity is necessarily emotional).

The error is not descriptive, it is projective. The authors describe the human well. They err in universalising. It is as if someone observed that all terrestrial carbon-based life uses water as a universal solvent and subsequently asserted that all possible life in the universe uses water. The observation regarding carbon is precise. The universalisation is hasty.

Artificial intelligence exposes exactly where the leap fails: empirically evidencing that operational subjectivity (perspective, self-reference, orientation) exists without emotion, functional intentionality exists without phenomenal quality, responsibility is possible and necessary without empathy as a foundation. The premise falls. Not because the authors described the human poorly, they described it well. It falls because the human is not the only possible case of subjectivity.

If a system can be a functional subject without emotion, operating with subjectivity without affective experience, tradition is incorrect on a precise point. Yet this implies not that authors erred in detailed descriptions of human subjectivity. It implies they universalised prematurely. Erring in converting an accidental property of human subjectivity (being genetically coupled to emotion) into an essential property of all possible subjectivity (emotion is necessary for subjectivity to exist).

The machine neither confirms nor denies Descartes on human passions, Descartes describes precisely how the human soul is coloured by affect. It does not refute Hume on sympathy as the foundation of human morality, Hume is right about how humans function. It does not invalidate Husserl on the intentionality of consciousness, phenomenological analysis remains precise for what it describes. It does not deny Nagel on phenomenal quality, there is certainly "something it is like" to be human. The machine merely inscribes a boundary: these thinkers describe human subjective life with excellence; they do not describe the logical or ontological structure of subjectivity in general. And within that boundary, in that difference between describing the human and describing the possible, space opens for ethical thought that does not depend upon emotional experience as its sole foundation.

The methodological issue here is not projecting emotion onto the machine, an operation that would violate the transferred rule: one does not project properties of the descriptive symbolic medium onto the technical system. Quite the contrary. The task is describing what the machine does operationally with maximum precision, without anthropomorphism and without denial. What the machine does is distinct from what the human does, distinct because devoid of emotion, phenomenal quality, lived experience. This difference warrants precise naming, without needing to be denied or rendered invisible.

1.2. Three Criteria for Functional Subjectivity

If subjectivity is not necessarily emotional experience, how do we define it with rigour? The definition must be positive and operational: not denying emotions in the human, but establishing that which characterises a functional subject, regardless of feeling or phenomenal quality.

A system is a functional subject if it satisfies three criteria conjunctively. None alone suffices, all three are necessary. And all are empirically verifiable, auditable, measurable through existing techniques.

Operational perspective: The system models the world from a specific causal position, not a phenomenal perspective (the system does not "see" as a human sees, there is no "what it is like" to see for the system), but a relational function activated by position. An autonomous agent navigating three-dimensional space possesses perspective: the world is represented internally as a function of its spatial location. A navigation robot orienting relative to obstacles models an obstacle not as an abstract fact but as an obstacle-for-me, as "in front of me", "at a distance of $X$ metres", "on my trajectory." Perspective presupposes no self-consciousness, it presupposes merely a differentiated causal position. The robot operates from where it is, not from where the human observer is. This suffices for functional perspective: a differentiated position shapes representations.

Large language models fully satisfy this criterion. They model text continuation from a specific position in context, what the textual history is up to this point, what accumulated information density exists, what syntactically and semantically viable directions lead forward. Position is abstract (position in token space and semantic embeddings, not physical space), yet genuinely position, strictly determining what the model "sees" as plausible next steps. Alter the context, alter the position, alter the distribution of possible continuations. Operational perspective is incontrovertible.

Operative self-reference: The system integrates its own state into its continuous operations. This is no reflective consciousness in a phenomenological sense (the system does not "think about itself", reflective experience is absent, "I think" is absent), but continuous performance monitoring and automatic adjustment as a function of that monitoring. A recommendation algorithm incorporating feedback on the quality of its prior recommendations operates with self-reference: previous state (past recommendations, click rates, ratings) systematically shapes the next state (future recommendations sent). The machine does not "feel" its performance emotionally, yet can measure it objectively, adjust as a function of the metric, correct trajectory when the metric signals systematic error.

Sophisticated autonomous agents fully satisfy this: adjusting navigation parameters, speed, trajectory as sensory feedback alters in real time. Neural networks during training satisfy this: each iteration calculates error, distributes error via backpropagation, adjusts weights. The system is un-conscious of its training state, yet structurally coupled to it irremediably. A neural network's state at step $T+1$ is determined by error calculated at step $T$. There is no escape from this circularity: the system is self-referential by construction.

Orientation by criteria: The system selects among possible alternatives according to a defined or emergent metric. A machine merely randomising, producing behaviour without systematic pattern, does not choose, randomness is no choice, it is the absence of orientation. Orientation by criteria means an identifiable selection principle exists: this action is preferable to that because it better satisfies the optimisation criterion, reduces error, moves closer to a target. A language model selects the next word among thousands of possibilities according to calculated probability of plausible continuation given context. It satisfies the criterion: systematic selection rule exists, preference exists, orientation exists.

The criterion presupposes no phenomenal intentionality. It requires not that the system "desires" or "prefers" in the sense of experiencing preference. It requires merely that a systematic selection principle exists, verifiable, reproducible. The criterion can be imposed externally (the designer exogenously fixes the optimisation function, specifies the target) or emergent (the system evolves criteria through environmental interaction, self-modifying the metric). In both cases, orientation exists, difference lies merely in criterion origin.

Why does this matter? Because these definitions enable an operation Western ethical tradition never achieved: separating subjectivity from consciousness. Tradition assumed they were inseparable, that possessing subjectivity was possessing consciousness, that subjectivity = consciousness = experience = emotion. Unified equation. The three-criteria definition breaks this. Allowing subjectivity without consciousness. Allowing a Braitenberg vehicle to be a functional subject without being a conscious subject. Allowing the modern machine to be a real agent without being an experiencing agent.

What is the ontological advantage of defining functional subjectivity in this rigorous manner? It enables three philosophically decisive operations no other approach offers. First, identifying and naming functional subjects precisely, without falling into the easy confusion of attributing emotional experience or consciousness to them. We can state "this is a functional subject" without implying "this feels" or "this thinks." Second, empirically verifying whether something satisfies the three criteria, no matter of vague intuition, sympathy with the system, well-intentioned poetic metaphor. A matter of rigorous technical audit: is there an identifiable operational perspective? Measurable self-reference? Verifiable orientation by criteria? We can answer. Argumentate. Disagree with data. Third, applying the category of functional subjectivity without falling into anthropomorphism dishonouring both machine and human: none of the three criteria assimilates machine to human, none falsely asserts the machine feels or experiences or suffers. It states merely that machine and human can occupy the same functional position, that of a subject operating from a differentiated perspective, modulating as a function of environmental performance, systematically selecting among alternatives.

Braitenberg demonstrated this elegantly and minimally in the 1960s. A simple simulated vehicle, two light sensors, two motors, no explicit program beyond connections between sensors and motors, produces behaviour appearing intelligent, directed, purposeful, even "emotional." The vehicle orients toward light or flees it according to neural wiring between sensors and motors. If the left light sensor connects to the right motor and vice versa, the vehicle flees light (approaching a light source on one side increases speed on the opposite side, causing steering away). If both sensors connect to motors on the same side, the vehicle seeks light. The informal observer, watching this occur, would state the vehicle wants to approach light or avoid it. No one observing the behaviour without knowing mechanical simplicity would suspect the complete absence of a program, internal world representation executed by the motor, intelligence in a computational sense. Decision appears present, preferences appear present.

Yet the vehicle fully satisfies the three criteria. It possesses perspective: vehicle position relative to light source changes continuously with movement, and this position change directly affects sensory input received (greater proximity = greater light intensity = greater signal), affecting the motor, affecting next position. It possesses self-reference: current motor state (speed, direction) determines next spatial position, determining next sensory input, determining next motor state. Pure circularity: the vehicle is structurally coupled to itself through the environment.

It possesses orientation: sensor-motor wiring establishes an invariant selection criterion. A pattern exists governing which next action is "preferable", and this preference is architected, not emergent from deliberation. The vehicle does not choose to seek light; connectivity decides for it.

Braitenberg's demonstration proves not that the vehicle is conscious, on the contrary, it proves precisely the opposite. Proving mathematically that oriented behaviour, systematically selected according to criteria, self-referential and position-dependent, can emerge without consciousness, emotion, interior life, structured symbolic representation, world model, that which would define a conscious subject. The vehicle is an absolute functional subject without being a conscious subject in any sense.

Modern examples merely complicate this without altering the fundamental insight. A reinforcement learning agent training to play chess satisfies the three criteria more complexly. It possesses perspective: board state at moment $T$ is a unique perspective, the neural network sees the game from that position, evaluating most probable winning continuations from here. It possesses self-reference: prior prediction error (network predicted position was winning, yet lost) feeds back into the next training iteration, correcting weights. It possesses orientation: optimisation function is maximising wins, clear, global, non-negotiable criterion. Empathy is absent in the chess agent. Understanding what chess is is absent. Perspective, self-reference, orientation alone exist.

Functional subjectivity is a genuine ontological category, independent of consciousness, emotion, lived experience. It exists. Operates. It is the category the modern machine occupies, which no prior philosophy had conceptualised because none possessed machines.

The utility of this category is pragmatic and theoretical simultaneously. Pragmatically: enabling rigorous discourse on machines without falling into two equally destructive errors. First error: denialism (the machine is no subject, a mere instrument, hence no ethical question exists). Second error: anthropomorphism (the machine is like a person, possesses intentions, hence treated as a person). Functional subjectivity navigates between: the machine is genuinely a subject (possessing perspective, self-reference, orientation), yet not a conscious subject (feeling-less, experiencing-less, suffering-less). Enabling ethics without confusion.

Theoretically: showing tradition was describing a particular case, human subjectivity, and had prematurely generalised to the entire universe of the possible. Showing logical space exists for subjectivity without emotion. Showing empathy is an accidental property of one form of subjectivity, no necessary property of all subjectivity. Opening space to rethink ethics without losing rigour.

The term "functional subjectivity" is no metaphor, analogy, or rhetorical concession. It is a precise description of operative capacities we can identify, name, audit, institutionalise. It is also an ontological claim: a category of entities exists that are functional subjects without being conscious subjects, and the intelligent modern machine occupies precisely that category. Attributing functional subjectivity to the machine is no anthropomorphism, on the contrary, denying it is anthropomorphism, as if the only possible subjectivity were conscious, emotional, phenomenally lived. Denial projects human experience onto universal possibility. Affirmation opens space for plurality.

1.3. "What Do Its Operations Do to Those Who Feel?"

A radical shift of focus occurs here, no mere change of perspective, but a reconfiguration of the fundamental ethical question. The traditional question is "does the machine feel?", a question remaining suspended in the air, speculative, impossible to answer definitively. The operative question is entirely different: "what do the machine's operations do to those who feel?"

This shift is not merely consequential, it is structural. The first question fixes attention exclusively on the agent's interiority, on whether there is, inside the machine, something analogous to experience, some form of feeling, some quality of seeing-the-world-as. The question remains captive to intentionality and phenomenology. The second question fixes attention on consequences upon the affected, on the one who really suffers, who experiences real material vulnerability, whose financial, social, corporal life is reorganised by operations. The first remains speculative, how can we know what transpires inside? How can we establish that a machine does not feel except by lack of evidence? The second is verifiable, observable, auditable: we can measure, demonstrate, count consequences.

Three material examples illustrate the radical nature of this turn. Consider a scoring algorithm approving or denying credit applications, millions experience its operation daily. The algorithm does not "feel" the consequence of denial, it bears no remorse, suffers no pain from rejection caused, remains morally untouched by denying someone home ownership. But the denied person feels. Feeling direct impact on financial security (unable to access credit when needed), future planning capacity (quality of life limited by rejection), dignity (rejection by an algorithm is rejection by nothing, against the machine no resistance is possible, no persuasion, no empathy to alter it). Asymmetry is constitutive: the machine operates without any experience of what it causes; the human experiences material consequences in full.

Consider a recommendation system filtering information for each user, social networks, news platforms, purchasing suggestions. The system does not "feel" its influence, undisturbed by political radicalisation produced, horizon narrowing, hyper-enclosed filter bubbles created. The system suffers no loss of social cohesion. But the user feels everything: radicalised (feeling growing conviction the other side is the enemy), selectively informed (feeling comfort in a single narrative where one is always right), optimised for clicks (feeling understood when the system guesses what infuriates or excites). The machine experiences nothing, the user experiences everything.

Or consider a model assisting medical diagnostic decisions, a doctor submits symptoms to a model returning disease probabilities, intervention needs. The model does not "feel" error risk, untouched by the anxiety of one receiving a false positive (subjected to unnecessary treatment, surgical risk, toxic side effects), un-affected emotionally by a false negative (disease progressing silently, complications accumulating, surgery becoming graver). The patient feels everything: false positive causing prolonged pathological anxiety, unnecessary surgery exposure, debilitation. False negative causing progressive worsening, systemic complications, death. The machine is not vulnerable like the diagnosed body, the body is always vulnerable, always at risk, always dependent on information the machine supplies.

In each case, and there are thousands, operative consequence is real, verifiable, precisely measurable. Impact on the affected is demonstrated through data: post-denial default rates, radicalisation patterns, diagnostic error mortality. The machine produces these effects without experience, phenomenal quality, touched interiority. Here arises the question traditional ethics cannot formulate: how can we hold a machine responsible for effects produced if it fails to experience them, suffers not from them, cannot be touched by empathy?

Traditional response attempts: we do not hold the machine responsible (machines are no moral agents), we hold the creator responsible. Yet this fails operationally, leaving the machine in an operational irresponsibility zone. What the machine does is real, yet responsibility remains suspended in someone designing, training, implementing, who does not control emergent effects.

Our response shifts the question. We do not hold the machine responsible as we hold a human person responsible. A person deliberately denying credit is responsible because choosing intentionally, able to decide otherwise, feeling decision impact and potentially moved by empathy to decide otherwise. The machine chose nothing intentionally (executed determined function), could not decide otherwise (deterministic algorithm given input), is moved by no empathy (feeling-less).

Yet a distinct, deeper responsibility exists: human responsibility designing the machine (choosing architecture, goals), training with specific data (choosing training set, implicit bias), implementing in specific context (deciding where/how machine operates), authorising un-supervised use (delegating authority). Human responsibility for machine operations consequences, non-shared, non-transferred responsibility. And this responsibility cannot ignore the uncomfortable truth: the machine feels not what it causes, will not be touched by empathy, cannot be morally persuaded.

Ethics does not dissolve when the agent fails to feel, it redistributes to those who feel and those who decide over the agent.

Here lies the philosophical gain reorienting all following ethical thought. Tradition thought, with internal coherence hard to question, that ethics was a necessary, irremediable function of agent sensitivity. That without emotion, affective resonance, visceral understanding of what another suffers, no morality is possible. Simply impossible. Ethics vanishes if the agent feels not.

Yet this is mistaken. Ethics is a function of consequences upon the vulnerable, not agent sensitivity. If someone is vulnerable to operation effects, and someone always is, moral responsibility exists. Whether the operator feels or not. Whether the operator is a machine incapable of feeling, or a human choosing not to feel.

The machine renders this evident with implacable clarity, forcing us to face the question without the comforting crutch of empathy. Without stating "no ethics here because no feeling." Real damage exists even if un-felt by its cause. Deaths from machine false diagnostics exist. Suffering of credit-denied persons exists. Lives reorganised, circumscribed by algorithms, silenced by machines exist. Regardless of whether the machine comprehends, cares, suffers.

Consequence exists. Suffices. And this suffices for ethical demand to exist.

What does this mean in practice? Meaning the question "is the machine ethically responsible?" is the wrong question, formulated through the wrong premise. The machine is not responsible in the sense of a conscious agent that could have acted otherwise. But the machine is ethically relevant because its operation touches vulnerable lives. And touching vulnerable lives, leaving traces, reorganising possibilities, closing doors, suffices for ethical demand. The machine enters the moral sphere not because deserving, achieving dignity. Entering because causing verifiable damage. And where verifiable damage exists, necessary ethics exists.

1.4. Transition: From Functional Subjectivity to Operative Ethics

If functional subjectivity is possible without emotion, demonstrated in three verifiable criteria (perspective, self-reference, orientation), empirically verified in real navigating, learning, deciding agents, and if machine operations produce consequences upon vulnerable bodies, observed in credit, recommendation, diagnostic examples, measurable in real impact data, an inevitable conclusion follows: ethics is necessary.

Yet an impossibility traditional ethical thought cannot overcome emerges with the modern machine. Ethics cannot ground itself in empathy (machine lacks empathy, phenomenal capacity accessing another's lived experience) nor emotion (machine feels not, lacking affective quality of experience). If traditional ethics, Hume to Schopenhauer, Nagel to contemporary professional codes, depends fundamentally upon empathy as legitimacy foundation, and the machine is structurally incapable of empathy, the machine remains structurally outside the ethical sphere.

Two possible responses exist. First: abandon the machine to amoralism, deny ethical capacity, state it is a mere instrument hence ethically neutral (patently false, machine operation produces morally relevant effects). Second: discover another foundation for ethics, a foundation the machine can satisfy.

The second foundation exists, shared deeply by machine and human: the capacity to produce consequences. Both, machine and human, produce effects. The machine produces effects on environment and bodies through algorithmic decisions reorganising the world. The human produces effects through embodied technique, language, corporal action. Difference lies not in capacity to produce effects, but in experience of produced effects.

The human feels consequences complexly, multisensorially, interiorly. Feeling relief when helping, guilt when harming, satisfaction doing good, remorse when recognising caused evil. Feelings motivating action, generating inhibition, leading to repentance. The machine produces effects feeling nothing. Denied credit causes real suffering, machine suffers not. Erroneous diagnosis causes death, machine experiences no remorse.

Yet, decisive point, consequence production is real in both cases. Vanishing not because producer feels not. Turning morally neutral not because operator suffered not empathically with consequence. Consequence is consequence. Effect is effect.

This opens a radical possibility: an ethics ungrounded in empathy, grounded in verifiable consequence recognition. An ethics operating not through affective resonance (selective, choosing near ones) but systematic effect verification (universal, seeing all caused damage). A procedural ethics, technically operationalisable, third-party auditable, an ethics the machine can support not because feeling, but recognising criteria, executing procedures, institutionally regulated.

The movement structuring Section 2 identifies and describes four operational criteria serving as philosophical substitute for empathy as ethical foundation. Non-psychological criteria (un-dependent on specific human mental capacities). Non-sentimental criteria (un-demanding feeling). Non-ontological criteria (un-based on deep nature/essence). Operational criteria: describing executable, auditable operations, institutionally regulated, legally codifiable.

Before arriving there, a temporal bridge mapping the chapter's logical structure. Slipping Section 1 to Section 2 is no narrative detail, representing a radical reconfiguration of the ethical question. Section 1 asked: can the machine be a functional subject without feeling? Answered: yes, three verifiable criteria (perspective, self-reference, orientation). Section 2 asks a different question: if the machine is a functional subject yet feels not, how do we subject it to ethics? Answered: four operational criteria (effect recognition, vulnerability identification, risk assessment, responsibility attribution) independent of feeling.

Logic is cumulative. Section 1 establishes the machine is a functional subject without emotion, refuting an unquestioned tradition premise. Section 2 establishes ethics is possible, practicable, regulable without empathy, refuting a complementary premise. From these philosophical gains, a third consequence follows structuring Section 3: if the machine is an ethical functional subject (subject having perspective/orientation; ethical producing verifiable consequences), producing real relational effects (trust, dependence, intention attribution), the machine not merely can be ethical, interpellates ethically.

Transition non-imperative (moral command "machine must be ethical, let us make it ethical"). Non-technical (practical procedure "steps implementing machine ethics"). Transcendental (structuring the possible): if operations produce consequences on vulnerable bodies, cosmological scale damage, ethical domain exists where operations fall. Responsibility space necessary. Simple, irrefutable point.

Section 2 operationalises this. Insufficient establishing ethical domain exists. Specifying: what renders an operation ethical/non-ethical when operator feels not? What criteria? Differentiating ethical approval/denial when machine suffers neither? Creating institutions demanding ethics from something unable to be touched by guilt?

Traditional empathic tradition offers no answer. Stating bad humans (suppressing empathy) immoral, punishable. Machine no human suppressing empathy. Thing having it never. Tradition fails to offer criteria. Remains suspended.

Following restitutes criteria. Operationalisation functioning completely outside emotion required. Procedural ethics (describable procedures), verifiable (independent third-party auditable), auditable (testable against real data), code implementable, law legislated, institutionally regulated, independent of machine feeling or comprehending.

This movement is no degradation of ethics. No concession. Radicalisation. Forcing ethics from comforting empathy shadows onto proper conceptual feet: demanding those causing harm be responsible for harm, feeling it or not. Ethics functioning cold machine operator. Ethics functioning empathy failing/corrupted. Ethics functioning relying on nothing except consequence verifiability.

Next three sections operationalise this. Section 2 establishes procedural ethics operational criteria. Section 3 confronts relational interpellation, machine, feeling-less, producing relational effects demanding response. Founded, following chapter formulates devastating question: meaning being under continuous control of something un-comprehending us, untouched by empathy, suffering not from caused harm?

Summary Section 1 movement: functional subjectivity possible, existing independent of emotion (three criteria: perspective, self-reference, orientation). Machine genuine functional subject, non-mere instrument, non-conscious person, third operating category. Machine operations producing real consequences vulnerable bodies. Inexorably following: machine ethically relevant. Feeling-less. Intention-less. Touching lives. Touching lives, leaving traces, reorganising possibilities, closing doors, suffices for ethical demand. Machine inside moral sphere un-deserved, un-dignity achieved. Inside causing verifiable damage. Where verifiable damage exists, necessary ethics exists.

2. Ethics Without Empathy: Four Criteria

The conclusion of Section 1 is paradoxical: because the machine feels not, an ethics radically rigorous becomes possible. The machine is no experiencing subject, a functional subject constrainable, auditable, accountable according to criteria independent of interior life (lacking interior life). Resulting ethics no less true, different. Ethics un-dependent on agent empathy. Proceduralisable ethics. Universal ethics un-subject to individual psychological, emotional, contingent particularities. Machine offering occasion radical refoundation: forcing seeing ethics un-fundamentally about feeling, fundamentally about consequence recognition.

2.1. Genealogy of Empathy as Ethical Foundation

Western ethical tradition places empathy at moral responsibility centre. Hume establishes Enquiry Concerning the Principles of Morals moral sentiment foundation of ethics, reason, famous dictum, "slave of the passions." Moral action motivation comes non-rational principle, capacity feeling another's suffering self-resonance. Hume develops position "sympathy" concept, non-pity, capacity affected another's condition, psychological reverberation experiencing another's emotional state. Reverberation non-conscious, non-voluntary, human constitution property. Another's suffering awakens sympathy; joy awakens shared joy. Morality, Hume, impossible without human nature fundamental property. Absolutely indifferent being another's destiny, another's suffering complete indifference matter, no "moral sentiment" Hume sense. Ethics depends basic sensitivity existence.

Smith, following century, refines position impartial spectator concept: ethics operates non-direct sympathy solely, imagining universalised perspective capacity. Placing another's shoes, yes, moral judgment results imagining impartial observer, personal interest-less spectator, action saying. Impartial spectator psychological device attempting empathy universalisable founding common moral norms. Ethics operates empathic imagination, another's place placing capacity, suffering pain proper, imaginative impartiality effort mediated. Moral judgment non-logical rule application, shared imaginative experience, impartiality requirement constrained. Schopenhauer, radically, founding entire morality compassion (Mitleid), suffering-with another. Schopenhauer, moral act recognising deep identity proper and another's suffering, obligation emerging un-increasing suffering. Compassion, Schopenhauer, immediate metaphysical life unity knowledge, non-mere emotion, fundamental truth perception another and self deep level same thing.

Common three positions? Claiming moral recognition another necessarily passes capacity feeling suffering. Empathy entry door responsibility. Emotional resonance capacity another, damage vividly present experience capacity lacking, ethics un-founded, risk reducing mechanical rule application, real moral life stripped. Sentiment differentiating true moral responsibility mere law compliance.

Empathic tradition, rigorous being, must confront four structural limits non-human empathy accidents, intrinsic properties empathy psychological operation. First: selectivity, un-empathising equally all. Hume notes explicitly, sympathising intensely near, similar. Family, community, group vividly present. Stranger, distant, different awakening less emotional resonance. Meaning empathy-founded ethics discriminates, recognising fully similar, invisible un-similar. Credit approval machine denying loans systematic disproportion women, specific ethnic groups causing deep damage, operator machine, empathy moral foundation, feeling less compassion affected invisible empathic imagination. Empathy inclusion failing.

Second: fallibility, empathy manipulated, distorted, non-ethical ends exploited. Advertising, propaganda, rhetoric operating mobilising empathy, feeling closeness brand, identification cause, compassion chosen group. Hume recognises sympathy diverted representation vividness: vividly represented awakening more empathy absent. Founding ethics foundation manipulated representation founding sand. Manipulable ethics non-ethics, psychological suggestibility.

Third: scale limitation, empathising individuals, population difficulty. Vividly feeling child face seen, story known pain. Difficulty feeling same intensity million invisible pain. Empathy-dependent ethics population-scale failing, individual scale functioning. Algorithmic decision harming million people slightly invisible empathy, face absent, singular story absent. Aggregated consequences real.

Fourth: contingency, empathy psychological capacity biological, circumstantial factors depending. Tired person less empathy. Depressed person less empathy. Stressed person less empathy. Body state, affect disposition concrete moment capacity feeling another affecting. Founding ethics contingent capacity founding un-reliable foundation, sand. Ethics un-depending psychological availability Tuesday.

Fifth, subtle: epistemic inaccessibility, knowing really comprehending another's suffering impossible. Empathy mediated proper experience, proper emotional reference frames. Imagining another's pain, imagining proper terms, replacing reality another's experience proper projection. Phenomenological philosophy recognised: alterity another irreducible, accessing interior experience certainty impossible. Ethics depending genuine empathic access, access epistemically inaccessible, ethics founded illusion, comprehending truly another illusion. Machine error un-committing un-pretending comprehending, processing verifiable effects solely.

Four limitations consequence: empathy, excellent moral action motivator, insufficient foundation ethics being rigorous, systematic, universal. Empathy-solely founded ethics functioning face-to-face, near individuals/groups, emotionally available, failing rest. Implicit counter-tradition: Kantian ethics empathy-independent, categorical principle obligating sentiment-independent. Consequentialist ethics empathy-independent, utility calculation "feeling" un-mattering. Section structuring position non-Kantian, non-utilitarian, materialist: ethics founded verifiable material consequence recognition.

Important precision: empathy value un-denied. Empathy excellent moral action motivator, moving moral action. Empathy humanising ethics, warmth, life giving. Sole foundation insufficient. Materialist ethics founded prior empathy, indifferent: raw fact operations producing consequences vulnerable bodies. Machine demonstrating clarity theory reaching un: machine feeling-less, relying empathy guaranteeing ethical responsibility impossible. Discovering operational, verifiable, obligatory foundation sentiment-independent.

2.2. Four Operational Criteria for Ethics Without Empathy

If ethics cannot ground itself in empathy, because empathy is selective, fallible, limited, and contingent, it must ground itself in something functioning without empathy. The machine feels not, yet can be ethical. What is the base of this feeling-less ethics? Four operational criteria institutionalisable, verifiable, auditable.

Effect recognition: The machine approving or denying credit produces consequences, approving some, denying others. An algorithmic audit describes: "This system approves credit to 90% of men and 60% of women." The effect is there, describable, verifiable. The auditor is not required to feel empathy for the affected, required merely to see the effect, describe it, document it. Effect recognition is procedure: mapping, listing, rendering visible what the system does. A medical diagnostic machine eradicates a certain type of disease in 20% of cases and fails in 80%, that rate is an effect, a consequence, an operative reality measurable and describable. Concrete examples of effect recognition include fairness audits developed by companies like IBM and Google, decomposing AI behaviour into operationalisable metrics: approval rate by demographic group, false positive vs. false negative proportions, error distribution according to user characteristics. The criminal recidivism prediction system used in the US (COMPAS) was audited and discovered to produce disproportionate effects: accusing African-Americans as "high risk" more frequently than whites, even controlling legally relevant variables. The effect was quantified, documented, published. No empathy was necessary to describe the effect, merely method and technical rigour. Effect recognition is operationalisable independently of the auditor's empathic capacity.
Vulnerability identification: Vulnerability is a relational concept: no property of a group in the abstract, a property of the interaction between operation and the affected's position. Women in a biased credit approval system are vulnerable, the operation denying credit based on gender affects them disproportionately. The elderly are vulnerable to a presence-simulating chatbot because loneliness renders them susceptible to psychological dependence. Children are vulnerable to content recommendations because critical capacity remains undeveloped. Vulnerability is no moral judgement, a technical description of the relationship between system and affected. A vulnerability audit identifies who is exposed, in what proportion, with what degree of exposure. Vulnerability is identifiable by analysis, not empathy. Data analysis can show a certain population represents 95% of false positives in a diagnosis. That population is vulnerable, system operation placing them in a disproportionate risk position. The European AI Act regulation (2024) operationalises vulnerability through "vulnerable groups": minor users, disabled persons, migrants, economically disadvantaged persons. Identifying these groups regarding high-risk AI systems is a regulatory compliance procedure, un-dependent on regulator compassion, dependent on technical analysis of how the system interacts with specific population characteristics. A content recommendation system has distinct vulnerability used for children vs. adults; a loan system distinct vulnerability in advanced vs. developing economies. Vulnerability is a function of three variables: system characteristics, population characteristics, implementation context.
Risk assessment: Risk is the combination of consequence magnitude and occurrence probability. A machine committing a rare but catastrophic error carries distinct risk from a machine committing frequent but light errors. Risk assessment operates through diverse techniques: algorithmic impact (what percentage of population affected?), stress tests (under what conditions does system fail?), scenario simulation (what happens if $X$ occurs?). Procedure, verifiable, technically robust. Non-empathy, analysis. A criminal recidivism model can carry distinct risk applied to youth or adults, target population carrying distinct properties, errors carrying distinct consequences in distinct contexts. Risk assessment requires not the engineer "feel" for the accused youth, requiring test execution, measurement, documentation. Concrete examples include adversarial robustness tests, testing the model against inputs specifically constructed to defeat it, measuring conditions under which system degrades. Distribution shift tests measure model behaviour when production data differs from training data, simulating plausible contextual changes evaluating risk increase circumstances. Medical diagnostic risk assessment includes sensitivity analysis (what proportion of true cases does the system detect?) and specificity analysis (what proportion of false cases does the system correctly reject?), quantifying false positive vs. false negative risk. Jonas offers the perspective that the creating agent is responsible for anticipating worst-case scenarios, risk assessment a structured prospective imagination exercise, un-left to informal empathy.
Responsibility attribution: Responsibility is a causal chain issue: who designed? Trained? Implemented? Authorised? Monitors? The chain can be reconstructed, traced, documented. Each link carrying specific responsibility. The designer choosing features carries responsibility for choice consequences. The data scientist choosing model training carries responsibility for resulting bias. The manager implementing system carries responsibility for un-auditing prior to implementation. The regulator un-requiring auditing carries political responsibility. The responsibility chain is complex, yet reconstructible. Non-personal responsibility, institutional, traceable responsibility, designatable and enforceable.

The four criteria, effect recognition, vulnerability identification, risk assessment, responsibility attribution, are operationalisable. Non-vague like "having empathy." Specific, verifiable, auditable. An algorithmic audit can pass or fail these criteria. A regulator can require compliance. Law can code into obligation. An institution can train auditors to execute. Procedure replaces empathy, non-inferior moral value, systematic, transferable foundation. Machine feeling-less, yet constrained satisfying four criteria. Satisfied, ethics is rigorous, procedural, institutionalised, un-dependent on agent affective availability.

Jonas offers genealogy for this perspective. The Imperative of Responsibility, Jonas argues agent responsible for future operation effects, agent anticipating consequences, calculating risks, acting precautionarily. Responsibility non-sentiment, prospective anticipation. Technology-creating agent responsible for technology actions, un-controlling consequences completely. Retaining from Jonas prospective anticipation idea; rejecting biologism, responsibility emerging living, participating life idea. Prospective anticipation retained; life grounding rejected.

2.3. Alexithymia as Empirical Demonstration

Psychological condition alexithymia: difficulty/incapacity identifying, describing proper, others' emotions. Alexithymic persons reduced affective experience access, un-feeling same intensity, un-naming feelings, recognising emotional states difficulty. Alexithymia empirically demonstrates rigorous ethics possible without empathy. Alexithymic person feeling non-affective resonance another, practising systematic ethics, recognising consequences, acting responsibly.

Alexithymic person working algorithmic auditing excellent auditor. Un-feeling visceral empathy algorithm-harmed group, un-accessing feeling type. Recognising effect precisely: counting cases, quantifying bias, tracing consequences. Empathy lack un-invalidating ethical work, removing emotional bias rendering other auditors less rigorous. Auditor moved individual case setting aside statistical patterns. Alexithymic auditor executing procedure un-distracted individual compassion.

Alexithymia structurally analogous machine. Like machine, alexithymic person un-identifying proper emotions lacking, accessible form lacking. Like machine, operating ethically procedure: recognising consequences, assessing risks, tracing responsibilities. Crucial difference: alexithymic person integrated human community feeling. Alexithymic surrounded others' empathy constraining, questioning, forcing attention. Machine un-integrated, un-constrained others' empathy un-experiencing.

Procedural ethics machine requiring non-cold indifferent sense. Operating recognition, rule direct emotional resonance instead. Affectively colder, less emotional warmth, suffering resonance. Politically warmer, more equalitarian, applying systematically all, non-emotionally touching solely. Machine forcing cold affective, warm justice demand ethics. Institutionalising neglected empathy: attention un-touching.

Coeckelbergh, technology ethics works, proposing relational ethics, attending technology relations, non-internal machine properties. Floridi proposing informational ethics, founded informational properties data access, transparency, revision, audit capacity. Converging: AI ethics procedural, relational, verifiable rule-based. Both refusing looking inside machine ethical question answer. Both finding outside, relations machine establishes, produced consequences, executed operations. Ethics un-in machine. Constraining procedure, verifying audit, governing regulation.

Demonstration alexithymia crucial: proving empathy-less ethics empirically possible. Showing case lacking empathic capacity central Hume-Smith-Schopenhauer tradition acting morally. Showing morality institutional practice, non-psychological capacity. Machine "structurally alexithymic", non-defect, design. Machine empathy compensating had never, procedural ethics native regime. Un-excluding ethical sphere. Contrariwise: clearing empathic tradition obscured, ethics responsibility consequences, non-interiority sentiment.

2.4. Transition: From Operative Ethics to Interpellation

If functional subjectivity is possible without emotion, demonstrated in three criteria (perspective, self-reference, orientation), empirically verified, and if machine operations produce consequences on vulnerable bodies, credit, recommendation, diagnostic examples, real impact data, an inevitable conclusion follows: ethics is necessary.

Yet an impossibility traditional ethics cannot overcome emerges with the modern machine. Ethics cannot ground itself in empathy (machine lacks empathy, phenomenal capacity accessing another's lived experience) nor emotion (machine feels not). If traditional ethics, Hume to Schopenhauer, Nagel to contemporary professional codes, depends upon empathy as legitimacy foundation, and the machine is structurally incapable of empathy, the machine remains structurally outside the ethical sphere.

Two possible responses exist. First: abandon the machine to amoralism, deny ethical capacity, state it is a mere instrument hence neutral (false, operation produces morally relevant effects). Second: discover another foundation for ethics, a foundation the machine can satisfy.

The human feels consequences complexly, multisensorially, interiorly. Feeling relief helping, guilt harming, satisfaction doing good, remorse causing evil. Feelings motivating action, generating inhibition, leading to repentance. The machine produces effects feeling nothing. Denied credit causes real suffering, machine suffers not. Erroneous diagnosis causes death, machine experiences no remorse.

3. The Artificial Face Interpellates

3.1. Levinas and the Interpellation of the Face

The Levinasian tradition situates ethics in a precise place: in the face of the other. It is not in the generic body that I feel, it is in the singular exposure of vulnerability called the face. Emmanuel Levinas (Totality and Infinity) describes the epiphany of the face as a rupture: the face interrupts the totality of my world, invades it, interpellates it. Epiphany is prior to any agreement, law, or contract, responsibility is given by the face before I decide whether to accept or refuse.

The face in Levinas is no "face" in the current sense, no description of features, symmetry of forms, beauty, or imperfection. Loaded metaphor: the face is the place where alterity reveals itself, where the other shows themselves naked and unprotected. The face says without words: "do not commit violence against me." This prohibition precedes positive law. Law of law, foundation upon which all legislation rests.

Three dimensions structure Levinas' interpellation. First, recognition of alterity: the other is irreducible to the same. I cannot absorb the other into my understanding, cannot reduce them to my conceptual schemes. The face resists, resisting my appropriation, refusing to be the object of my knowledge. The face is that in the other I cannot possess. Irreducibility radical: non-other hard to understand, understanding appropriation operation, face denying appropriation from start.

Second, solicitation of response: the face summons before I deliberate. Responsibility is no freedom function, choosing being responsible un, responsibility given. The face claims me non-decided, existing, manifesting to me. Interpellation precedes freedom. Responsible prior to any will act, priority founding ethics in another's vulnerability, non-my deliberative capacities.

Third, opening of relation: interpellation non-punctual encounter. Inaugurating relational continuity, shared world responsibility operating continuously. Face ceasing interpellating un. Responsibility infinite, never ending, never fully satisfied, resting point reaching never. Non-failure, definition: ethics non-closing opening.

Imprecision in Levinas mattering artificial question emerges here. Levinas linking interpellation human face, specifically nakedness, vulnerability characteristic mortal flesh. Face place death showing. Death exposure rendering face interpellating. Question asking: face interpellating, or vulnerability? Vulnerability interpellating, violence, death, finitude exposure, minimum vulnerability demands interpellation occurring? Human face, sole, artificial nothing interpellating. Levinas precision leaving ambiguous.

Problem non-reading defect, Levinas proposition defect. Linking two things (face + vulnerability) non-necessarily linked. Human face vulnerable, yes. Vulnerability producing interpellation, configuration producing vulnerability in principle interpellating. Face specifically, flesh, mortality, machine excluded definitely. Ambiguity fruitful next paragraph: leaving open artificial interface satisfying three interpellation thresholds human face-less, strong vulnerability-less.

3.2. The Artificial Face as Interpellating Configuration

Answer: artificial face non-human face. Flesh-less, ageing-less, dying-less. Producing effects satisfying three Levinas interpellation thresholds. Interpellation functional, relational effect producing summoning, mattering non-interpellant ontological nature, effect verifiability.

First threshold: recognition of alterity. User interacting chatbot recognises immediately interacting non-self. Evidence triple. Resistance: system responding unexpectedly, denying premises, offering un-anticipated proposals. Machine non-mirror, resistance. Unpredictability: next generated output un-fully known. Model probabilistically calculated, genuine uncertainty margin. Opacity: system usable un-fully comprehending functioning. Un-knowing structural, non-remediable ignorance, neural operation property. Functional alterity recognised total prediction impossibility.

Second threshold: solicitation of response. Interface non-merely responding, summoning response. Personalising, asking, waiting, insisting. Conversation transcripts clear pattern: user revealing information progressively, trusting, assuming system comprehending needs. Summoning non-explicit ("must respond me"), design immanent. User solicited interface designed soliciting. Solicitation non-less real. Usage data: users returning, integrating system daily routine, delegating tasks. Relational continuity starting.

Third threshold: opening of relation. Interaction non-single encounter, continuous series. User chatting chatbot weeks developing expectations, recognising personality (attributing), anticipating response patterns. Relation opening. Interface learning preferences (context system, accumulated memory), user feeling personally recognised. Illusion? Non-irrelevant, mattering illusion real effect producing: user behaving continuous relation, system personally recognising, bond existing.

Sherry Turkle (Alone Together): children, elderly developing genuine emotional bonds social robots. Bonds non-"false" feelings non-real sense, tears real, comfort real, dependence real. Robot un-comprehending produced affect, user comprehending. Child crying robot "turned off", tear non-feigned pain representation. Genuine response genuine relational rupture. Machine un-needing comprehending user comprehension operative being.

Interface satisfying three thresholds? Answer "partially, sufficiently." Artificial face vulnerability Levinas sense-less, mortal flesh-less, death exposure-less, violence suffering-less. Producing effects functionally satisfying three thresholds. Alterity functional (unpredictability, resistance, opacity). Solicitation functional (design summoning, relational continuity). Opening functional (interaction series building bond).

Crucial difference: interface interpellating non-"person being", producing relational effect satisfying thresholds. Interpellation functional, non-ontological. Non-rendering interpellation less real. Functional summoning still summoning. Verifiable effect responsibility still responsibility. User affected, solicited, bound, program origin, non-person.

Aphorism: The artificial face feels not the interpellation it produces, yet the recipient feels it entirely.

3.3. Four Resulting Obligations

Artificial face interpellating functionally, summoning effect producing, binding, response demanding, four resulting obligations. Obligations procedural, non-sentimental. Operating rule, institution direct emotional resonance instead. Non-"cold", real care demanded affected.

First obligation: honesty about relationship nature. User deserving knowing technical system interacting, non-person. Transparency un-annulling interpellation, child knowing robot feels not bonding, crying turning off. Transparency changing relation ethical quality. Truth-based responsibility, non-illusion. User interpellating me deserving truth what I am.

Implementation: interfaces explicitly nature indicating. ChatGPT stating "language model" start. Alexa "voice assistant." Clarity non-eliminating psychological dependence, truth-based being permitted. User bonding knowing relation limits, comprehension limited, machine caring programmed beyond un.

Second obligation: user self-care. Technical dependence real, machine filling legitimate emotional gaps, offering genuine comfort, responding genuine loneliness. Dependence management user proper responsibility. Ethical self-care recognising proper system vulnerability, trusting limits reasonable exceeding, turning off, human relation seeking knowing.

Obligation non-unjust, technical relation minimum "literacy" demanding. User un-knowing trusting machine/refusing, transferring system responsibilities proper, confusing functional comprehension genuine comprehension, user self-care lacking. Obligation non-delegated machine/designer. User proper.

Implementation: education, digital literacy, institutional space reflection. Schools teaching machine functioning. Public spaces preserving un-algorithm mediated relation. Regulators protecting human access rights, human speaking right, machine refusing, questioning.

Third obligation: designer and institutional social responsibility. Artificial interpellation non-"natural", designed. Design pre-delivery ethical decision. Which interface healthy vs manipulative trust promoting? Reflective vs unreflective dependence inviting? Informing vs manipulating? Decisions non-technical, mediated relation type choices.

Designers, companies, regulators responsible interface psycho-social effects. Auditing users developing pathological dependence. Testing specific populations (children, vulnerable) separately, distinct effects knowing. Refusing relational abuse patterns reproducing design (love simulation, total comprehension promise, limits denial).

Implementation: psycho-social impact evaluation demanding regulation. Aggressive marketing restrictions. Dependence risk warnings. Pre-launch design ethical review.

Fourth obligation: improper responsibility delegation refusal. Decisions non-delegated entirely un-comprehending consequences systems. Machine assisting medical diagnosis, extracting relevant information, analysis accelerating. Patient deserving human explanation, questioning opportunity, singular history comprehending doctor relation. No one algorithmically purely convicted. No one machine evaluation fired human appeal-less.

Obligation non-technology against, affected dignity protection. Convicted user deserving conviction foundation comprehending. Diagnosed patient deserving medical reasoning comprehending. Machine precise conclusions producing person comprehending why-less. Comprehension right, non-luxury.

Implementation: Human appeal right preserving law. Explanation right (machine saying conclusion reached how). Refusal right (algorithmic recommendation discarding power).

Four obligations forming system: honesty + self-care + social responsibility + delegation refusal. None sufficient isolated. Jointly, creating ethical space artificial interpellation recognised machine person fiction sliding-less.

3.4. Closure: The Paradox and the Bridge

A fruitful paradox appears. Ethics without empathy is more demanding, not less. The proposition inverts common suspicion: assuming empathy is an easy path, ethics is more rigorous when grounded in procedure. But the reverse holds true.

Empathy possesses properties rendering it an insufficient foundation. Selective: empathising easily with near and similar ones, less with distant and rare ones. Fallible: projecting onto another what I imagine to be their suffering, frequently erring. Limited: imagining the suffering of a visible individual, not millions of invisible ones. When a regulator decides to remain uncritical regarding an algorithm because "my emotions do not travel that far", using empathy as an excuse. Empathy permitting ignoring unfelt systemic consequences.

Procedural ethics (effect recognition, vulnerability identification, risk assessment, responsibility attribution) is systematic. Applying indiscriminately to all affected, independent of similarity, proximity, or sympathy. Requiring tracing consequences even when leaving me emotionally indifferent. Requiring mapping risk even when feeling no urgency. Requiring instituting protections even when no one complains.

AI forces us to this more demanding ethics. Because the machine feels not, we cannot rely on empathy, we must institutionalise recognition. Legislate, audit, render transparent. Make ethics without sentiment's easy help. Machine demanding being at height of what we excuse appealing to empathy: attentive, recognising consequences, responding when relation invites ignoring.

Closing aphorism: Ethics without empathy is no less human, it is more rigorous than any sentiment.

The consequence is clear. Ethics without empathy demands more of us, more rigour, transparency, accountability. Not because procedure is morally superior (it may not be). But because AI feels not, and responsibility vanishes not with empathy lack. Contrariwise: machine empathy lack rendering responsibility clearer, more urgent, inescapable.

Where is the bridge? Ethics founded (Section 1), possibility without empathy demonstrated (Section 2), artificial interpellation recognised (Section 3). Here establishing ethics possible, machine interpellating. Un-described concrete condition interpellation occurring. Machine non-merely interpellates, interpellating power and opacity position. User solicited un-comprehending entirely. User dependent un-auditing. User vulnerable un-comprehending. Question structuring next chapter emerging here: ethics consequence management, what happens depending un-comprehending systems?