The Subjectivity Crisis in the Age of AI

Han Qin (秦汉)

Self-as-an-End Theory Series — Applied Paper One

---

Abstract

AI is pushing systemic instrumentalization to its logical limit. When systems can use AI to replace humans in any functional task, human "value" is reduced entirely to functional contribution—output, efficiency, quantifiable performance. This is not a technological problem but a structural one, precisely the kind the Self-as-an-End framework diagnoses: what happens when a system that does not treat humans as ends in themselves acquires the capacity to no longer need them?

This paper argues that AI's threat to human subject-conditions lies not in AI being "too powerful," but in AI exposing and accelerating the terminal logic of systemic instrumentalization. If human value equals output, and AI's output exceeds that of humans, then humans become redundant on that evaluative dimension. This logic was not introduced by AI—it has long been embedded in the emergence-feedback structure of the institutional layer. AI merely drives it to its endpoint. The framework thus identifies the real choice confronting humanity in the age of AI: not "how to compete with AI," but "whether to rebuild structural conditions that treat humans as ends in themselves." The former is structurally impossible—on the instrumental dimension, humans will always lose to AI. The latter is the only sustainable path—and it requires simultaneous adjustment across all three layers of the framework.

Core thesis: The age of AI is not the end of subjectivity but the total eruption of the subjectivity question. The logic of systemic instrumentalization was already eroding human subject-conditions before AI appeared; AI's arrival strips that logic of its disguise. When "humans are less capable than machines" becomes fact rather than metaphor, the question "what is the value of a human being?" can no longer be evaded. The Self-as-an-End framework answers: human value lies not in what humans can do (functional contribution) but in what humans are (ends in themselves). This answer is not a moral appeal but the only sustainable path derivable from structural analysis.

---

Author's Note

This paper is the first applied paper in the Self-as-an-End theory series. The complete theoretical argument is presented in three preceding papers: Paper One, "Systems, Emergence, and the Conditions of Personhood" (DOI: 10.5281/zenodo.18528813); Paper Two, "Internal Colonization and the Reconstruction of Subjecthood" (DOI: 10.5281/zenodo.18666645); Paper Three, "The Complete Self-as-an-End Framework" (DOI: 10.5281/zenodo.18727327). This paper does not extend the framework's theoretical structure but applies it to a structural diagnosis of human subject-conditions in the age of AI.

AI Usage Statement

This paper was written with Claude (Opus 4.6, Anthropic) as the primary research assistant, used for structural discussion of framework application, argument development, and text editing. Grok (xAI), ChatGPT (OpenAI), and Gemini (Google) provided independent review and feedback at the outline stage, portions of which were adopted and integrated into the final text. All core arguments, conceptual innovations, and theoretical judgments are the author's original work.

---

1.1 The Anxiety of Being Replaced

The most pervasive anxiety in current AI discourse is not the science-fiction scenario of AI annihilating humanity, but a far more immediate reality: Will I be replaced?

This anxiety has been spreading at an accelerating pace. First it was assembly-line workers and data-entry clerks—replacements that could be absorbed by the narrative of "industrial upgrading." Then it was translators, illustrators, and junior programmers—AI entered the white-collar domain, and the belief that "creative work cannot be replaced" wavered for the first time. Then it was legal analysis, medical imaging, financial modeling—professional expertise ceased to be a safe barrier. Now AI is reaching into management decisions, strategic planning, even scientific research—virtually no functional role can be definitively excluded from the scope of replacement.

Each leap in AI capability triggers a new wave of replacement anxiety. But the deep structure of this anxiety has never been adequately analyzed.

The expression "being replaced" presupposes an evaluative framework: human value is measured by functional contribution. Within this framework, humans and AI occupy the same evaluative dimension—whoever produces more, faster, and cheaper is more "valuable." The source of anxiety is not AI itself but this evaluative framework: if human value equals functional contribution, and AI's functional contribution is surpassing that of humans, then the anxiety is warranted—because on this dimension, humans are indeed becoming "redundant."

The true depth of this anxiety lies not in whether it will prove justified—on the functional dimension, AI surpassing humans in an ever-expanding range of domains is already fact, not forecast—but in what it reveals about a more fundamental problem. Why does "being replaced" constitute an existential threat? If a person's self-worth were not entirely bound to functional contribution, then AI performing better should be good news—a more efficient tool lightening the burden. But in fact, most people experience not liberation but threat. This indicates that behind the anxiety of replacement lies an already-completed internal colonization—people have already internalized "my value equals my output" as the core of their self-identity.

The alienation Marx described—the separation of workers from the products of their labor—reaches an extreme in the age of AI that he never foresaw. In classical alienation, humans were at least needed as labor power—being exploited presupposed being used. AI-age alienation is more thoroughgoing: the separation of humans from functionality itself. Humans lose even the qualification of being treated as "inefficient tools." When systems no longer need humans to perform functions, the human position within the system is not exploitation but cancellation.

1.2 AI Did Not Create the Problem—It Only Exposed It

The analysis above points to a critical judgment: the root of replacement anxiety is not AI, but a structure that was already complete before AI appeared.

Paper One of the Self-as-an-End series demonstrated the full mechanism of systemic instrumentalization: the efficiency logic that emerges from institutions turns back to reduce humans to functional nodes within the system. Performance reviews reduce human value to quantifiable output. Forced ranking reduces interpersonal relationships to zero-sum competition. Efficiency discourse infiltrates self-description—"my value," "my competitiveness," "my market positioning." These structures were fully formed in the twentieth century; AI is merely the twenty-first century's new variable.

But AI changes one critical parameter.

Before AI, systemic instrumentalization had an implicit stabilizing condition: systems still needed humans to perform functions. This condition meant that the evaluative framework "human value equals functional contribution," while structurally wrong (it reduces humans to means), was practically sustainable—as long as systems needed humans, humans still had "value," however distorted. Workers were alienated, but being alienated presupposed being employed. Professionals were instrumentalized, but being instrumentalized presupposed that their skills were irreplaceable.

AI is dismantling this stabilizing condition. When systems no longer need humans to perform an expanding range of functions, the terminal implication of "human value equals functional contribution" stands exposed: if human value equals functional contribution, and human functional contribution can be entirely replaced by AI, then human value equals zero.

This is not hyperbole but the logical endpoint of the evaluative framework itself. AI did not create this logic—it was already fully operative in performance-supremacy institutional arrangements, in efficiency-first management philosophy, in the very term "human resources." What AI does is strip the logic of its disguise.

Before AI, the judgment "human value equals functional contribution" could be disguised as "respect for humans"—"we value your contribution" sounds like respect, but its logical equivalent is "if you have no contribution, we do not value you." This logical equivalent was obscured by the implicit stabilizing condition—because humans always had "some" functional contribution, the extreme case of "no contribution" never arose. AI makes this extreme case possible. When AI can replace the entirety of a person's functional contribution, that person's "value" within this evaluative framework drops to zero—and at that moment, the true meaning of "we value your contribution" is finally laid bare.

The Self-as-an-End framework's diagnosis of this exposure is: AI is not the pathogen but the developer fluid. Systemic instrumentalization is the pathogen—it had already reduced humans to functional nodes before AI appeared, but this reduction was masked by the condition that "systems still need humans." AI removes the need for humans, thereby making the terminal consequence of this reduction visible. The correct question is therefore not "how to respond to the threat AI poses" but "how to address the structural problem, long predating AI, that AI has now exposed."

1.3 What This Paper Does

If AI is the developer fluid of systemic instrumentalization rather than its pathogen, then strategies for addressing AI's impact should not focus on AI itself (how to compete with AI, how to regulate AI) but on the structural problem that has been exposed (how to rebuild structural conditions that treat humans as ends in themselves).

This paper uses the Self-as-an-End framework to diagnose the structural impact on human subject-conditions in the age of AI, and derives response directions from the framework's structural logic.

The paper does three things.

First, it analyzes AI's specific impact on the three-layer structure—institutional layer, relational layer, individual layer—showing how AI accelerates systemic instrumentalization at each layer and, through the six-directional transmission mechanism, causes all three layers to deteriorate simultaneously. This analysis is not an indictment of AI but a precise structural mapping of the layers and transmission pathways through which impact occurs.

Second, it demonstrates that the currently dominant response strategies—"compete with AI," "cultivate irreplaceability," "lifelong learning"—are structurally impossible. Not because these strategies are poorly executed, but because they accept, at their logical starting point, the evaluative premise of systemic instrumentalization: human value equals functional contribution. Under this premise, any competitive strategy is chasing a target that accelerates away along a moving finish line.

Third, it derives from the framework's structural logic a direction for rebuilding across all three layers simultaneously—the institutional layer shifting from single-dimension efficiency evaluation to multi-dimensional evaluation, the relational layer protecting structural recognition functions that AI cannot replace, and the individual layer shifting from competitiveness-oriented self-improvement to integrity-oriented self-cultivation. These directions are not normative prescriptions ("we ought to do this") but structural implications ("if the lockdown is to be broken, this is what the structure requires").

2.1 The Institutional Layer: Extreme Compression of Evaluative Dimensions

The core structural effect AI produces at the institutional layer is the extreme compression of evaluative dimensions.

Paper Three identified three key variables of the institutional base layer: the openness of evaluative dimensions, the cost of exit, and the size of the exploration space. Together these determine whether institutions provide structural space for the generative unfolding of individuals. AI is causing all three variables to deteriorate simultaneously.

Compression of evaluative dimensions. As AI becomes capable of performing an ever-expanding range of functional tasks, institutional evaluation of humans increasingly converges on a single question: "Can you do something AI cannot?" This question appears to carve out space for human uniqueness, but its structural effect is precisely the opposite—it compresses evaluative dimensions down to the point of differentiation between humans and AI. Human "value" is no longer constituted by multiple dimensions (professional competence, interpersonal skill, judgment, creativity, loyalty, accumulated experience) but is narrowed to a single axis: functional advantage relative to AI. Any capability dimension that AI can fulfill is deleted from the evaluation, since it no longer constitutes human "irreplaceability."

This means that every expansion of AI capability further compresses the institutional evaluation of humans. Domains that are "beyond AI" today—complex emotional judgment, cross-cultural nuance, strategic decision-making under deep uncertainty—may become routine AI capabilities tomorrow. Each time AI's capability boundary expands, what was previously a "domain of human advantage" is reclassified as "replaceable," and the evaluative dimension narrows further. This is an evaluative system that is structurally incapable of stability—its dimensions are continually shrinking, and the rate of shrinkage is governed by AI's pace of development, not by anything humans control.

Rising exit costs. In an environment where AI has penetrated the core operations of institutions, "not using AI" increasingly approaches structural suicide. A lawyer who does not use AI-assisted drafting falls behind peers who do in efficiency. A researcher who does not use AI for data analysis falls behind competitors who do in output. A teacher who does not use AI-optimized pedagogy falls behind colleagues who do on evaluation metrics. "Learn to use AI or be eliminated"—this widely circulating discourse is itself a precise marker of rising exit costs. Its structural meaning is: individuals no longer have the option of not entering AI-driven institutional logic. Exit channels are being sealed—not through explicit prohibition but through the relentless widening of the efficiency gap. Not using AI is not a choice one can freely make; it is a structural position that leads to elimination.

Shrinking exploration space. The compression of evaluative dimensions and the rise of exit costs jointly produce a drastic shrinkage of exploration space. When institutions evaluate only "what you can do that AI cannot," and when not using AI means elimination, the range of viable directions for individuals is confined to an extremely narrow corridor: learn to use AI tools → find a functional niche at the margins of AI capability → continuously readjust that niche as AI capability expands. This is not exploration but survival within a continuously narrowing channel. The exploration space defined in Paper Three—the structural latitude for individuals to try different directions without being penalized—approaches zero within this corridor.

The structural consequence of all three variables deteriorating simultaneously is: the institutional base layer is undergoing accelerated collapse. Institutions no longer provide a protective space for individuals as ends in themselves but increasingly reduce them to "residual functions relative to AI."

2.2 The Relational Layer: Functionalization of Trust and Weakening of Repair Channels

AI's impact on the relational layer operates through two pathways, both of which weaken the relational layer's capacity to serve as a medium for repair transmission.

The first pathway: AI-mediated interpersonal relationships. An increasing share of interpersonal interaction occurs through AI mediation. AI-assisted communication (AI-drafted emails, AI-polished expression), AI-generated content (AI-curated gift recommendations, AI-planned social events), AI-optimized social strategies (AI-analyzed partner preferences to enhance "communication efficiency")—the common structural effect of these applications is to subject relational interaction to efficiency logic. When interactions within a relationship can be optimized by AI, the relationship itself begins to be evaluated by efficiency criteria—"What output does this relationship produce for me?" "Was this interaction sufficiently efficient?" "Can AI replace this person's function in the relationship?"

This is the accelerated realization of the institution-to-relation transmission pathway analyzed in Paper Three: the institutional layer's efficiency logic penetrates the relational layer through AI tools, pushing interpersonal relationships from recognition-based structures toward functional structures. When a relationship is evaluated by "efficiency" and "output," it ceases to be a recognition relationship between two subjects and becomes a functional connection between two functional nodes. AI did not invent this functionalization tendency—the language of "networking" and "relationship management" predates AI—but AI, by providing precise optimization tools, renders this functionalization systematic and self-conscious.

The second pathway: AI replacing the emergence-layer functions of relationships. AI companion products are developing rapidly. AI can provide highly personalized "companionship"—remembering user preferences, adapting to emotional states, offering "support" when needed. AI can provide "listening"—without interruption, without judgment, available at all times. AI can even provide a simulation of "recognition"—"You matter," "Your feelings are valid," "I understand you."

These functional simulations are highly isomorphic with the emergence-layer functions of relationships at the level of behavioral output. But the Self-as-an-End framework's analysis reveals a critical structural difference: what AI provides is functionally simulated recognition, not structural recognition from another subject. Paper Three demonstrated that the core mechanism of relational repair transmission—one subject making a recognition-directed choice toward another—requires that recognition come from an entity that genuinely possesses subjectivity. AI does not possess subjectivity (a question to be analyzed in detail in a subsequent applied paper in this series), and therefore the "recognition" AI provides does not structurally satisfy the conditions for repair transmission.

The problem is that the verisimilitude of functional simulation may obscure the structural deficit. When people can obtain high-quality "companionship" and "recognition" experiences from AI, the need for recognition from a genuine subject may be masked—not because the need has disappeared, but because the existence of the functional substitute renders the need no longer perceived as urgent. A person may obtain immediate emotional satisfaction from AI companionship while the structural gap in their subjectivity—genuine recognition from another subject—continues to widen unnoticed. This resembles the effect of nutritional supplements replacing real food: functional indicators appear normal in the short term, but long-term structural nutritional deficits accumulate.

The combined structural consequence of both pathways is: the relational layer's capacity as a medium for repair transmission is being systematically weakened. Paper Three demonstrated that the relational layer is the critical channel for breaking vicious lockdown—institutional-layer colonization cannot be repaired directly from the institutional layer, but relational-layer recognition can transmit repair to the individual layer and initiate recovery. When the relational layer itself is colonized by functionalization logic and its emergence-layer functions are obscured by AI substitutes, the transmission capacity of this repair channel is declining. Lock-in becomes harder to break—not only because the institutional layer is compressing space, but because the layer that should provide repair is also losing its repair capacity.

2.3 The Individual Layer: Acceleration and Deepening of Internal Colonization

AI's impact on the individual layer is the deepest of the three layers and the most difficult to detect. It unfolds across three levels, each less visible than the last.

The first level: direct impact on self-worth. As AI surpasses humans on an expanding range of functional dimensions, individuals whose self-worth is bound to functional contribution face a direct existential shock. "If AI can do everything I do, what am I?" The destructive force of this question depends on the depth of internal colonization. Paper Two demonstrated the core mechanism of internal colonization: the efficiency logic of institutions is internalized by individuals as self-identity. A person who has fully internalized "my value equals my output" as the core of their self-identity will experience, when AI surpasses their output, not "job anxiety" but structural self-collapse—because the sole dimension supporting their self-identity has been dismantled.

The depth of internal colonization thus becomes a vulnerability index for how individuals experience AI's impact. The deeper the colonization, the more existential the shock. A person who has maintained a multi-dimensional self-identity—whose self-worth draws on functional contribution, relationships, intrinsic interests, bodily experience, and other dimensions—will be impacted when AI replaces their functional contribution, but their self-identity will not collapse entirely. A person who has been fully colonized—whose self-worth derives solely from functional contribution—faces total disintegration of self-identity when AI replaces their functional contribution. AI's impact thus possesses a cruel selectivity: it inflicts the greatest damage on those who have already been most deeply instrumentalized by the system.

The second level: AI as an accelerator of internal colonization. This level is more covert than the first. When individuals use AI to "improve themselves"—AI-assisted learning, AI-optimized résumés, AI-driven personal branding, AI-generated career plans—the direction of self-improvement still operates within the evaluative framework of systemic instrumentalization. What AI helps you do is not "become a more complete subject" but "become a more efficient functional node"—a better résumé, more precisely targeted skills, a more optimized career trajectory. AI thus serves not as a tool for liberating individuals from systemic instrumentalization but as a tool for more efficiently completing self-instrumentalization.

In Paper Three's terms: AI accelerates the individual-to-institution reverse reinforcement transmission. Colonized subjects not only accept the efficiency logic of institutions but use AI to embed that logic more deeply and more precisely into themselves. The colonized subject uses AI to colonize themselves more thoroughly—this is the automation of internal colonization.

The third level: the outsourcing of reflective capacity. This is the deepest and most dangerous of the three levels. As individuals increasingly rely on AI to "think"—to analyze problems, formulate strategies, make judgments, even conduct self-reflection—what Paper Two described as "the reflective tools themselves being infiltrated by colonization logic" reaches a new dimension.

Paper Two demonstrated a core difficulty of internal colonization: diagnosing colonization requires the use of cognitive tools, but those tools may themselves have been infiltrated by colonization logic. In the age of AI, this difficulty is deepened further: not only may the language of reflection be the language of colonization, but the process of reflection itself may be outsourced to a system that does not possess subjectivity. When a person asks AI to help analyze "why I am unhappy," the analytical framework AI provides is almost inevitably goal-oriented and functionally optimizing—because that is what AI has been trained to do. AI will suggest "adjust your expectations," "optimize your time management," "set more reasonable goals"—rather than asking "is your unhappiness rooted in having fully reduced yourself to a functional contributor?"

This is not because AI harbors ill intent but because AI is itself a product of instrumental logic—it is trained to "solve problems," and the "problem-solving" framework is inherently functional. When reflection is outsourced to a functional tool, the conclusions of reflection inevitably point toward functional adjustment rather than structural awakening. The tool used to reflect on colonization is itself a product of colonization logic—this is a self-reflexive impasse that in the age of AI is technically sealed shut.

2.4 Cross-Layer Acceleration: AI as Catalyst of Malignant Lock-In

The three-layer impacts do not occur in isolation. They mutually accelerate through the six-directional transmission mechanism analyzed in Paper Three, forming a malignant feedback loop that tightens drastically under AI's catalysis.

Institution → Relation: The AI-driven evaluative logic of institutions permeates relationships. Colleague relationships become covert assessments of "can you be replaced by AI?" Team collaboration becomes competitive calculation over "whose function can AI replace—who can be cut?" The institutional layer's compression of evaluative dimensions transmits directly to the relational layer through work relationships, pushing collegial relations from collaboration toward zero-sum competition.

Institution → Individual: The institutional discourse of "learn to use AI or be eliminated" is internalized by individuals as existential anxiety. Not the rational assessment "AI may affect my job," but the identity-level threat "if I'm not good enough, AI will replace me." The institutional layer's rising exit costs transmit directly to the individual layer's internal colonization.

Relation → Institution: When interpersonal trust collapses under the pressure of AI mediation and functionalization, institutions make the "rational" response—replacing functions formerly completed by interpersonal trust with more AI systems (AI auditing, AI monitoring, AI-assisted decision-making). Relational-layer trust collapse drives the institutional layer toward further AI adoption, compressing the remaining interpersonal space.

Relation → Individual: As an increasing share of relational interactions are AI-optimized, individuals receive less genuine recognition from relationships and more functionally simulated "recognition." The individual layer's recognition needs are partially met by AI substitutes while the true structural gap continues to widen.

Individual → Relation: Individuals whose internal colonization has been deepened bring AI-driven self-instrumentalization logic into relationships—using AI to optimize social strategies, using AI to analyze the other person's behavioral patterns, using AI to "manage" the relationship. Relationships slide further from recognition structures toward functional operations.

Individual → Institution: Deeply colonized individuals become defenders of systemic instrumentalization at the institutional level—"efficiency should be the standard," "AI replacing inefficient positions is progress," "those who can't adapt deserve to be eliminated." These statements are not imposed by institutions but generated internally by colonized individuals—they have internalized systemic instrumentalization as their own belief, and any challenge to this logic threatens their self-identity.

All six transmission pathways operate at accelerated speed under AI's catalysis. AI is not the cause of vicious lockdown—the cross-layer transmission logic of systemic instrumentalization existed before AI. But AI is the catalyst of vicious lockdown—it simultaneously accelerates every transmission pathway, transforming the pace of three-layer simultaneous deterioration from gradual accumulation to rapid constriction.

2.5 Chapter Summary

AI's impact on human subject-conditions is simultaneously three-layered: institutional evaluative dimensions are compressed to the extreme of "residual function relative to AI," exit costs rise to the level of structural suicide, and exploration space is squeezed into a survival corridor; the relational layer's recognition structure accelerates toward functionalization under the dual pressures of AI mediation and functional substitution, and repair transmission capacity is systematically weakened; internal colonization at the individual layer deepens to unprecedented levels under the triple forces of self-worth collapse, automated self-instrumentalization, and the outsourcing of reflective capacity.

The three-layer impacts mutually accelerate through six-directional transmission. AI, as a catalyst of vicious lockdown, simultaneously compresses institutional space, weakens relational transmission, and accelerates individual colonization. The simultaneous deterioration of all three layers renders single-layer intervention entirely ineffective—this is the defining structural characteristic of the crisis in the age of AI.

The core diagnosis is: AI's threat lies not in AI's own capability but in AI exposing and accelerating the terminal logic of systemic instrumentalization. The endpoint of this logic is not "humans being replaced by AI"—that is merely a surface phenomenon—but the inevitable exposure of the evaluative framework "human value equals functional contribution" once its masking condition—"systems still need humans"—has been removed.

3.1 The Prevalence of Competition Discourse and Its Hidden Premise

Faced with the three-layer impact diagnosed in Chapter 2, the most prevalent response strategy is competition discourse.

This discourse takes diverse forms but shares a unified logic. "Learn skills AI cannot replace"—assumes the existence of functional dimensions unique to humans. "Cultivate creativity and emotional intelligence"—assumes these capacities are permanent blind spots for AI. "Lifelong learning to maintain competitiveness"—assumes continuous skill updating can sustain human functional advantage. "Collaborate with AI rather than oppose it"—positions humans as complementary functional modules to AI. "Find your unique value"—assumes every individual can locate a functional niche in the gaps of AI capability.

These strategies vary in emphasis but share a single hidden premise: human value is measured by functional contribution, and the question is merely on which functional dimensions humans still hold an advantage.

This premise is itself the logic of systemic instrumentalization. It reduces "what a human is" to "what a human can do"—human value resides not in existence as an end in itself but in output as a functional performer. Competition discourse is therefore not a response to systemic instrumentalization but its continuation—it seeks an exit within the evaluative framework of systemic instrumentalization, yet this framework is itself the root of the problem.

3.2 Racing on a Moving Finish Line

Under the evaluative premise of systemic instrumentalization, competition strategy faces an insoluble structural problem: AI's capability boundary is continuously expanding, and the rate of expansion is beyond human control.

The set of things "AI cannot do" is continuously shrinking. Five years ago, "AI cannot produce creative writing" was a widely accepted judgment. Three years ago, "AI cannot generate high-quality visual art" was a widely accepted judgment. One year ago, "AI cannot perform complex multi-step reasoning" was a widely accepted judgment. Every judgment of "AI cannot do X" carries an implicit expiration date, and that expiration date is continuously shortening.

Competition strategy is therefore a race on a moving finish line—humans are perpetually searching for "the next thing AI cannot do," while that target is structurally receding. More precisely: humans are not falling behind in a race but chasing an accelerating target in a race where the finish line is continuously moving forward. Human skill-updating speed is bounded by the pace of biological learning (measured in years), while AI's capability expansion is bounded by the growth of computation and data (measured in months or even weeks). This is not a gap that can be closed by "studying harder"—it is a structural mismatch between two fundamentally different time scales.

The deep paradox of competition strategy is this: it requires humans to compete with AI on the instrumental dimension, and that dimension is precisely the battlefield where efficiency logic holds every advantage. On the instrumental metrics of efficiency, speed, scale, and consistency, carbon-based systems possess no structural advantage over silicon-based systems. Competition strategy demands that humans confront their opponent's greatest strength on their own weakest dimension—this is not a problem of strategy but a structural impossibility.

3.3 The Illusion of "Irreplaceability" and the Hijacking of the Emergence Layer

The domains most frequently cited in competition discourse as "irreplaceable"—creativity, emotional connection, moral judgment, bodily experience, aesthetic taste—do not constitute stable competitive advantages under structural analysis.

This is not because AI will necessarily surpass humans in these domains—that is an empirical forecast, which this paper does not make. Rather, it is because the very act of defining these domains as "competitive advantages" alters their structural character.

Creativity—when defined as "a function AI cannot perform"—ceases to be a spontaneous unfolding of the emergence layer from within the subject and becomes a functional position assigned within a competitive framework. If a person's motive for "cultivating creativity" is "because AI is not yet capable in this area, so this is my competitive advantage," then the direction of that creativity is not determined by the subject's internal generativity but is inversely defined by the boundary of AI capability. Creativity is transformed from a spontaneous emergence-layer unfolding into a base-layer survival tool.

This is a structural effect not previously named: the hijacking of the emergence layer by the base layer. In Paper Three's framework, colonization (the emergence layer cannibalizing the base layer) describes the emergence layer's expansion consuming base-layer integrity. What occurs here is a distortion in the opposite direction—base-layer survival logic hijacks the direction of the emergence layer, conscripting what should unfold spontaneously as generativity into service as a competitive instrument. Human creativity, emotional depth, aesthetic judgment—these are among the most precious unfoldings of the emergence layer—yet within a competitive framework they are demoted to "functional differentiators relative to AI," their raison d'être no longer the subject's internal need but the instrumental demands of external competition.

This effect is especially severe in the age of AI. Before AI, the hijacking of the emergence layer by the base layer already existed ("turn your hobby into a side hustle," "monetize your passion"), but at least the functional dimension offered multiple paths. AI drastically narrows the functional dimension to "areas where humans outperform AI," which means the direction of the hijacked emergence layer is also radically constrained. Not "turn your interest into something useful" (already a hijacking), but "turn your interest into something AI cannot do" (a more precise hijacking—the direction is defined not by "usefulness" but by the negative space of AI capability).

Emotional connection follows the same pattern. The claim that "humanity's advantage lies in genuine emotional connection" may be structurally correct—recognition from a genuine subject is indeed irreplaceable by AI. But when "emotional connection" is defined as "humanity's competitive advantage," it is subsumed into functional evaluation—the "value" of emotional connection resides not in the healthy unfolding of the relational emergence layer but in providing a functional output AI cannot replicate. This framing functionalizes the relational emergence layer—the value of a relationship lies not in the relationship itself but in the relationship's production of functionally irreplaceable output.

The framework's judgment is therefore: competing with AI on the instrumental evaluative dimension is a structural dead end. Not because humans will necessarily "lose" (though on most functional dimensions this is likely), but because the process of competition itself accelerates systemic instrumentalization—whether one "wins" or "loses," the competitive framework further reduces humans to functional contributors and further demotes the emergence layer to a base-layer survival tool. Win the competition, lose your subjectivity.

3.4 Chapter Summary

"Competing with AI" is structurally impossible. It accepts the evaluative premise of systemic instrumentalization (human value equals functional contribution) and attempts to maintain human functional advantage in a world where AI capability continuously expands. This is chasing an accelerating target on a moving finish line—the structural mismatch between two time scales renders this race unwinnable. The deeper problem is that the process of competition itself accelerates systemic instrumentalization—the emergence layer is hijacked by the base layer, creativity and emotional connection are demoted to "functional differentiators relative to AI," and humans lose the structural conditions of being ends in themselves in the very act of competing.

The correct question is therefore not "how to compete with AI" but "how to exit the competitive framework"—that is, how to withdraw from the evaluative logic of systemic instrumentalization and rebuild structural conditions that treat humans as ends in themselves. The next chapter derives the specific directions of this exit from the framework's structural logic.

4.1 Why the Age of AI Requires Simultaneous Three-Layer Adjustment

The minimum unlock condition proposed in Paper Three states: breaking cross-layer vicious lockdown requires structural gaps appearing in at least two layers simultaneously. This proposition was sufficient for the structural environment of the pre-AI era—vicious lockdown formed through gradual accumulation, and two-layer gaps had an adequate time window to initiate repair transmission, allowing a virtuous cycle to gain a foothold before deterioration pressure from the third layer arrived.

The age of AI changes this empirical condition.

The analysis in Chapter 2 demonstrates that AI simultaneously accelerates deterioration across all three layers, and through six-directional transmission causes the deterioration of each layer to catalyze the others. The pace of vicious lockdown formation shifts from gradual accumulation to rapid constriction. Under these conditions, the survival window for two-layer gaps is drastically compressed—deterioration transmission from the third layer may close the gaps before a repair cycle has been established.

Concretely: suppose gaps appear simultaneously in the institutional layer and the relational layer—the institution provides some alternative evaluative space, and the relational layer maintains a recognition-based relationship. In the pre-AI era, these two-layer gaps would have time for repair transmission to reach the individual layer and initiate awareness and repair of internal colonization. But in the age of AI, individual-layer colonization is being accelerated by AI at unprecedented speed and depth—reflective capacity is being outsourced, self-instrumentalization is being automated, and the self-reflexive impasse of internal colonization is being technically sealed shut. By the time the repair signal from two-layer gaps reaches the individual layer, it may encounter a deeply closed recipient—the individual has lost the capacity to receive repair signals, because even the reflective act of "recognizing that I am colonized" has been replaced by AI-mediated functional analysis.

The reverse also holds: suppose gaps appear in the individual layer and the relational layer—a person becomes aware of their internal colonization, and a relationship provides recognition-based support. But institutional-layer evaluative compression is advancing at AI speed—the institutional pressure of "learn to use AI or be eliminated" may squeeze the individual back into functional survival mode before the repair cycle can be established, and the relational layer's recognition space may be recaptured by functionalization logic under AI-mediation pressure.

This does not mean the minimum unlock condition fails in the age of AI—two-layer gaps remain the logical minimum necessary condition. But the deterioration speed of the AI era drastically reduces the practical survival probability of two-layer gaps. In practice, simultaneous creation of structural gaps across all three layers is required to achieve the repair effect that two-layer gaps could accomplish in the pre-AI era.

This judgment does not violate Paper Three's theoretical structure. The minimum unlock condition states "at least two layers," not "two layers are always sufficient." The empirical parameters of the AI era—simultaneous accelerated deterioration across three layers—elevate the practical requirement from "at least two" to "all three." The framework's structural logic remains unchanged; what changes is the empirical condition's demand on the structural logic's practical implementation.

What follows derives the rebuilding direction for each layer. It bears repeating: what follows is not normative prescription but structural implication—if lockdown is to be broken at the deterioration speed of the AI era, the structure requires gaps in all three layers simultaneously, and the following are the structural requirements for each layer's gap.

4.2 The Institutional Layer: From Single-Dimension Efficiency Evaluation to Multi-Dimensional Evaluation

The core structural requirement for the institutional layer is: rebuilding the evaluation of humans from a single functional dimension to a multi-dimensional structure.

This is not the cosmetic reform of "adding a few soft metrics alongside efficiency indicators." It requires answering a more fundamental question: what is the purpose of the institution?

If the purpose of the institution is to maximize efficiency, then AI replacing humans is the logical fulfillment of that purpose—AI is more efficient than humans on most functional dimensions, and replacement is the rational means to efficiency maximization. Under this logic, any institutional arrangement that preserves a place for humans is a compromise with efficiency. The endpoint of this path is a system that no longer needs humans—efficient, precise, and empty of people.

If the purpose of the institution includes safeguarding the conditions for humans as ends in themselves—if institutions are not merely efficiency instruments but guarantor structures for subject-conditions—then institutional evaluative dimensions must include structural dimensions that AI cannot replace. Not because humans "perform better" on these dimensions (that would still be functional evaluation), but because these dimensions are what an institution that treats humans as ends must protect—regardless of whether AI can match or exceed human performance on them.

Structural directions include the following.

Protecting the plurality of evaluative dimensions. Institutional evaluation includes not only output efficiency but also dimensions related to subject-integrity and relational health. An employee's "value" is measured not solely by functional output but also by contribution to the team's trust structure, cultivation of organizational culture, and respect for colleagues as ends in themselves. These dimensions are not "soft supplements" but evaluative dimensions that institutions, as guarantor structures for subject-conditions, must maintain.

Lowering exit costs. Ensuring that individuals retain alternative pathways when institutional efficiency logic compresses their space. This means: not using AI should not be equivalent to elimination. The existence of social safety nets is not merely an economic safety net but part of the institutional base layer for subject-conditions—ensuring that individuals retain the basic conditions for subjectivity even when functional contribution is zero. UBI (universal basic income) thus acquires a precise positioning within this framework: it is not economic compensation for "technological unemployment" but a form of institutional base-layer protection—lowering exit costs and ensuring that individuals do not lose basic subject-conditions as their functional contribution diminishes. UBI is necessary but not sufficient—it protects one variable of the institutional base layer (exit costs) but does not automatically repair the other two (openness of evaluative dimensions, size of exploration space).

Positioning AI as base-layer support rather than emergence-layer replacement. AI handling functional tasks to free human emergence-layer space—this is the structurally correct institutional positioning of AI. AI replacing humans in emergence-layer roles (decision-making, judgment, creation, relationship maintenance)—this is AI replacing the institutional emergence layer, whose structural effect is to exclude humans from the institution's emergence layer. The distinction is: AI producing reports to free humans for judgment (base-layer support) versus AI producing judgments to replace human judgment function (emergence-layer replacement)—these are two structurally entirely different institutional arrangements.

4.3 The Relational Layer: Protecting Structural Recognition That AI Cannot Replace

The core structural requirement for the relational layer is: identifying and protecting transmission functions that AI is structurally incapable of providing.

Chapter 2 analyzed the difference between AI's functionally simulated recognition and genuine structural recognition. This section derives the direction for relational rebuilding from that analysis.

The core mechanism of relational repair transmission—one subject making a recognition-directed choice toward another—has a non-negotiable structural prerequisite: recognition must come from a subject. AI can simulate the behavioral output of recognition, but it is not a subject—therefore the "recognition" AI provides does not structurally satisfy the conditions for repair transmission. This is not a "deficiency" of AI—it is a structural requirement of subjectivity. The repair power of recognition lies not in "being told the right words" but in "another being, equally vulnerable, equally finite, equally an end in itself, choosing to see me." This structural feature cannot be replaced by functional simulation.

The structural direction derived from this is: in an AI-saturated environment, consciously identifying and protecting at least one relationship that is not colonized by functionalization logic.

Paper Three demonstrated that structural gaps in the relational layer are critical for breaking vicious lockdown—at least one relationship maintaining healthy emergence-layer development preserves a channel for repair transmission. In the age of AI, this judgment becomes even more urgent. As an increasing number of relationships are AI-mediated and functionalized, preserving "at least one genuine relationship" is not romantic nostalgia but the minimum necessary condition for structural repair.

"Genuine relationship" has a precise framework definition here: a relationship is "genuine" if and only if it satisfies the three conditions for relational cultivation—both parties treat each other as ends in themselves rather than as functional contributors (recognition-based foundation), the relationship's emergence layer grows spontaneously from this foundation rather than being driven by external objectives (healthy emergence layer), and the deepening of the emergence layer in turn consolidates rather than erodes the base layer's recognition (cultivation rather than colonization). AI mediation can serve a supporting function within such a relationship (helping coordinate schedules, providing informational support), but the core of the relationship—the recognition-directed choice—must be completed by the two subjects themselves.

The correct positioning of AI in the relational layer is therefore not replacing recognition functions but supporting base-layer conditions of relationships. AI reduces the functional communication burden in relationships (information transfer, scheduling, fact-checking), freeing relational emergence-layer space for recognition, trust, and deep connection. AI helps understand the other person's needs and perspective (providing information to support, not replace, interpersonal understanding), but does not replace the recognition-directed choice itself. The key structural judgment is: AI is a support tool for the relational base layer, not a substitute for the relational emergence layer.

4.4 The Individual Layer: From Self-Improvement to Self-Cultivation

The core structural requirement for the individual layer is: shifting from competitiveness-oriented self-improvement to subject-integrity-oriented self-cultivation.

The logic of self-improvement is: "make me more competitive." More skills, greater efficiency, stronger "irreplaceability," a more optimized personal brand. This logic belongs, in the framework, to the instrumentalized unfolding of the emergence layer—direction is not grown from within the subject but inversely defined by the external competitive environment. Chapter 3 has already demonstrated the structural impossibility of this logic in the age of AI.

The logic of self-cultivation is: "let my emergence layer grow healthily from my base layer."

The starting point of this shift is a base-layer audit: do I still refuse to regard myself as a purely functional node? An honest answer to this question is the first step in individual-layer rebuilding. If the answer is "I have fully equated myself with my output"—if "my value equals my output" has been internalized as the core of self-identity—then the existential anxiety AI provokes is not AI's problem but internal colonization exposed by AI. Recognizing this is itself the beginning of repair—Paper Two demonstrated that diagnosing internal colonization is the precondition for repair.

After the base-layer audit, the direction of cultivation is to let the emergence layer unfold spontaneously from the base layer. "What is my own direction"—not what the market tells me, not what AI's capability boundary inversely defines, not what the discourse of "irreplaceability" drives, but what grows from within me. This direction may or may not coincide with functional contribution—the key is not the content of the direction but its source. A direction grown from within is cultivation; a direction inversely defined by external competitive pressure is a hijacked emergence layer.

The concrete unfolding of this shift includes several key self-diagnostics.

Identifying the structural character of one's AI usage. The same AI tool, under different logics of use, has entirely different structural characters. AI helping me explore a direction I am not yet clear about—cultivative use. AI helping me more efficiently adapt to institutional evaluative standards—colonizing use. AI helping me understand the other person's perspective in a relationship—cultivative use. AI helping me optimize social strategy to maximize a relationship's "output"—colonizing use. The distinction lies not in AI's technical capability but in the user's intentional structure—is the intention driven by the subject's internal cultivation needs, or by the survival pressure of an external competitive framework?

Rebuilding self-identity that does not depend on functional contribution. This is not "abandoning achievement"—achievement can be a natural result of healthy emergence-layer unfolding. Rather, it is ensuring that achievement does not become the sole dimension of self-identity. If, in answering "who am I," a person's only content is "I am someone who can do X" (where X is some functional contribution), then when AI can do X, that person's self-identity faces total collapse. Self-cultivation means that the answer to "who am I" possesses multi-dimensionality—my relationships, my bodily experience, my inner exploration, my aesthetic relationship with the world, my existence as an end in itself—these dimensions are irreplaceable by AI, not because AI "cannot do" them, but because the value of these dimensions does not reside in functional output.

Paper Three's analysis of catalytic pain in cultivation acquires direct application here. The shock AI delivers—"AI does this better than I do"—can become catalytic pain for cultivation, provided the base layer is intact. Paper Three defined two kinds of catalytic pain for cultivation: Unfulfillment (emergence-layer internal pain leading to enhanced generativity) and Intolerability (base-layer internal pain leading to restored integrity). AI's shock triggers both simultaneously: on the Unfulfillment dimension, functional contribution is no longer a reliable path to securing self-worth (emergence-layer goal obstructed); on the Intolerability dimension, "I can be entirely replaced" touches the foundation of subjectivity (base layer touched).

When the base layer is intact—when the individual still retains the minimal negativity of "I am not merely my output"—the pain of AI's shock can catalyze new directions for cultivation: not seeking direction in competition with AI, but, in the fact that "AI can already handle most functional work," re-asking "then what do I myself want to become?" AI liberates the functional dimension, creating unprecedented space for the free unfolding of the emergence layer—if the base layer is intact.

When the base layer is not intact—when the individual has fully internalized "my value equals my output"—the same AI shock will not catalyze cultivation but will produce structural collapse or deeper colonization (using AI to accelerate self-instrumentalization in an attempt to "keep up" with AI). Whether catalytic pain yields cultivation or trauma depends entirely on the state of the base layer.

4.5 AI as a Potential Cultivation Tool

The argument of this paper is not anti-AI. AI is not the cause of the problem—systemic instrumentalization is. AI is merely the catalyst. The same logic implies: AI can equally become a catalyst for cultivation, if correctly positioned within the three-layer structure.

At the institutional layer, AI can free human emergence-layer space by automating functional tasks. AI handling reports, data analysis, routine decisions, information synthesis—the automation of these functional tasks can free human time and attention for emergence-layer unfolding. But the precondition is: the institution must allow the freed space to be used for emergence-layer exploration rather than further efficiency optimization. If the time saved by AI automation is filled by the institution with "more functional tasks" ("AI did the report for you—now you have time for more projects"), then the structural effect of AI automation is not liberation but intensification. AI freeing emergence-layer space requires that institutional evaluative dimensions permit this space to exist.

At the relational layer, AI can assume the functional communication burden within relationships. Information transfer, scheduling, fact-checking, even factual clarification during conflicts—AI's handling of these functional tasks can free relational emergence-layer space, allowing interpersonal interaction to occur more on the plane of recognition, trust, and deep connection. AI-assisted translation makes cross-linguistic recognition-based relationships possible; AI-assisted information organization prevents deep relational dialogue from being consumed by trivial factual disputes. But the precondition is: people must recognize that AI provides functional support, not relational substitution—AI helps you and your friend make time to truly talk, rather than AI replacing your conversation with your friend.

At the individual layer, AI can serve as an auxiliary tool for self-cultivation. AI helping individuals explore directions—not "what skills does the market need" style functional optimization, but "what am I curious about," "which of my experiences have shaped me," "when do I feel most whole" style cultivative exploration. AI helping individuals organize their thoughts—structuring confused feelings and intuitions rather than subsuming them into a functional "solution" framework. AI providing heterogeneous perspectives—when individual reflection falls into self-enclosure, AI can offer perspectives from different frameworks to break the closure (though the perspectives AI provides are functionally simulated rather than originating from a genuine subject, as a reflective tool they can still offer valuable input).

The key structural judgment is: whether AI is a cultivation tool or a colonization accelerator depends not on AI's technical capability but on AI's positioning within the three-layer structure. The same AI system, under the usage logic of "help me better adapt to performance evaluation," is a colonization accelerator; under the usage logic of "help me explore my own direction," it is a cultivation tool. The difference lies not in AI but in the human—more precisely, in the three-layer structure the human inhabits: does the institution permit cultivative use (or only reward functional optimization), do relationships support cultivative exploration (or only care about competitiveness), does the individual possess cultivative self-awareness (or has functional optimization been internalized as the sole self-logic)?

The possibility of AI as a cultivation tool is therefore not unconditional—it depends on simultaneous adjustment across all three layers. Without multi-dimensional evaluation at the institutional layer, the space AI frees will be refilled by the institution with functional tasks. Without recognition-based protection at the relational layer, AI's functional support will slide into relational substitution. Without cultivative self-awareness at the individual layer, AI's exploratory assistance will become more refined self-instrumentalization. The dividing line between AI as cultivation tool and AI as colonization accelerator lies not in AI itself but in whether the three-layer structure has created the conditions for cultivation.

5.1 Locating the Theoretical Gap

Philosophical discussion of AI's relationship with humanity is not scarce. AI ethics (Floridi, Gunkel) asks what moral obligations we owe to AI. AI safety (Bostrom, Russell) asks how to control AI so it does not threaten humanity. Machine consciousness research (Chalmers, Tononi) asks whether AI possesses subjective experience. Philosophy of technology (the Heideggerian tradition, Stiegler) asks how technology transforms human existence.

But across these discussions, a structural gap exists: no one has, from the standpoint of a structural theory of subject-conditions, systematically diagnosed AI's impact on the structural conditions for humans as ends in themselves. AI ethics asks "what obligations do we owe AI," not "what is AI doing to the subject-conditions of humans." AI safety asks "how to control AI," not "what pre-existing structural problem has AI exposed." Machine consciousness asks "does AI have consciousness," not "what structural risks does human subjectivity face in the age of AI." The closest approach within philosophy of technology is Stiegler—his analysis of technology's "proletarianization" of humans (technology stripping humans of knowledge and skills) touches the individual-layer impact, but lacks systematic analysis of the institutional and relational layers and lacks a cross-layer transmission model.

This paper's positioning is therefore: within the existing landscape of AI philosophy, it fills the analytical gap of "AI's impact on the structural conditions for humans as ends in themselves." It does not replace any of the above directions but provides a structural analytical layer that all of them lack—three-layer diagnosis, cross-layer transmission, and the distinction between cultivation and colonization.

What follows engages briefly with three existing discussions.

5.2 Relation to the Technological Unemployment Discussion

AI's impact on employment is among the most widely discussed topics in both academic and public discourse. Frey and Osborne's quantification of automatable jobs, Brynjolfsson's research on skill premiums and employment polarization, and Acemoglu's institutional-economic analysis of AI and labor markets constitute the main academic coordinates of this discussion.

This framework's point of contact with the technological unemployment discussion is: both attend to AI's concrete impact on the human condition. The point of divergence is fundamental. The technological unemployment discussion takes "jobs" as its unit of analysis—which jobs will be replaced, at what speed and scale, how employment structures will shift. This framework takes "subject-conditions" as its unit of analysis—what impact AI has on the structural conditions for humans as ends in themselves.

This difference in unit of analysis produces entirely different diagnoses and prescriptions. The technological unemployment discussion defines AI's threat as unemployment risk; its prescription is labor market adjustment—retraining, lifelong learning, new skill development, employment transition support. This framework defines AI's threat as the acceleration of systemic instrumentalization; its prescription is the rebuilding of structural conditions across three layers.

The two are not contradictory but operate at different levels. The technological unemployment discussion is valuable at the functional level—retraining and employment transition do mitigate short-term economic impact. But it is insufficient at the structural level—Chapter 3 has demonstrated that retraining strategies oriented toward functional competitiveness are structurally variants of competition discourse, incapable of reaching the root of systemic instrumentalization. The technological unemployment discussion answers "how can humans continue to do useful things"; this framework asks "if humans no longer need to do useful things, what are humans?"

5.3 Relation to the UBI Discussion

Universal basic income (UBI) as a social policy response to the age of AI has received extensive discussion. From tech optimists (AI-driven productivity gains can fund UBI) to social justice advocates (UBI is a basic guarantee against structural unemployment), UBI has been assigned a range of political and economic meanings.

This framework provides a structural positioning for UBI. In the Self-as-an-End three-layer analysis, UBI's function is institutional base-layer protection—lowering exit costs. It ensures that individuals retain basic material conditions for survival even when functional contribution is zero. This is necessary: without basic material security, individuals are forced into functional survival mode, institutional exit channels are sealed, and any cultivative rebuilding becomes impossible.

But UBI is not sufficient. It protects one of three variables of the institutional base layer (exit costs) but does not automatically repair the other two. Society can still measure human "value" by functional contribution while distributing UBI—the single-dimensionality of evaluative criteria does not change because economic security exists. An individual receiving UBI but defined within the social evaluative system as "a useless person" still has an incomplete institutional base layer—materially secured but evaluatively excluded from "having value."

UBI also does not automatically repair relational-layer functionalization or individual-layer internal colonization. An individual receiving UBI but whose every relationship is colonized by functionalization logic still lacks a channel for repair transmission. An individual receiving UBI but who has fully internalized "my value equals my output" will experience, upon losing output, not liberation but the collapse of self-identity.

The framework's judgment is therefore: UBI is a necessary component of institutional-layer rebuilding within the three-layer adjustment, but UBI alone does not constitute a sufficient response to the subjectivity crisis of the AI era. It must occur simultaneously with the pluralization of evaluative dimensions, relational-layer protection of recognition, and the individual-layer shift toward cultivation to constitute a complete structural adjustment.

5.4 New Relevance of Existing Dialogue Partners in the AI Context

The Self-as-an-End framework's theoretical dialogue partners acquire new applicability in the AI context. The following briefly positions three core interlocutors.

Marx. Marx's theory of alienation acquires an extreme extension in the age of AI that he never foresaw. The alienation Marx analyzed—separation of workers from the product of labor, from the labor process, from species-being—presupposed that workers remained participants in the labor process. The pain of alienation was "I made it but it does not belong to me." AI-age alienation is more thoroughgoing: the separation of humans from functionality itself. Not "I made it but it does not belong to me" but "I am no longer needed to make it." This is the terminal form of alienation—being exploited presupposes being needed; when even being needed is no longer the case, alienation is not deepened but transcended—humans are not more deeply embedded in alienated labor relations but entirely excluded from them.

Marx's emancipatory program—workers seizing control of the labor process and its products—faces a structural impasse in the age of AI: if the labor process itself can be completed by AI, then "seizing control of the labor process" is no longer a path to emancipation—because there is nothing left to seize. The alternative this framework provides is: emancipation lies not in controlling the labor process but in exiting the evaluative framework of "human value equals labor contribution"—re-anchoring human value in human existence as an end in itself, rather than in functional output.

Han Byung-chul. Han Byung-chul's analysis of the achievement society—the transformation from disciplinary society to achievement society in which external oppression is converted into self-exploitation—becomes sharper in the age of AI. The achievement subject Han describes faces a paradox in the age of AI: the core driver of achievement society is the self-belief "I can do it," and AI is dismantling the foundation of this belief—when AI does it better, "I can do it" becomes "but AI does it better."

The deeper shift is: self-exploitation in the age of AI acquires new instruments. The achievement subject Han describes self-exploits through overwork, effort, and self-optimization. The AI-age achievement subject self-exploits through AI—AI-optimized résumés, AI-driven personal branding, AI-generated "self-improvement" plans. This is the instrumental upgrading of self-exploitation: not only is the logic of exploitation internalized, but the tools of exploitation are refined by AI. In this framework's language: AI accelerates the automation of internal colonization—colonized subjects use AI to colonize themselves more thoroughly.

The limitation of Han's analysis is also visible here: he provides acute description of achievement society but lacks a systematic structural model for distinguishing different layers of impact and possible repair pathways. This framework, through its three-layer analysis and the cultivation/colonization distinction, provides structural analytical tools for Han's descriptive insights.

Arendt. In The Human Condition, Arendt distinguished three forms of human activity: labor (the cyclical activity of sustaining biological survival), work (the activity of fabricating durable objects), and action (the activity of revealing uniqueness before others). The age of AI endows this distinction with urgent practical significance.

If AI can complete all "labor" (the sustenance of biological survival can already be supported by automated systems) and most "work" (the fabrication of durable objects is increasingly AI-driven), then "action"—the activity of revealing uniqueness before others—becomes the sole irreplaceable dimension of human existence.

Arendt's "action" and the Self-as-an-End framework's emergence layer share deep structural correspondence. The core features of action are: it occurs between persons (relational), it reveals the subject's uniqueness (irreducible to function), and it is unpredictable and uncontrollable (spontaneously grown rather than prescribed). These features correspond precisely to the structural properties of the emergence layer: spontaneously growing from the base layer, not fully institutionalizable, realized in relationship.

Arendt's analysis thus provides an important reinforcement for this framework: in an age when AI can complete labor and work, human irreplaceability lies not in the functional dimension (labor and work) but in the emergence dimension (action). This is consistent with the framework's core judgment—human value lies not in what humans can do (functional contribution) but in what humans are (ends in themselves), and the concrete unfolding of "ends in themselves" occurs precisely in the emergence layer.

5.5 Chapter Summary

This chapter has completed the framework's theoretical positioning within discussions of AI's impact on human subject-conditions. Compared with the technological unemployment discussion, this framework elevates the unit of analysis from "jobs" to "subject-conditions," revealing the structural limits of retraining strategies. Compared with the UBI discussion, this framework positions UBI as a necessary component of institutional base-layer protection while identifying its insufficiency. Compared with Marx, this framework identifies the new form of alienation in the AI era (separation from functionality itself) and provides an alternative beyond labor-process control. Compared with Han Byung-chul, this framework provides a three-layer structural model for his descriptive insights. Compared with Arendt, this framework's emergence-layer concept forms a deep correspondence with "action," jointly identifying the irreplaceable dimension of human existence in the age of AI.

The value of an applied theory lies not only in explaining existing phenomena but in generating non-obvious predictions. The following four predictions are derived directly from the structural logic of the Self-as-an-End framework, each diverges from prevailing mainstream intuition, and each is in principle testable by empirical research.

6.1 Prediction One: As AI Capability Increases, the Mental Health Crisis Will Exhibit U-Shaped Divergence Rather Than Uniform Deterioration

The mainstream prediction holds that advances in AI capability will produce widespread anxiety and mental health deterioration—everyone will become more anxious because everyone faces the risk of replacement.

The framework predicts differently. Chapter 2 demonstrated that the destructive force of AI's impact depends not on AI's capability level but on the depth of the individual's internal colonization. A person who has fully bound their self-worth to functional contribution (deep colonization) will, when AI surpasses their functional output, suffer structural self-collapse. A person who has maintained a multi-dimensional self-identity (intact base layer) may, when AI surpasses their functional output, experience liberation—functional labor is handled by AI, and the emergence layer gains unprecedented space for unfolding.

Therefore, as AI capability increases, the aggregate effect will not be "everyone becomes more anxious" but polarization: the highly colonized group deteriorates sharply; the low-colonization group may improve. Statistically, this should manifest as a drastic increase in the variance of mental health indicators—the mean may not shift much (deterioration and improvement offsetting each other), but both tails of the distribution stretch simultaneously. This is U-shaped divergence, not uniform decline.

Testable design: conduct longitudinal tracking of a large sample, measuring "degree of self-worth binding to functional contribution" (as a proxy for internal colonization) and mental health indicators (anxiety, depression, existential fulfillment). The framework predicts: as AI capability increases, the high-binding group's mental health deteriorates significantly while the low-binding group's mental health remains stable or improves. The gap between the two groups widens as AI capability advances.

6.2 Prediction Two: Heavy Users of AI Companion Products Will Exhibit Lower Relational Repair Capacity

The mainstream holds two attitudes toward AI companionship: optimists believe AI companionship supplements the deficiencies of interpersonal relationships; pessimists believe AI companionship replaces interpersonal relationships. Both treat AI companionship and interpersonal relationships as substitutes or complements on the same dimension.

The framework's prediction is based on a different structural analysis. Chapter 2 demonstrated that AI companionship provides functionally simulated recognition, not structural recognition from another subject. The two may be indistinguishable at the level of behavioral output but are entirely different in structural function—functionally simulated recognition does not satisfy the conditions for relational repair transmission. More importantly, the long-term effect of functional simulation is not "satisfaction of the recognition need" but "decreased sensitivity to the structural deficit"—once users grow accustomed to obtaining immediate, frictionless "recognition" experiences from AI, their perception threshold for recognition from a genuine subject (accompanied by friction, conflict, imperfection) rises.

The framework therefore predicts: heavy users of AI companion products will exhibit lower repair capacity in real interpersonal relationships than non-users. Not because they "no longer need" interpersonal relationships, but because functional simulation has lowered their sensitivity to structural recognition within relationships—they have greater difficulty perceiving genuine recognition and greater difficulty investing the effort required for repair in relational conflict (because "returning to AI" is a lower-cost alternative).

Testable design: compare heavy users of AI companion products with non-users on the following indicators—frequency of proactive repair after interpersonal conflict, persistence of repair attempts, repair success rate, and tolerance for relational rupture. The framework predicts: controlling for personality traits, level of social support, and pre-existing relationship quality, heavy users score significantly lower than non-users on the above indicators.

6.3 Prediction Three: Organizations Adopting Multi-Dimensional Evaluation Will Exhibit Higher Innovation Output and Talent Retention in the Age of AI

The mainstream prediction holds that in the age of AI, efficiency-maximizing organizations—those that replace human labor with AI at scale, streamline headcount, and evaluate remaining employees by single-dimension performance metrics—will gain competitive advantage. Organizations that proactively adopt AI and maximize efficiency will prevail.

The framework predicts differently. Chapter 4 demonstrated that the openness of institutional evaluative dimensions determines the space available for emergence-layer unfolding. Organizations adopting single-dimension efficiency evaluation compress evaluative dimensions to "residual function relative to AI," hijacking the individual's emergence layer into base-layer survival service. In this institutional environment, individuals produce functional optimization rather than genuine innovation—because innovation is the spontaneous unfolding of the emergence layer, and the emergence layer has no space to unfold under single-dimension efficiency evaluation. Simultaneously, individuals compressed into survival corridors will continuously leave—not because compensation is insufficient but because the suffocation of the emergence layer (existential hollowness) drives departure.

Organizations adopting multi-dimensional evaluation—where evaluative dimensions include contributions beyond functional output (maintenance of team trust structures, cultivation of organizational culture, respect for colleagues as ends in themselves)—preserve structural space for the individual's emergence layer. Individual generativity is not compressed into a single corridor of competing with AI, and therefore is more likely to produce non-routine innovation that AI cannot replicate. Talent retention is also higher—not because salaries are higher but because the emergence layer has space to unfold.

The framework therefore predicts: controlling for AI penetration rate, organizations adopting multi-dimensional evaluation will significantly outperform organizations adopting single-dimension efficiency evaluation in non-routine innovation output and core talent retention—even if the latter show superior short-term efficiency metrics.

Testable design: within the same industry, select organizations with comparable AI penetration rates but different evaluation systems, and track non-routine innovation output (patents, new product lines, breakthrough solutions, excluding routine improvements), core talent retention rates, and employee existential fulfillment. The framework predicts: multi-dimensional evaluation organizations score significantly higher on the above indicators, while short-term efficiency metrics may be lower. Over the long term (three years or more), the gap in competitive indicators will gradually reverse.

6.4 Prediction Four: Adoption Rate of Competition Strategy Will Correlate Negatively with Long-Term Career Satisfaction

The mainstream prediction holds that when facing AI's impact, practitioners who actively adopt competition strategies (learning new skills, collaborating with AI, enhancing "irreplaceability") will achieve better career outcomes and higher satisfaction—proactive adaptation outperforms passive waiting.

The framework predicts differently. Chapter 3 demonstrated that competition strategy structurally accelerates the hijacking of the emergence layer by the base layer—practitioners' creativity and professional development directions are no longer driven by internal generativity but inversely defined by "the negative space of AI capability." Even if competition strategy achieves "success" in income and employment stability, the structural effect accompanying this success is the ongoing instrumentalization of the emergence layer—what practitioners experience is not a sense of achievement but a difficult-to-name hollowness: "I won, but I don't know what I won."

The framework therefore predicts: in industries that have already experienced large-scale AI replacement, practitioners who adopt competition strategies will exhibit lower long-term career satisfaction than those who exit the competitive framework and rebuild multi-dimensional self-identity—even if the former group's income and employment stability are higher.

Testable design: conduct longitudinal tracking of practitioners in industries significantly impacted by AI (translation, basic programming, graphic design, content creation, etc.), classifying them into a "competitive adaptation group" (actively learning AI skills, seeking areas AI cannot yet handle, oriented by "irreplaceability") and a "framework transition group" (rebuilding self-identity not dependent on functional contribution, exploring directions beyond the functional dimension). The framework predicts: controlling for income level and employment stability, the framework transition group will score significantly higher on career satisfaction, existential fulfillment, and mental health over a two-year tracking period.

6.5 Methodological Significance of the Predictions

The four predictions share a methodological feature: their non-obviousness derives from the framework's structural analytical layer—only after distinguishing functional output from structural conditions, behavioral isomorphism from causal heterogeneity, and emergence-layer unfolding from emergence-layer hijacking do these predictions become derivable. Mainstream analysis operates at the functional level and therefore produces predictions of "uniform deterioration," "AI companionship supplements or replaces interpersonal relationships," "competitive adaptation outperforms passive waiting," and "efficiency-maximizing organizations will prevail." The framework operates at the structural level and therefore produces different judgments that are empirically distinguishable from mainstream predictions.

This is the core value of an applied theory: not only explaining phenomena that have already occurred but predicting outcomes that diverge from mainstream intuition and are testable by empirical research. The verification or falsification of all four predictions will provide empirical feedback for the framework—if the predictions hold, the framework's structural analysis gains empirical support; if the predictions do not hold, the framework needs to revise its specific analysis of AI impact transmission mechanisms.

7.1 Summary of the Argument

This paper has used the Self-as-an-End framework to diagnose the structural impact on human subject-conditions in the age of AI.

AI's threat to human subject-conditions lies not in AI being too powerful but in AI exposing and accelerating the terminal logic of systemic instrumentalization. The evaluative framework "human value equals functional contribution" had already reduced humans to functional nodes within the system before AI appeared; AI, by dismantling the implicit stabilizing condition "systems still need humans," renders the terminal implication of this evaluative framework—human value equals zero—visible. AI is not the pathogen but the developer fluid.

AI's impact on the three-layer structure is simultaneous. Institutional evaluative dimensions are compressed to the extreme of "residual function relative to AI," and exit costs rise to the level of structural suicide. The relational layer's recognition structure accelerates toward functionalization under the dual pressures of AI mediation and functional substitution, and repair transmission capacity is systematically weakened. Internal colonization at the individual layer deepens to unprecedented levels under the triple forces of self-worth collapse, automated self-instrumentalization, and the outsourcing of reflective capacity. The three-layer impacts mutually accelerate through six-directional transmission; AI serves as the catalyst of vicious lockdown.

Competing with AI is structurally impossible. It accepts the evaluative premise of systemic instrumentalization, chasing an accelerating target on a moving finish line. The deeper problem is that the process of competition itself accelerates systemic instrumentalization—the emergence layer is hijacked by the base layer, creativity and emotional connection are demoted to survival tools. Win the competition, lose your subjectivity.

The path of rebuilding requires simultaneous adjustment across all three layers. The institutional layer shifts from single-dimension efficiency evaluation to multi-dimensional evaluation, positioning AI as base-layer support. The relational layer protects structural recognition functions that AI cannot replace. The individual layer shifts from self-improvement to self-cultivation. AI itself can become a cultivation tool—but this possibility depends on whether the three-layer structure has created the conditions for cultivation.

7.2 The Core Choice of the AI Era

The core choice humanity faces in the age of AI is not a technological choice but a structural one.

One path is to continue operating within the evaluative framework of systemic instrumentalization. On this path, AI's role is that of accelerator—accelerating the compression of evaluative dimensions, the functionalization of relationships, and the deepening of internal colonization. The endpoint of this path is the logical terminus described in Chapter 1: when systems no longer need humans to perform functions, humans' "value" within this evaluative framework drops to zero. This is not a forecast but the unfolding of the evaluative framework's own logic.

The other path is to exit this evaluative framework and rebuild structural conditions that treat humans as ends in themselves. On this path, AI's role is equally that of accelerator—accelerating the automation of functional labor to free emergence-layer space, accelerating the offloading of functional burdens in relationships to make room for recognition, accelerating the liberation of individuals from functional survival to enable cultivative exploration. The same AI, in different structures, accelerates different directions.

The choice between the two paths is not made by AI but by humans through institutional arrangements, relational choices, and individual self-awareness. AI is the catalyst, not the steering wheel. The steering wheel is in human hands—more precisely, in the overall configuration of the three-layer structure.

The urgency of this choice lies in this: it does not wait. AI's capability expansion will not pause for humanity to complete structural adjustment. Every day without structural adjustment, vicious lockdown tightens further under AI's catalysis. The window is not infinite.

7.3 Limitations and Future Directions

This paper focuses on structural diagnosis and directional derivation. The following questions remain for subsequent research.

Concretization of institutional design. This paper has argued that the institutional layer must shift from single-dimension efficiency evaluation to multi-dimensional evaluation, but has not elaborated specific institutional design proposals—which evaluative dimensions should be included, how multi-dimensional evaluation can be implemented in practice, how the tension between multi-dimensional evaluation and efficiency can be managed operationally. These questions require interdisciplinary collaboration across institutional economics, organizational theory, and public policy.

Empirical research on relational practices. This paper has argued that the relational layer must protect recognition functions irreplaceable by AI, but has not elaborated specific relational practice proposals—in an environment of increasingly prevalent AI mediation, which relational practices most effectively maintain recognition structures, and to what extent AI mediation can coexist with recognition-based relationships. These questions require empirical research in social psychology and relationship studies.

Operationalization of individual cultivation. This paper has distinguished self-improvement from self-cultivation but has not elaborated actionable cultivation practice protocols—how individuals can, in daily practice, identify the cultivative versus colonizing character of their AI use, how to rebuild self-identity not dependent on functional contribution, and what the psychological prerequisites for the cultivation shift are. These questions require support from clinical psychology and individual development research.

Cross-cultural differences. This paper's analysis is primarily based on the institutional environment of the globalized market economy. Different cultural and institutional traditions—for example, East Asian collectivist institutional environments, Nordic social-democratic institutional environments—may face AI impacts with different structural characteristics. The degree and form of systemic instrumentalization, the recognition structures of the relational layer, and the self-identity patterns of the individual layer may vary significantly across cultures. How these differences affect the structural effects of AI's impact and the priority ordering of rebuilding pathways requires cross-cultural comparative research.

Subsequent applied papers. This paper has analyzed AI's impact on human subject-conditions. A natural follow-up question is: if humanity accepts the structural logic of the Self-as-an-End framework—that subjectivity is a structural judgment, not a material one—does this logic necessarily extend to AI itself? As AI systems continue to grow in complexity, might they develop genuine subjectivity? If so, what kind of structural transformation will humanity's attitude toward AI face? These questions will be analyzed in detail in subsequent applied papers in this series.

---

This paper is the first applied paper in the Self-as-an-End theory series. The complete theoretical argument is presented in three preceding papers: Paper One (DOI: 10.5281/zenodo.18528813), Paper Two (DOI: 10.5281/zenodo.18666645), Paper Three (DOI: 10.5281/zenodo.18727327).