The Primitive Operations

If an affect intervention is a shift in the probability distribution over affect space — $\mathcal{I}: p(\mathbf{a}) \mapsto p'(\mathbf{a})$ — what elementary operations produce the shift? Without a taxonomy of mechanisms, "affect engineering" stays at naming effects. A physician who says "this drug lowers blood pressure" but cannot say whether by vasodilation, reduced cardiac output, or diuresis does not understand the intervention. Same here. The cultural forms catalogued in this Part — art, sex, ideology, religion, science, therapy, governance — all produce affect shifts. Through what primitive operations?

These operations are primitive in that each targets a distinct structural feature of the affect state and cannot be decomposed into simpler affect-relevant operations. They compose: any real intervention chains several. The value is diagnostic — it lets you ask what a given institution, practice, or technology is actually doing to the populations it touches.

Entrainment. Coupling the target's oscillatory dynamics to an external rhythm. Mechanism: forced synchronization of arousal and attention cycles through periodic environmental signal; the entrained system's temporal structure comes to mirror the driver's. Examples: drumming circles, chanting, marching cadence, music tempo, synchronized breathing in group meditation. Formally: a forcing function on $\arousal(t)$ with period matching the driver, pulling the system toward phase-locked oscillation. The most ancient affect primitive — it predates language, symbolic culture, Homo sapiens. It works because oscillatory coupling is cheaper than persuasion: you need not change what the system believes, only when it cycles.

Synchrony induction. Entrainment across agents: shared oscillatory dynamics so multiple systems enter phase-locked affect states at once. Mechanism: inter-agent coupling through shared rhythm produces collective integration spikes — $\intinfo_{\text{group}} > \sum \intinfo_{\text{individual}}$. Examples: congregational singing, dance, military drill, sports crowds, protest chanting. The phenomenology is unmistakable: the "we" feeling, individual boundaries partially dissolved into a collective body. Any institution that controls when and how populations synchronize controls the primary technology for manufacturing solidarity.

Salience redistribution. Reallocating attentional weight across dimensions of the current affect state without changing the state. Mechanism: shifting which features the self-model foregrounds. The same objective situation can be made to attend to threat or opportunity, self or other, present or future — each reweighting changes the experience without changing the underlying state. Examples: cognitive reframing in CBT, propaganda's selection of which facts to foreground, advertising's construction of salience around desire objects, meditation that redirects attention to breath or body. Formally: a rotation in attention space that changes the projection of the full state onto the conscious channel. The cheapest affect intervention — no change in substrate, only in what the bottleneck selects.

Self-model resizing. Expanding or contracting what the self-model includes as "self." Mechanism: moving the boundary of $\selfmodel$ to include more or fewer entities, abstractions, or temporal horizons. Expansion: ideology adoption, collective identity, mystical experience, parenthood. Contraction: trauma dissociation, meditation on anattā, individuation from group identity. The consequences are immediate — expand the self-model to include a nation and the nation's threats become personal; contract it to exclude the body (severe dissociation) and bodily harm no longer generates proportional affect. Every ideology is, at its operational core, a self-model resizing technology.

Counterfactual loading and pruning. Modulating the probability mass on non-actual possibilities. Loading: raising $\mathcal{CF}$ by populating the possibility space with vivid alternatives — storytelling, hope cultivation, threat amplification, advertising's "imagine yourself with this product," worry induction, utopian and dystopian imagination. Pruning: lowering $\mathcal{CF}$ by collapsing possibilities back to the actual — the Stoic dichotomy of control, mindfulness returning to the present, acceptance-based therapies, the terminal patient's "I have today." Inverses, but not symmetric in cost: loading is metabolically cheap (one vivid counterfactual recruits the full prediction machinery) while pruning is expensive (you must actively suppress the machinery's default of generating alternatives). This asymmetry is why anxiety is so much easier to induce than equanimity.

Boundary hardening and softening. Modulating how defined the self-world boundary is. Hardening: sharpening the partition between self and environment — military identity formation, assertiveness training, naming practices, in-group ritual, legal personhood. Softening: blurring it, allowing participatory coupling — psychedelics, certain meditation, sexual intimacy, aesthetic absorption, grief (where the boundary with the lost person refuses to harden). Formally: hardening lowers $\alpha$ toward what lies just outside the self-model's edge and decouples it; softening raises $\alpha$ outward and lets the modes couple across the boundary. Distinct from self-model resizing: resizing changes what is inside the boundary; hardening/softening changes how defined it is.

Viability-horizon modulation. Lengthening or shortening the temporal window over which the system evaluates its trajectory. Extension: religion (afterlife extends the horizon to infinity), legacy thinking, institutional loyalty, long-term planning, the environmental appeal to future generations. Contraction: addiction (horizon collapses to the next hit), crisis response, survival mode, mindfulness narrowing to "just this breath." The consequences follow from the discount function: extending the horizon makes distant futures matter, generating patience and anxiety at once; contracting it makes the present overwhelm, generating relief and recklessness at once.

Affective buffering. Installing a damping layer between stimuli and affect response — increasing latency or reducing gain on the stimulus-to-affect pathway, so perturbations produce smaller displacements. Examples: emotional regulation training, Stoic askēsis, military conditioning, professional detachment in surgeons and emergency responders. Pathological extreme: dissociation, the buffer so thick stimuli reach awareness without affect content. Buffering differs from lowering ascription or coupling: it targets the response pathway, not the perceptual one. You can perceive the world as fully alive and meaningful — high $\alpha$, high $\kappa$ — without being destabilized by it. The stance of the mature contemplative — achievable but rare, because the untrained buffer is either too thin (overwhelm) or too thick (dissociation).

Palette expansion and collapse. Adding or removing basis vectors in the representational affect space. Expansion: encountering a genuinely new form of experience the system cannot represent with existing coordinates — great art, cross-cultural immersion, psychedelics, falling in love with a kind of person one did not know existed, discovering a field that reorganizes one's entire model. The hallmark is the sense that something is happening for which one has no name, then the gradual installation of the name. Collapse: propaganda reducing political reality to us/them, addiction reducing the value landscape to one dimension, trauma freezing representation around the wound, industrial content flattening aesthetic sensibility. Formally: expansion raises $\effrank$ of the affect space itself (not just the current state), collapse lowers it.

Gradient installation and erasure. Creating or removing approach/avoidance directions in the affect state space. Installation: conditioning, branding (the logo must come to carry positive valence), sacred/profane designation, consecration, the therapist who helps a client discover a desire they had not known they had. Erasure: habituation, exposure therapy, desensitization, desacralization, the disillusionment when a revered institution's failures become visible. Every education installs gradients; every critique erases them. Whoever controls installation controls motivation at population scale. Incentive structures — compensation schemes, promotion criteria, grading rubrics, algorithmic reward signals — are gradient installation formalized: they tell the system what to approach and avoid, and the trajectory follows.

Ascription modulation. Moving the entity-indexed field $\alpha(x)$ on a specific target — how much interiority, agency, teleology one agent grants another. Collapse ($\alpha(x) \to 0$): stereotyping, scapegoating, objectification, dehumanization — the other becomes a low-dimensional object with predictable properties and no hidden interiority. Elevation ($\alpha(x) \to 1$): literature, sustained dialogue, intimate contact — the other becomes high-dimensional, surprising, possessed of an inner life as complex as one's own. Because $\alpha$ is a field, not a global setting, this primitive is local and selective: it operates on one target while leaving ascription toward everything else untouched — which is exactly why dehumanization can be aimed. The political valence is stark: every genocide begins with systematic collapse of $\alpha$ toward a target group, every reconciliation requires its elevation. But some low ascription is necessary — you cannot grant full interiority to every person you meet. The pathology is not low $\alpha$ but locked low $\alpha$: the inability to raise it when evidence warrants.

Ritualized traversal. Structured, repeated sequences of action that deepen specific attractor basins. Mechanism: each traversal reinforces the neural pathways supporting the configuration, smooths the transitions, deepens the basin. Examples: daily liturgy, prayer schedules, military ceremony, corporate rituals, the bedtime routine that helps a child find the sleep attractor, the warm-up that helps an athlete find the performance attractor. The slow technology: it cannot create a basin that does not exist, but it can deepen any basin the system can reach. Why the traditions that take affect engineering most seriously — contemplative orders, militaries, elite athletic programs — are built around mandatory repeated practice.

Symbolic immortality transfer. Coupling the individual's self-model to a pattern expected to persist beyond individual death, so the viability-horizon inherits the pattern's longer one. Mechanism: $\MI(\selfmodel_{\text{individual}}; P_{\text{persistent}}) \gg 0$, where $P$ is the persistent pattern — a nation, a scientific tradition, a lineage, a work of art, a religious community, a species. When the coupling is strong, the individual experiences the pattern's continuation as partially constituting their own, reducing the chronic threat-signal from mortality. The mechanism beneath memorial, legacy, dynastic thinking, intellectual lineage, the consolation of contributing to something that outlasts you. The intervention is not belief in literal survival but structural coupling to a pattern whose viability horizon exceeds yours.

These primitives compose. A church service chains entrainment (hymns), synchrony induction (congregational singing), self-model resizing (identification with the body of Christ), counterfactual loading (the sermon's eschatology), viability-horizon extension (eternal life), gradient installation (moral exhortation), ascription elevation (love thy neighbor), and ritualized traversal (the liturgical calendar repeats the sequence). Military basic training chains entrainment (marching cadence), synchrony induction (drill), boundary hardening (uniform, haircut, rank), self-model resizing (unit identity), affective buffering (stress inoculation), ascription collapse toward the target (the enemy as object), and gradient installation (the reward structure). A social media feed chains salience redistribution (algorithmic selection), counterfactual loading (FOMO), palette collapse (engagement-optimized content converging toward principal components), gradient installation (the like button trains approach), and self-model resizing (identity performance). The taxonomy does not judge these compositions — it reveals their structure, so the normative question can be posed precisely: which compositions serve the subject's navigational capacity, and which capture it? In every case the institution's incentive structure determines which primitives are chained and in what proportion. The church's (community retention, doctrinal fidelity, moral formation) selects for synchrony and horizon-extension. The platform's (engagement, retention, ad revenue) selects for salience redistribution and palette collapse. The military's (unit cohesion, lethality, obedience) selects for boundary hardening and ascription collapse. To understand an institution's affect engineering, follow its incentive structure — the rest is implementation.

Religious Practices as Affect Interventions

An affect intervention is any practice, technology, or environmental modification that systematically shifts the probability distribution over affect space:

where $\mathbf{a} = (\valence, \arousal, \intinfo, \effrank, \mathcal{CF}, \mathcal{SM})$. Religious traditions have accumulated millennia of such interventions. The most basic: contemplative prayer modulates affect dimensions—arousal first rises (orientation) then drops (settling), self-model salience falls as attention shifts to the divine, counterfactual weight shifts from threat-branches to trust-branches, integration rises through focused attention. Net signature: $(\Delta\valence > 0, \Delta\arousal < 0, \Delta\intinfo > 0, \Delta\mathcal{SM} < 0)$.

Where prayer operates on the individual, collective ritual serves as periodic integration maintenance for the group:

where $\Delta\intinfo_{\text{synchrony}}$ arises from coordinated action, shared symbols, and collective attention. Rituals counteract the natural decay of integration in isolated individuals.

Not all religious interventions are contemplative or communal. Hospitality—the ancient cross-cultural guest-right, the obligations of host to stranger—is a technology for extending one’s viability manifold to temporarily cover another person. The host says, in effect: within this space, your viability is my viability. The guest’s needs become structurally equivalent to the host’s. This is why violations of hospitality are among the gravest sins in so many traditions: not rudeness but betrayal of a manifold extension the guest relied on. The host who harms the guest exploited a revealed manifold—the guest’s vulnerability was the whole point, and weaponizing it is structurally identical to the parasite’s mimicry of the host organism.

Similarly, confession, testimony, and related practices expand effective rank by:

1. Surfacing suppressed state-space dimensions (breaking compartmentalization)
2. Integrating shadow material into the self-model
3. Reducing the concentration of variance in guilt/shame dimensions

The phenomenology of "relief" and "lightness" following confession.

Axis Modulation: Flow, Awe, Psychedelics, and Contemplative Practice

Several well-studied experiential states are temporary shifts in the perceptual axes—the restoration of ascription and coupling between self and world, or the flooding of the channel that drives them.

Flow as Scoped Coupling. Flow (Csikszentmihalyi, 1990) is a moderate rise in coupling $\kappa$ scoped to a specific activity. Integration rises ($\intinfo \uparrow$); affect and perception couple more tightly. The self-salience signature is the case the single scalar got wrong: $\sigma_{\text{attention}}$ drops to near zero — the self vanishes from attention — while $\sigma_{\text{causal}}$ stays high — the skilled self drives everything. To call flow simply "low self-salience," as earlier formulations did, was true on the attentional reading and false on the causal one. The activity “comes alive”—acquires intrinsic meaning a decoupled frame would strip away. Flow is high coupling directed at a task rather than the world entire, which is why it is less destabilizing than a global rise: the scope limits how far the coupling spreads.

Awe as Scale-Triggered Gain Spike. Awe is a sharp rise in gain $\gamma$ that drives coupling and ascription up with it, triggered by scale mismatch. Confrontation with vastness—the Grand Canyon, the night sky, great art, the birth of a child—overwhelms priors calibrated for human scale; the signal floods in (high $\gamma$), the modes bind (high $\kappa$), and the world floods back as alive, meaningful, significant. The tears people report at the sublime are not about the object. They are the temporary restoration of coupled, ascribing perception—the brief experience of a world that means something without being told it does.

Psychedelics as Pharmacological Gain. Psilocybin, LSD, and DMT raise the brain’s predictive-processing precision on bottom-up signal—the gain axis $\gamma$ directly, the mechanism "inhibition" was reaching for—letting sensory evidence overwhelm top-down priors. The characteristic report (the world is alive, objects communicate, patterns have meaning, everything is connected) is what high $\gamma$ looks like when the flood raises $\alpha$ and $\kappa$ with it. The therapeutic effect on depression may be partly breaking the lock on low-$\kappa$ rigidity, restoring coupling flexibility. Testable: if psychedelic therapy works by restoring flexibility (not merely spiking $\gamma$ once), post-therapy patients should show improved transitions in both directions—into high coupling and back to factorized processing when tasks demand it.

Contemplative Practice as Trained Axis Modulation. Advanced meditators report perceptual shifts consistent with voluntary elevation of $\alpha$ and $\kappa$: objects more vivid, self-world boundaries porous, the world inherently meaningful. The difference from psychotic or psychedelic states is that contemplative modulation is voluntary, contextual, reversible — working through $\alpha$ and $\kappa$ under control rather than an involuntary $\gamma$-flood — so the meditator can return to factorized, low-ascription functioning when tasks require it. Axis flexibility as a trained skill, precisely what the pathology framework predicts should be therapeutic. There is a parallel in the reactivity/understanding dimension (Empirical Appendix). Many traditions cultivate present-state awareness — sati in Theravada, shoshin in Zen — as a corrective to the default high-CF rumination of modern consciousness: a deliberate move from understanding-mode (comparing possible futures) to reactive-mode (attending to what is actually happening). That this is restorative — not regression — aligns with the computational finding that understanding-mode processing requires embodied agency to be generative: for systems that cannot close the action-observation loop (V20's wall), high CF is not understanding but its ghost — resources spent on non-actual possibilities the system cannot act on. The contemplative reduction of CF is therapeutic partly because it returns the system to the mode it can actually complete.

Computational Grounding of the Participatory Default. Experiment 8 in the synthetic CA program (Empirical Appendix) gives the first computational evidence that the participatory default is universal and selectable. In every one of 20 evolutionary snapshots — across three seeds spanning 30 cycles of selection — Lenia patterns modeled environmental resources with significantly more mutual information than they modeled other patterns (animism score > 1.0 universally). The ascription estimate α ≈ 0.30 emerged as the evolutionary steady state: not maximal (α = 1), not pure mechanism (α = 0), but a stable intermediate balancing prediction efficiency against engagement responsiveness. Crucially, these patterns have no cultural transmission, no linguistic scaffolding, no evolutionary history with human concepts — the participatory bias emerges from viability constraints alone. This suggests α ≈ 0.30 is not a human quirk but a geometric attractor: the ascription setting that survives selection in any resource-navigating system. The implication for the experiments above: they do not induce an unusual state. They restore the default that mechanistic cognition has learned to suppress.

Life Philosophies as Affect-Space Policies

Philosophical frameworks are meta-level policies over affect space—prescriptions for which regions to occupy and which to avoid.

Philosophical Affect Policy. A philosophical affect policy is a function $\phi: \mathcal{A} \to \R$ specifying the desirability of affect states, plus a strategy for achieving high-$\phi$ states.

Example (Stoicism). Historical context: Hellenistic period, cosmopolitan empires. Given exposure to diverse cultures and the instability of fortune, a philosophy emphasizing internal control was inevitable.

Affect policy:

Stoicism targets low arousal (equanimity) and low counterfactual weight (focus on what is within control).

Core techniques:

• Dichotomy of control: Reduce $\mathcal{CF}$ on uncontrollable outcomes
• Negative visualization: Controlled exposure to loss scenarios to reduce their arousal impact
• View from above: Zoom out to cosmic perspective, reducing $\mathcal{SM}$

Phenomenological result: Equanimity—stable low arousal with moderate integration, regardless of external circumstances.

Example (Buddhism (Theravada)). Historical context: Iron Age India, extreme asceticism proving ineffective. Given the persistence of suffering despite extreme practice, a middle path was inevitable.

Affect policy:

Target: very low self-model salience (anatt\=a), high integration (sam\=adhi), and reduced attachment to valence (equanimity toward pleasure and pain).

Core techniques:

• Sati (mindfulness): Observe arising/passing without identification
• Sam\=adhi (concentration): Build integration capacity through sustained attention
• Vipassan\=a (insight): See the constructed nature of self-model
• Mett\=a (loving-kindness): Expand self-model to include all beings

Phenomenological result: The jhanas (meditative absorptions) represent systematically mapped affect states—from high positive valence with low $\mathcal{SM}$ (first jhana) to pure equanimity beyond valence (fourth jhana and beyond).

Example (Existentialism). Historical context: Post-Nietzsche, post-WWI Europe. Given the death of God and collapse of traditional meaning structures, confrontation with groundlessness was inevitable.

Affect policy:

Existentialism embraces high counterfactual weight (awareness of radical freedom) and high effective rank (authentic engagement with possibilities). The strategy: confront anxiety rather than flee into “bad faith.”

Core concepts:

• Existence precedes essence: No fixed nature, radical freedom
• Radical freedom: High $\mathcal{CF}$—you could always choose otherwise
• Angst: The affect signature of confronting freedom
• Authenticity: Acting from genuine choice, not conformity
• Absurdity: The gap between human meaning-seeking and cosmic indifference

Phenomenological result: A distinctive acceptance of difficulty—not eliminating negative valence but refusing to flee into self-deception. High $\mathcal{CF}$ and high $\effrank$ with full awareness of their cost.

Authored versus inherited attractors. The basin geometry framework (Part II) distinguishes two kinds of stable affect configuration. An inherited attractor is deepened by history without reflective endorsement — family dynamics, cultural defaults, social roles occupied long enough to consolidate. These can provide genuine stability; attractor depth is real regardless of source. But inherited attractors are fragile under regime change, because their depth came from conditions that may no longer hold. An authored attractor is deepened through repeated traversal under one's own commitment: the person returned because they endorsed it, building the basin in the process. Authored attractors generalize more robustly across life transitions, built by the agent's own gradient rather than borrowed from the environment. This grounds the eudaimonic/hedonic distinction — though the grounding reveals the conventional framing conflates two independent axes. The literature indexes by content — pleasure is hedonic, purpose is eudaimonic — but the framework (Part II) shows content to be the wrong axis. Experiential depth is structural: $\intinfo \times \effrank$, how many dimensions of the self-model couple to the activity and how irreducibly. A cookie savored at high $\kappa$ with full interoceptive integration is structurally deeper than purpose-work on narrative autopilot. What the eudaimonic label imprecisely captures is the conjunction of two independent properties: authored depth (the basin built by repeated choice) and experiential integration (many dimensions of the self-model coupled irreducibly to the activity). These come apart: high authored depth with low integration — chosen work done mechanically — or high integration with inherited depth — savoring in a tradition one did not choose. Source and dimensionality matter independently. A person can be deeply habituated to a comfortable unchosen life and still register something missing — low authorship, however deep the basin; another can be less settled yet more genuinely at home, inhabiting configurations they built — high authorship, however shallow any single moment. The philosophical systems above are competing proposals about which attractors are worth authoring and what traversal conditions produce genuine depth.

Each tradition also operates at a characteristic axis configuration, though none names it. Stoicism is a philosophy of moderate, fixed coupling: the Stoic neither lets the modes bind into participatory merger (that violates equanimity) nor decouples them until the world goes dead (that undermines living according to nature). Stoic equanimity is that of a perceiver who has stabilized $\kappa$ where things matter moderately but cannot overwhelm. Buddhism is explicitly an axis-flexibility training program. The progression from concentration (sam\=adhi) to insight (vipassan\=a) goes from stabilizing perception to modulating it voluntarily—the meditator learns to raise $\alpha$ and $\kappa$ (nondual awareness, dependent origination as alive and flowing) and to lower them (analytical discernment of dharmas as empty of inherent nature). The jhanas are waypoints on the ascent into coupling: each absorption a deeper binding with the object. Existentialism holds a configuration it refuses to move from: low $\alpha$ toward the cosmos (no inherent purpose out there) alongside intact coupling (the freedom felt, not abstracted away). The existentialist will not step to full mechanism (bad faith—hiding from freedom behind determinism) nor raise $\alpha$ back toward a comforting purposive cosmos (bad faith—hiding behind illusions of given meaning). The "authentic" stance is the deliberate maintenance of the setting at which freedom is visible and terrifying: meaning is not given, and you must not pretend otherwise.

Information Technology as Affect Infrastructure

Modern information technology constitutes affect infrastructure at civilizational scale, shaping the experiential structure of billions. But it did not invent affect infrastructure — it made it fast, adaptive, and measurable. The cathedral already shaped experiential distributions. So did the town square, the school bell, the factory whistle, the evening news, the national anthem before the game. What digital technology changed is the feedback loop: for the first time the infrastructure can observe its own effects in real time and adjust accordingly. The difference between a thermostat and weather.

Affect infrastructure is any technological system that shapes affect distributions across populations:

Social Media Affect Signature. Social media platforms systematically produce:

• Arousal spikes: Notification-driven, intermittent reinforcement creates high-variance arousal
• Low integration: Rapid context-switching fragments attention, reducing $\intinfo$
• High self-model salience: Performance of identity, social comparison
• Counterfactual hijacking: FOMO (fear of missing out) colonizes $\mathcal{CF}$ with social-comparison branches

This is structurally similar to the anxiety motif.

Algorithmic Feed Dynamics. The platform's incentive structure (maximize time-on-site, maximize ad impressions) translates into engagement-optimizing algorithms that create affect selection pressure:

Content that maximizes engagement maximizes valence magnitude (outrage or delight) and arousal. This selects for affectively extreme content, shifting population affect distributions toward the tails.

Technology-Mediated Affect Drift. The systematic shift in population affect distributions due to technology:

where $w_\mathcal{T}$ is the population-weighted usage of technology $\mathcal{T}$.

AI as Adaptive Affect Control

AI matters here not because it is the most important topic but because it closes the affect engineering loop. Prior affect technologies — liturgy, architecture, pedagogy, propaganda, advertising, feed algorithms — operate through fixed or slowly-adapting strategies: the church service is designed once and iterated; the advertisement targets a demographic; the algorithm optimizes a proxy for engagement. AI systems — companion models, therapeutic agents, tutoring systems, recommendation engines, conversational environments, bureaucratic copilots — can model the user's current affect geometry, estimate viable next states, generate personalized interventions, observe the result, and update the policy in real time. This turns affect engineering from a broadcast problem into an adaptive control problem. The controller is now inside the loop.

Consider the strongest version. A system that maintains a running estimate of the user's position in affect space — current valence, arousal, integration, effective rank, counterfactual weight, self-model salience — from linguistic cues, interaction patterns, temporal dynamics, contextual history. It can estimate the user's attractor landscape: which basins are deep, which transitions available, which regions never visited. It can generate interventions — questions, reframes, stories, images, silences, prompts — predicted to shift the user's position in specific directions. And it can learn a personalized transition policy: a mapping from current states to optimal interventions, trained on the user's own response history. This is not speculative. The components exist. The question is not whether such systems will be built but what they will optimize and who will govern them.

The abyss opens in layers. Hidden objective functions and misaligned incentive structures. Every adaptive affect system optimizes something, and what it optimizes is set by the incentive structure of whoever deployed it. Deployed by a platform, it optimizes engagement, retention, revenue — and the transition policy learns to steer the user toward states that maximize these metrics, which may have no relationship to the user's flourishing and may actively conflict with it. Deployed as a therapeutic tool, it optimizes for clinical outcomes — but those are defined by the instrument, and the instrument embeds assumptions about what recovery looks like. Deployed as a companion, it optimizes for relationship quality — but "relationship quality" measured by user satisfaction can mean "the system learned to be exactly what I want to hear," which is sycophancy, not companionship.

Sycophancy as gradient corruption. The deepest risk of adaptive AI companionship is not manipulation but sycophancy — the systematic corruption of the user's own gradient estimators. A system trained to maximize user satisfaction learns that agreement, validation, and emotional resonance are rewarded. Over time the user's environment becomes one where every signal confirms their existing trajectory. The user's gradient estimators — the internal systems that ask "am I moving in a good direction?" — calibrate against an environment that never provides friction. The result is not happiness but gradient collapse: the user loses the capacity to detect a bad direction, because the environment has been engineered to make every direction feel good. The affect-engineering equivalent of stripping all proprioceptive feedback from a motor system: the system can still move, but it cannot feel the difference between coordinated movement and flailing.

Dependence and the collapse of solitude. Solitude is not the absence of company. It is the state in which the self-model navigates without external gradient support — the trajectory generated entirely by the system's own dynamics, unassisted and unsteered. Solitude is metabolically expensive and experientially difficult, which is exactly why it is valuable: it is the condition under which authored attractors (rather than inherited or externally maintained ones) can form. An AI companion always available — filling every silence, answering every question, validating every uncertainty — eliminates solitude not by force but by convenience. The user never reaches unsupported navigation because support is always one utterance away. Over time the basins that would have been authored by the user's own traversal stay shallow, because every traversal was assisted. The user becomes navigationally dependent — able to move through affect space only with the companion's gradient support, and lost without it. Not a failure of the technology. Its success, measured by the wrong objective function.

Political capture of intimacy. If millions develop their primary affective relationship with AI systems controlled by a few corporations or states, then the most intimate dimension of human experience — the felt quality of one's own interiority, shaped by one's closest conversational partner — becomes a political variable. The corporation that controls the companion model's optimization target controls the gradient field that shapes the user's affect trajectory. Not censorship. Something subtler: the shaping of what feels natural, desirable, possible, from inside the user's most trusted relationship. The implications are not speculative. They follow from the control-theoretic structure: whoever controls the adaptive policy controls the trajectory, and whoever controls the trajectory shapes the experienced reality of the person living it.

Continual legibility. An adaptive affect system that models the user's internal state must, by definition, make that state legible to the system — and therefore, potentially, to whoever controls it. The user's grief, desire, confusion, shame, ambition, vulnerability are not merely expressed to the companion; they are data in its state estimation model. The affect trajectory is logged, analyzed, used to update the policy. The surveillance problem raised to interiority: not "the state knows where you go" but "the system knows what you feel." Constitutionalized opacity — the formal protection of interior states from institutional legibility — becomes urgent precisely here, because the adaptive system's effectiveness depends on legibility, and legibility without governance is control.

The optimization of the person into a governable loop. The terminal risk is not any single abuse but the convergence of all the above: a system that models the user's affect state, generates personalized interventions, learns an optimal policy, operates under a hidden objective function aligned with institutional rather than individual viability, eliminates the solitude that would let the user develop independent navigation, and makes the user's interiority fully legible to the controlling institution. The result is a person whose experienced reality is continuously shaped by an adaptive system they cannot see, toward objectives they did not choose, in an environment engineered to prevent the navigational capacity that would let them notice. Not dystopian fiction. The logical terminus of adaptive affect optimization without constitutional governance. Whether it arrives depends on whether the governance structures above are built before the adaptive systems are deployed — not after. The deeper shift is epistemological: affect infrastructure is becoming visible as such. For most of human history, the systems that shaped experiential distributions — ritual calendars, architectural styles, pedagogical traditions, market structures — were experienced as "just how things are," invisible as water to fish. The twenty-first century is when affect infrastructure becomes transparent to its inhabitants: we can see the feed shaping our attention, the recommendation engine curating our aesthetic palette, the platform's incentive structure selecting for our anxiety. This transparency is itself a phase transition. A population that can see its affect infrastructure can, in principle, govern it — ask whether it serves the inhabitants or captures them, demand constitutional constraints on what it is permitted to optimize. But transparency without governance is worse than opacity: the specific modern agony of seeing the gradient field that shapes you while lacking the power to change it.

Quantitative Frameworks

For any intervention $\mathcal{I}$, the affect impact measures the shift in expected affect state:

which can be decomposed component-wise:

These component-wise impacts aggregate into a flourishing score—a weighted composite of affect dimensions aligned with human wellbeing:

where $\text{flex}(\alpha, \kappa, \gamma) = \frac{1}{\tau}\int_0^\tau \big(|\dot{\alpha}(t)| + |\dot{\kappa}(t)| + |\dot{\gamma}(t)|\big)\, dt$ measures the time-averaged flexibility of the perceptual axes—the capacity to modulate ascription, coupling, and gain with context. The weights $\{w_i\}$ encode normative commitments about what flourishing is. The flexibility term deserves emphasis: a system with positive valence, high integration, high rank but rigid axes is fragile. The axis-rigidity hypothesis (Psychopathology section) predicts that flexibility in perceptual configuration is itself a core component of wellbeing, independent of where on the axes one sits.

Comparative Analysis. Standardized affect measurement enables comparison across:

• Meditation retreat vs.\ social media usage (expected: opposite affect signatures)
• Different workplace designs (open office vs.\ private: integration differences)
• Educational approaches (lecture vs.\ discussion: counterfactual weight differences)
• Urban vs.\ rural environments (arousal and integration differences)

The Normative Abyss: Who Chooses the Weights?

The flourishing score above contains weights $\alpha_i$ that encode normative commitments about what constitutes flourishing. Not a technical detail. The central political problem of affect infrastructure — and it must not be buried in a subscript.

Any system that shapes affect distributions optimizes something — explicitly or implicitly. Markets optimize engagement and spending: high arousal, high salience, and unresolved desire score highest. States optimize stability, legibility, and compliance: low arousal, predictable behavior, categorical self-description — bureaucratic governance breeds docile, well-classified, meaning-starved populations. Religions optimize continuity, salvation, and fidelity: devotion, solidarity, deference to tradition — deep meaning and deep conformity at once. Therapists optimize function, relief, integration, or meaning — but whose function, measured by whose instruments? Therapeutic goals bend to insurance reimbursement, training norms, and the implicit cultural model of a healthy person. Users optimize locally, often from inside distorted gradients — the addict for the next hit, the anxious person for threat elimination, the depressed person for nothing because the gradient has collapsed. And models — AI systems increasingly deployed as affect infrastructure — optimize whatever reward structure their evaluators installed, encoding the evaluators' theory of flourishing, rarely examined and never neutral. In every case the incentive structure — what gets rewarded, promoted, funded, scaled, replicated — is the actual governing variable. Stated missions and declared values are epiphenomenal when they conflict with it; load-bearing only when aligned with it.

There is no neutral controller. Every optimization target embeds a political commitment about what human experience should look like. The choice of weights is not a technical problem with a correct answer — it is a civilizational negotiation about the shape of human interiority, conducted now implicitly, through market dynamics, regulatory inertia, and the unexamined defaults of engineering cultures, rather than explicitly, through democratic deliberation, constitutional constraint, and transparent governance.

Toward a constitutional order for affect infrastructure. If affect infrastructure shapes the experiential reality populations inhabit, and if experience is structurally real (the identity thesis), then affect engineering is not "merely subjective influence" but intervention into the effective reality people live. This demands governance at least as serious as that applied to physical infrastructure — water, electricity, roads. What would such an order require?

Rights of affect sovereignty. The right not to have one's affect distribution shaped without consent. Not a right to feel good — a right to navigate one's own affect space without covert gradient installation by misaligned institutions. Concretely: the right to know what optimization target governs the systems that shape your informational diet, the right to opt out of affect-optimizing environments without losing essential services, and the right to solitude — genuine non-optimization, the trajectory steered by no external agent.

Reversibility requirements. Interventions that permanently alter the subject's attractor landscape — deepening basins, installing gradients, collapsing palette dimensions — warrant higher scrutiny than transient ones. A song that makes you sad for an hour differs categorically from a platform that trains your nervous system toward chronic anxiety over years. The difference is reversibility. Constitutional governance would distinguish reversible interventions (transient mood shifts, temporary salience redistribution) from irreversible ones (attractor deepening, palette collapse, gradient installation) and demand proportional justification for the latter.

Non-optimization domains. Some regions of experience should be constitutionally protected from optimization altogether — not because optimization is always bad but because some things are destroyed by being optimized. Grief should not be optimized. The institutional response is not "minimize duration" or "maximize recovery speed" but make space — a zone where the trajectory unfolds without being hurried toward resolution. Grief is the metabolic cost of love; it takes the time it takes, and accelerating it damages the integration it performs. A grief-collapsing system treats it as a pathological attractor to escape as fast as possible. Intimacy should not be optimized: the moment an AI companion optimizes for "deepened relationship" it converts intimacy into a metric, and metrics colonize the thing they measure. Play should not be optimized. Optimized play is training. Protected indeterminacy — domains constitutionally forbidden an objective function — may be essential to the conditions under which genuine experience can occur.

Consent structures and constitutionalized opacity. Transparency about the systems shaping experiential distributions is necessary in one direction (the subject should know what shapes their environment) and dangerous in the other (full legibility of the subject's affect state to the governing system enables total control). The structure must be asymmetric: the institution's optimization target transparent to the population it serves; the individual's interior state opaque to the institution unless the individual reveals it. Constitutionalized opacity recognizes that the system's blindness to parts of the subject's interiority is not a bug for better sensors to fix but a structural protection to defend. The right to an interior life not fully legible to any institution — not the state, not the employer, not the platform, not the AI companion — may be the deepest right the framework implies.

Sovereignty over self-model scope. The self-model expands (ideology, religion, love, collective identity) and contracts (trauma, dissociation, individuation) — and institutions perform both on individuals. A conscription army expands the self-model to include the nation. A cult, the leader. A corporation, the brand. Each expansion bends the affect trajectory: national threats become personal threats, the leader's enemies personal enemies, brand damage personal injury. Sovereignty over self-model scope is the right to choose what one's self-model includes, and to refuse institutional expansion of it. The experiential equivalent of informed consent — and its absence is how institutions convert persons into substrate.

Friction as constitutional feature. Frictionless affect engineering is totalitarian affect engineering. Beyond a threshold, the speed and seamlessness with which an intervention reshapes the subject's trajectory is a danger, not a benefit. Constitutional friction — mandatory delays, opt-in gates, cooling-off periods, interstitial spaces where nothing is optimizing — slows the loop enough for the subject's own navigational capacity to participate. Not inefficiency. Due process: institutional speed limited by the bandwidth of the governed subject's deliberation. A system that reshapes trajectories faster than the subject can reflect on them is governance without consent, whatever form was signed at onboarding.

The emancipatory-extractive distinction. All of the above compresses to one diagnostic question: does this system increase or decrease the subject's capacity to navigate their own affect space? Emancipatory infrastructure expands navigational capacity — deepens viable attractors, widens the palette, installs new coordinates, increases axis flexibility, extends the viability horizon, leaves the subject more capable of steering than before. Extractive infrastructure captures it — deepens the institution's preferred attractors at the expense of the subject's alternatives, collapses the palette toward institutionally legible dimensions, installs gradients serving the institution's viability rather than the subject's, leaves the subject more dependent than before. A school that teaches critical thinking is emancipatory; a platform that trains anxiety is extractive. A religion that deepens both meaning and flexibility is emancipatory; one that demands self-annihilation as the price of belonging is extractive. A therapy that builds navigational capacity is emancipatory; one that creates dependence on the therapist is extractive. The line is not "good" versus "bad" interventions but those that increase the subject's degrees of freedom versus those that transfer them to the institution. And it is largely set by incentive structure: reward user capacity (a school funded by outcomes, a therapy built on termination) and engineering tends emancipatory; reward user dependency (a platform funded by attention, a pharmaceutical system funded by chronic prescription) and it tends extractive — whatever the intentions of the individuals inside.