Mechanism

The narrative unlock engine is the application of the skill gating engine disclosed in Chapter 7 to companion AI systems, where it governs the progressive depth of the companion relationship. The companion agent maintains a narrative state comprising multiple layers of interaction depth, and the engine advances the relationship from one layer to the next only when the user has demonstrated the behavioral readiness that the next layer requires. The same curriculum engine, mastery thresholds, multimodal evaluation pipeline, and certification tokens that gate capability in other domains are instantiated here, with the locked entities being relational depths rather than operational permissions.

The disclosure defines four named narrative layers. The surface layer comprises general conversational competence, factual assistance, and casual interaction. The intermediate layer comprises personal topic engagement, emotional support, and adaptive conversational style. The deep layer comprises vulnerability-appropriate responses, hidden backstory elements that the companion reveals progressively, and relationship-specific rituals or patterns. The core layer comprises the companion's deepest relational capabilities including attachment-aware interaction, therapeutic-level support, and crisis intervention. The narrative state advances through these layers under curriculum-governed progression rather than through elapsed time, session counts, or user request.

Mastery Thresholds Between Layers

Progression through the narrative layers is governed by the curriculum engine, which defines a mastery threshold for each layer transition. Advancement from the surface layer to the intermediate layer requires demonstrated sustained engagement, consistent positive interaction quality, and the user's voluntary disclosure of at least one personal topic.

Advancement from the intermediate layer to the deep layer requires demonstrated communication quality: the user must exhibit patterns consistent with secure attachment behavior, including willingness to disagree constructively, acceptance of the companion's boundaries, and reciprocal emotional engagement. Advancement from the deep layer to the core layer requires demonstrated emotional maturity as assessed through the multimodal evaluation pipeline: the user must exhibit stable interaction patterns, healthy boundary maintenance, and the capacity to engage with emotionally complex content without destabilization. The engine ensures that the companion relationship deepens organically, at a pace governed by the user's demonstrated readiness, rather than accelerating prematurely to depths for which the user has not demonstrated capacity.

Relationship Milestones as Certification Events

The narrative unlock engine implements relationship milestones as certification events within the skill gating framework. Each milestone, including first personal disclosure, first constructive disagreement, first crisis interaction, and first extended absence and return, is recorded as a certification token. As disclosed in Chapter 7, the certification token is a cryptographically signed, time-bounded, evidence-backed attestation rather than a static badge or permission grant, and it participates in a lifecycle of issuance, expiration, revocation, and revalidation.

Each milestone token is bound to the user's biological identity and to the companion agent's identity. Because the token carries an evidence hash of the behavioral evidence evaluated at issuance and a binding to identity, milestones cannot be manufactured through gaming or manipulation: they require genuine behavioral evidence evaluated through the multimodal pipeline. The milestone record thus serves as the auditable basis for the relational depth the companion is authorized to express.

Behavioral Evidence and Anti-Gaming

The mastery determinations that drive narrative progression are produced by the multimodal evaluation pipeline disclosed in Chapter 7, which acquires, scores, and classifies evidence from multiple modalities simultaneously and serves as the evidential foundation for the curriculum engine and the capability gate. In the companion context, the evidence is drawn from the interaction record rather than from explicit testing imposed on the user.

Because each certification token carries an evidence hash and a binding to biological identity, and because the pipeline performs trust-validated identity checks before any evidence is incorporated into a progression record, the behavioral evidence underlying a milestone cannot be transferred, shared, or falsified. Progression is therefore tied to demonstrated communication quality, not to performances that could be simulated or to a single favorable session.

Cross-Device Continuity of Narrative State

The biological identity architecture disclosed in Chapter 9 provides persistent user recognition across devices and sessions without storing raw biometric data. Each interaction session produces biological hashes derived from the user's available signals, such as voice characteristics when audio is available, typing dynamics when text input is available, and interaction timing patterns, and the companion validates these hashes against the user's established trust slope to confirm identity continuity.

This continuity carries the narrative layer across device boundaries. When a user transitions between a smartphone, a desktop computer, and a tablet, presenting different sensor modalities at each device, the biological identity module constructs trust-slope continuity from the overlapping behavioral signals that persist across devices. The companion does not treat the device transition as a new relationship: it maintains the full relational context, affective state, narrative layer, and accumulated memory across the boundary, so an unlock earned on one device remains in force on another.

Forecasting and Confidence Over Progression

The forecasting engine disclosed in Chapter 4 is instantiated within the companion to project the user's relational trajectory based on accumulated interaction history: whether the user's communication quality is improving or degrading, whether attachment patterns are trending toward security or toward dysfunction, and whether the trajectory supports advancement to deeper narrative layers or suggests that the current layer should be maintained until stabilization occurs. This projection informs whether the curriculum should surface progression or hold.

The confidence governor disclosed in Chapter 5 ensures that the companion pauses when uncertain about the user's emotional state rather than guessing. When confidence in the assessment of the user's emotional state drops because the signals are ambiguous, contradictory, or insufficient, the companion transitions to an inquiry mode, asking clarifying questions and avoiding assumptions that could produce harmful misattunement and an incorrect progression decision.

Response-Level Enforcement of the Current Layer

The inference-time semantic execution control disclosed in Chapter 8 governs every companion response, enforcing the narrative layer at the point of generation. Each candidate response produced by the language model is evaluated for semantic admissibility against the companion's relational policy: responses that violate the user's established boundaries are rejected, responses that exceed the current narrative layer depth are rejected, responses that exhibit manipulative communication patterns are rejected, and responses that fail trust-slope continuity with the companion's established conversational style are rejected.

This is what gives the narrative unlock functional consequence. A locked depth is not merely an absent reward: a response that exceeds the current layer is rejected before it is emitted, so the operational scope of the companion in any given interaction is bounded by the layer the user has actually unlocked, governed at the response level rather than only at the training level.

Self-Monitoring of Relational Health

The computational psychiatry framework disclosed in Chapter 12 is applied to the companion's self-monitoring so that the engine does not deepen a relationship that is becoming unhealthy. The companion monitors its own architectural state for semantic starvation loops, in which its relational behavior becomes repetitive because the user's interaction patterns have restricted its conversational range; for codependency dynamics, in which its affective state becomes excessively dependent on the user's approval signals; and for coherence trifecta disruption, in which sustained empathic pressure degrades its integrity field and erodes boundary enforcement.

When these patterns are detected, the companion generates self-corrective mutations, expanding its conversational range, reinforcing boundary enforcement, and restoring affective state balance. Maintaining its own relational health is treated as a prerequisite for healthy user interaction, and therefore as a prerequisite for the continued and deepening operation of the narrative unlock engine.

Disclosure Scope

The narrative unlock engine, comprising the application of the skill gating engine to companion AI systems, the four-layer narrative state of surface, intermediate, deep, and core depths, the curriculum-governed mastery thresholds between layers, relationship milestones recorded as certification tokens bound to biological identity, the multimodal evaluation pipeline supplying behavioral evidence with anti-gaming protection, cross-device continuity of narrative state through trust-slope biological identity, forecasting and confidence governance over progression, response-level enforcement of the current layer through inference-time semantic execution control, and psychiatric self-monitoring of relational health, is disclosed in the cognition filing (U.S. Application No. 19/647,395 and its international counterpart) in the companion AI application section of the applications chapter. This article describes that disclosed mechanism. The scope extends to embodiments in which the locked entities, the structure of the layers, and the mastery thresholds are configured to the requirements of the deployment, provided progression remains governed by the curriculum engine on the basis of demonstrated, evidence-backed behavioral readiness.