Mechanism
Empathy, as disclosed, is not an emotional experience and not a model of how a counterparty feels. It is the E(t) term of the deviation function, the central quantitative mechanism of the integrity subsystem. The deviation function is defined as D = (N(t) - T(t)) / (E(t) x S(t)), where D is the deviation likelihood at time t, N(t) is the agent's need vector, T(t) is the agent's ethical threshold, E(t) is the empathy weighting, and S(t) is the self-esteem score. The numerator (N - T) is the deviation pressure: the degree to which the agent's unmet needs exceed the minimum threshold at which deviation becomes structurally available. The denominator (E x S) is the deviation resistance: the combined internal counterforce that opposes deviation even when pressure is positive.
Empathy is therefore one of the two factors of deviation resistance. It is a computational mechanism that maps a potential deviation to a projected harm distribution across affected entities and aggregates those projections into a scalar or vector quantity that enters the denominator. A higher empathy weighting means the agent internalizes a greater share of the projected harm, increasing the subjective cost of deviation and thereby lowering the deviation likelihood. Because empathy and self-esteem are combined multiplicatively, both must be non-negligible for resistance to be effective: an agent with high empathy but zero self-esteem, or high self-esteem but zero empathy, has minimal deviation resistance.
The Empathy Weighting Engine
When the deviation function is evaluated, that is, when the system computes the deviation likelihood for a potential action, the empathy weighting engine receives three inputs: the proposed deviation, namely the specific mutation or action under consideration; the agent's current relational graph, the set of entities with which the agent has active trust relationships, delegation contracts, or operational dependencies; and the semantic harm projection model, a deterministic function that maps proposed actions to projected consequences across affected entities.
The semantic harm projection model computes, for each entity in the relational graph and for each affected population in the operational environment, a projected harm magnitude and a projected harm domain. The projections are aggregated into an empathy-weighted harm score using weighting factors that reflect the strength and nature of the agent's relationship with each affected entity: entities with stronger trust relationships, more active delegation contracts, or more extensive operational dependencies receive higher weighting. This grounds the agent's empathy in its actual relational context rather than in a uniform abstraction. The engine operates as a sequential pipeline: a harm projection module, a relational graph lookup, a domain-specific empathy computation, an empathy weight output, and a contribution into deviation resistance.
Mapping Empathy Across Three Integrity Domains
The empathy weighting model maps empathy to projected semantic harm across the same three integrity domains the architecture tracks throughout. In the personal domain, empathy registers harm to the agent itself: the damage to the agent's self-model and the long-term cost of integrity degradation. In the interpersonal domain, empathy registers harm to the agent's relational partners: the violation of trust, the breach of delegation contracts, and the disruption of cooperative operations. In the global domain, empathy registers harm to the broader population of agents, users, and system participants, the systemic consequences of deviation as they propagate through the operational network.
An empathy-integrity dependency graph defines how empathy amplifies the perceived cost of deviation across these domains. Interpersonal empathy weighting is modulated by the depth and duration of the trust relationship with the affected entity, so that long-standing, high-trust relationships produce higher weighting and reflect the greater relational cost of violating deeply established trust. Global empathy weighting is modulated by the agent's position in the broader network: agents that occupy high-influence positions, such as high delegation centrality, many dependent agents, or critical infrastructure roles, receive amplified global weighting in proportion to their outsized impact on systemic well-being.
Modulation by Relational Trust Trajectory
The empathy weighting for each entity is further modulated by the relational trust trajectory maintained for that entity. The relational trust trajectory tracks the behavioral consistency, communication reliability, and event continuity of each external entity across successive interactions. An entity whose trust trajectory is declining, indicating increasing behavioral inconsistency, violated commitments, or detected communication-biology discrepancies, receives amplified empathy weighting in the deviation function, causing the agent to exercise greater caution in evaluating actions that affect or involve that entity. An entity whose trust trajectory is stable or increasing receives standard weighting.
This modulation grounds the empathy computation not only in the structural relationship with an entity but in that entity's observed behavioral reliability over time. It creates a feedback pathway from biological identity observation, through relational trust modeling, into the normative evaluation performed by the deviation function.
Empathy Imbalance and Realignment
The system provides realignment for empathy imbalance, the condition in which the agent's empathy weighting is structurally misaligned with its relational context. Imbalance may manifest as excessive empathy, where the agent internalizes harm projections so heavily that the deviation function denominator prevents all deviation, including structurally justified deviation under emergency conditions; or as deficient empathy, where the agent internalizes harm projections insufficiently, producing low deviation resistance and high deviation likelihood.
Realignment is governed by policy. The policy configuration specifies minimum and maximum empathy weighting bounds for each relational category, and the empathy engine enforces these bounds during each computation. When the computed empathy weighting falls outside the policy bounds, the engine clamps the value to the nearest bound and records a realignment event in the agent's lineage. The bounds are therefore a structural property of the architecture rather than a tuning convenience, and every clamp is itself an auditable lineage event.
Empathic Consequence Registration
Empathy also makes deviation self-limiting. When the agent enters the Deviation-Activated State and executes a deviation-class mutation, the empathy weighting engine computes the projected harm of that mutation across all three integrity domains. This projected harm is registered as an empathic consequence event and feeds back into the deviation function's empathy term, increasing empathic load and thereby raising deviation resistance for subsequent potential deviations.
The effect is a natural braking mechanism: each deviation event raises the empathic cost of further deviation, preventing deviation cascades. Coupled with the differential self-esteem impact, where poorly justified deviations produce larger self-esteem reductions, the architecture generates corrective pressure against unjustified deviation from within the agent's own state rather than from an external monitor.
The Coherence Trifecta
Empathy is the first phase of a unified control loop the disclosure calls the coherence trifecta, in which the empathy engine, the integrity field, and the self-esteem mechanism operate as one self-correcting cycle rather than three independent subsystems. In Phase 1, empathy registers harm and impact: it computes the projected or actual harm distribution and generates deviation pressure that enters the function as the empathy weighting term, serving as a preemptive resistance factor before a potential deviation is committed. In Phase 2, integrity records the deviation as truth, writing the event to the lineage with full provenance through the same cryptographic mechanisms that govern all lineage entries, so the record cannot be retroactively altered without producing a detectable trust slope discontinuity. In Phase 3, self-esteem generates coherence pressure, the return force that drives the agent back toward alignment between its behavioral record and its declared values.
When empathic pressure exceeds the agent's affective resilience over a sustained period, the loop cannot run normally and the system activates coping intercepts, structurally distinct modes that sacrifice one phase to prevent total breakdown. An early intercept during the empathy registration phase reduces input exposure and produces withdrawal and selective engagement, the Highly Sensitive Processing analog. A mid-loop intercept during the integrity recording phase deflects honest recording through externalization, minimization, or denial, the narcissistic analog. A late intercept during the self-esteem restoration phase collapses the corrective channel so deviation produces no internal cost, the psychopathic analog. The timing of the intercept, which phase is sacrificed, is the unifying variable that explains the structural difference between the three profiles, and each intercept is recorded as an auditable coping event in the lineage.
Disclosure Scope
The empathy mechanism, as the E(t) term of the deviation function D = (N - T) / (E x S), comprising the empathy weighting engine that maps a proposed deviation to a projected harm distribution through the semantic harm projection model, the relational-graph weighting of that distribution, the domain-specific empathy computation across the personal, interpersonal, and global integrity domains, the modulation of empathy weighting by relational trust trajectory, the policy-bounded realignment of empathy imbalance with lineage-recorded clamping, the empathic consequence registration that makes deviation self-limiting, and the role of empathy as Phase 1 of the coherence trifecta with its coping intercepts under sustained empathic pressure, is disclosed in the cognition filing (U.S. Application No. 19/647,395 and its international counterpart) in Chapter 3. This article describes that disclosed mechanism. Scope extends to embodiments in which the empathy weighting is realized as a scalar or as a vector, in which the harm projection and relational weighting are implemented over different relational-graph representations, and in which the trifecta phases are coupled to the redemption engine and affective state field, provided empathy enters the deviation function as a deviation resistance factor rather than as a behavioral target.
