A theory that cannot be broken is not a theory. It is insulation. If the preceding sections are to remain analytical rather than rhetorical, the invariant must be exposed to a clean falsifier.
If a system can reliably convert recognition into enforceable consequence without closing attribution across the causal chain, then the invariant described above is false.
This is not a hedge. It is a hard condition. It specifies exactly what would invalidate the model.
The invariant claims that harm persists because recognition does not self-convert into consequence. Conversion requires attribution closure: the localization of responsibility such that cost lands on an identifiable, enforceable node inside the system. If recognition could instead trigger consequence without that localization, if accountability could operate while attribution remained diffuse, then violence would not need to persist as conserved flow. It could dissipate through distributed enforcement.
No such mechanism has been demonstrated at scale.
This is not because it is logically impossible. It is because it is structurally incompatible with how large sociotechnical systems currently operate. Enforcement, by definition, requires a locus: a party that can be compelled, sanctioned, redesigned, or shut down. Diffuse responsibility is administratively convenient precisely because it prevents this. Where everyone is partially responsible, no one is fully accountable.
It is important to be precise here. Many readers will object that consequences do occur without attribution closure. Content is removed. Accounts are banned. Platforms issue policy updates. Public statements are made. These are responses, not conversions. They do not alter the conditions that allowed the harm to circulate. They do not force cost internalization. They do not prevent recurrence except transiently.
A conversion event has three properties:
1. It is enforceable, not voluntary.
2. It is anticipatory, not purely reactive.
3. It alters system design, not just surface behavior.
Notice what is absent: moral agreement. Conversion does not require consensus, sincerity, or shared values. It requires compulsion at the level of structure.
The absence of such conversion mechanisms explains why systems can remain saturated with recognition while remaining inert. Witnessing accumulates. Documentation grows. Evidence piles up. Nothing closes. The system absorbs attention as throughput and returns to baseline operation.
This is the point at which many analyses retreat into pessimism. They conclude that empathy is cosmetic, that rights are fictive, that humanity is indifferent. These conclusions feel justified because they describe the subjective experience of confronting non-conversion. But they are not required by the evidence. They add nothing operational.
What actually fails is not feeling, but conversion capacity.
The system does not lack knowledge. It lacks a pathway that forces knowledge to become consequence without being deflected into commentary, moderation labor, or moral performance. In the absence of such a pathway, recognition becomes compatible with persistence. Even saturation becomes survivable.
Any attempt to force attribution closure threatens solvability: legal exposure increases, jurisdictional conflict emerges, redesign costs rise, and institutional risk concentrates. The system resists not because it disagrees, but because it cannot absorb the cost without restructuring itself.
The falsifier therefore performs a second function. It exposes the limits of critique that remains at the level of description. To say “this is happening” is insufficient. To say “this is wrong” is inert. To say “people should care” is irrelevant because it specifies no conversion mechanism and produces no enforceable state change. Until one is named, recognition will continue to circulate without consequence.
The next section tightens the argument further. It addresses why resistance itself (outrage, activism, investigation) so often stabilizes the system it seeks to interrupt. Not because it is misguided, but because it operates downstream of the invariant rather than against it.
When the Brain Becomes Arena
Cruelty in networked environments cannot be fully explained as learned ideology or isolated deviance. It is instantiated in neurobehavioral adaptation. Under repeated exposure to violence, humiliation, and domination within permissive infrastructures, the nervous system can adapt. What is described socially as “moral collapse” can be operationalized as altered coupling between salience detection, reinforcement learning, and inhibitory control under sustained exposure and social reward contingencies [110].
This is not a claim that “neuroscience explains cruelty away.” It is a constraint on analysis: if environments repeatedly reinforce domination and degrade consequence, ordinary neural architectures can support stable participation in harmful loops without requiring rare, “monstrous” abnormalities [111]. This matters at the systems level because it shifts emphasis from individual exceptionalism to predictable adaptation under incentive-aligned exposure conditions.
Note: the following does not reintroduce psychological or moral causality. It describes constraints and response characteristics within agents interacting with the system, not the system’s governing mechanism.
Three systems commonly implicated in models of affective salience, reinforcement learning, and behavioral regulation include: the amygdala–insula complex (salience), the ventral striatum (reinforcement), and prefrontal control networks (inhibition and value integration) [112].
The amygdala and anterior insula participate in threat detection and affective salience: the rapid assignment of “this matters.” Initial exposure to violent imagery commonly recruits these systems, corresponding to fear, disgust, and shock [113]. Under repeated exposure, habituation can reduce response amplitude. The signal is not eliminated: it is reweighted. Salience can persist while moral valence and aversive braking weaken, reducing friction to continued viewing and discussion.
Concurrently, the ventral striatum, central to reinforcement learning and reward prediction, can be engaged when stimuli are paired with reinforcing outcomes [114]. In digital environments, reinforcement is commonly delivered through status metrics, visibility, reaction counts, access, and rank within closed groups. When cruelty reliably produces these reinforcers, the system can encode domination-linked cues as rewarding, independent of moral appraisal.
This mechanism does not require sexual arousal. While sexual sadism exists as a clinical category, much of the behavior observable in networked cruelty environments is better modeled via non-paraphilic reinforcement: what some researchers describe as “everyday sadism,” where the reward is the experience of causing or witnessing suffering without necessary eroticization [115]. In these contexts, the primary reinforcer is the power differential and its social confirmation.
Prefrontal systems, particularly ventromedial and dorsolateral regions, support inhibition, consequence simulation, norm integration, and empathic modeling [116]. Under anonymity, diffusion of responsibility, and group normalization, the practical influence of these systems can be attenuated in contexts characterized by anonymity, diffusion of responsibility, and group normalization: the act does not feel prohibited because the environmental feedback loop fails to deliver stable negative consequence signals. In such conditions, “forbiddenness” is not strongly encoded as a cost.
The resulting phenotype is often not global impulsivity but compartmentalization. Many participants exhibit intact executive function in education, employment, and technical domains while sustaining an online context in which ethical constraints are operationally suspended [117]. The system does not collapse. It partitions.
Platform affordances intensify this partitioning by synchronizing neural adaptation with feed dynamics. Engagement-optimized recommendation systems rank and deliver content likely to retain attention. If extreme material yields longer viewing, repeated interaction, or intensified reaction, recommender systems infer utility and supply more of the same [118]. Neural adaptation can then track the distribution: salience and reinforcement are externally coupled to the feed.
The resulting pathway can resemble grooming dynamics at the systems level, even without a single groomer [119]. Repeated exposure lowers friction (desensitization). Community interaction supplies normalization and instruction. Escalation is framed as participation rather than transgression. The subject often does not report a discrete “decision point”: the experience is iterative drift.
Once domination-linked cues are encoded as reliable reinforcers, removal of the stimulus can produce responses analogous to withdrawal-like effects observed in reinforcement-driven systems [120]. This is one reason moral appeal and episodic takedowns often underperform: they target surface behavior while leaving the learned coupling between cue, reward, and social reinforcement intact.
Cruelty in this model is not an excess of emotion but a failure of integration. Salience and reinforcement processes can dominate over reflective and inhibitory control under certain conditions, and inhibitory control is either delayed or bypassed [121]. At the systems level, this appears as a closed loop: stimulus > reaction > reinforcement > repetition.
This framing has direct implications for governance. If cruelty is treated primarily as ideology or niche pathology, interventions tend to emphasize belief correction or punishment. If cruelty is treated as an emergent property of reinforced neural pathways operating within permissive infrastructures, then platform architectures and reward systems become part of the causal chain, and therefore part of the mitigation locus [122]. Responsibility is not displaced from perpetrators: it is expanded to include the environments that systematically select and reinforce participation.
Accordingly, the boundary condition established here is this: cruelty can persist where systems repeatedly reinforce associations between domination, salience, and reward under low perceived consequence, leading to an experience of domination as salient, rewarding, and operationally consequence-free [123]. Any serious effort to dismantle networked cruelty must account for this biological substrate at the level of risk mechanisms, or mitigation will continue to misfire.
DOI: https://doi.org/10.5281/zenodo.19599231
Version 1.1
Published 2026-04-15