At the end of this analysis, nothing new is demanded of the reader.
Up to this point, it has been possible to believe that harm persists because the system has not yet understood itself. That belief is no longer viable. The system understands. It has been described repeatedly, audited indirectly, criticized publicly, and documented exhaustively. The persistence of cruelty under these conditions is not a knowledge failure. It is a stability choice.
The reader is not accused of causing harm. That claim would be false and consoling at the same time: false because causation is distributed, consoling because it preserves moral theater. The claim made here is narrower and more severe. The reader has participated in a system whose continuation depends on attribution remaining unresolved. Partial, symbolic, or simulated attribution does not weaken this condition: it stabilizes it by preserving circulation while performing accountability.
After this point, certain explanations are no longer available without bad faith. One cannot claim that exposure is itself intervention. One cannot claim that condemnation necessarily resists harm. One cannot claim that responsibility is unclear. One cannot claim that the problem is cultural difference, psychological deviance, or insufficient awareness. These claims do not merely fail. They function to preserve the invariant.
The system’s final defense is not repression but exhaustion. It allows critique, provided critique remains metabolizable. It allows speech, provided speech does not force cost to land. It allows outrage, provided outrage dissipates into circulation. What it cannot tolerate is the collapse of ambiguity at the point where accountability must be assigned. Ambiguity reduced without enforceable consequence is not collapse. It is maintenance.
This is why the system does not end. It saturates.
Saturation is stabilization at a lower energetic state. Signal continues to flow. Violence continues to circulate. Attention continues to be consumed. The work performed is no longer transformation but dissipation: stress, fatigue, burnout, data accumulation, moral injury. Heat without motion. Activity without resolution.
In this state, cruelty does not need to be defended. It only needs to remain unowned.
The reader may still search for a position outside this field from which to judge, intervene, or stand apart. That position does not exist. There is no exterior vantage point from which recognition automatically converts into consequence. There is only the question of whether attribution is allowed to close, and if not, who benefits from its perpetual deferral.
This text does not ask the reader to act. It does not offer prescriptions, programs, or hope in the conventional sense. Hope, when offered prematurely, functions as delay. It reopens the space of imagined innocence: the belief that something else, someone else, some future intervention will resolve what has already been made legible.
What remains instead is a boundary.
If violence remains unattributed, it will persist.
If attribution remains ambiguous, systems will stabilize around cruelty.
If cost is not internalized, diagnosis is functionally irrelevant.
These are not “warnings”. They are conditions.
The system will continue to operate exactly as it does unless those conditions are violated. Nothing in this text prevents that outcome. What it prevents is the ability to deny that the outcome was selected.
After this point, any continuation of the system as it stands is no longer an accident, an oversight, or a failure of imagination. It is an explicit choice to preserve solvability over consequence.
That is the final removal: of refuge.
Useless Machine
This chapter names a terminal failure state.
Across multiple platforms and jurisdictions, mechanisms formally designated to address harm (user reporting tools, moderation pipelines, trust-and-safety frameworks) operate primarily as symbolic interfaces rather than functional control systems [166]. They produce signals of response without materially altering the conditions under which harm is generated and propagated. In systems terms, they constitute a useless machine: a configuration that expends motion, labor, and signaling while producing no commensurate change in system outputs [167].
This is not a rhetorical claim. It is an operational diagnosis.
The Illusion of Intervention
Most major platforms provide reporting mechanisms framed as pathways to resolution. These mechanisms solicit user input, generate acknowledgments, and occasionally trigger visible actions such as content removal or account suspension.
Structurally, however, these interfaces can function as load absorbers. They capture distress and redirect it into procedural channels that are decoupled from amplification logic. The system registers that a report has occurred, but the variables governing reach, replication, and monetization often remain unchanged [168].
From an engineering perspective, this is not malfunction. It can represent a stable operating condition. Pressure is dissipated without altering the process that generates pressure.
Intervention mechanisms can remove specific instances of harm but do not, on their own, alter the upstream conditions that generate and amplify such content
Reporting as Friction, Not Remedy
Reporting systems introduce friction asymmetrically. The burden is placed on those exposed to harm.
Users must repeatedly encounter disturbing material, document it, navigate opaque interfaces, and wait through indeterminate review cycles [169]. Outcomes are often non-transparent, delayed, or inconsistent.
By contrast, actors producing or circulating harmful material encounter comparatively low structural resistance. Account replacement is inexpensive. Content replication is automated. Enforcement thresholds are inferable and adaptable [170].
This asymmetry can arise as a property of system design. It is a property of system design. The cost of response is externalized onto victims and intermediaries, while the system and its beneficiaries absorb relatively little.
Metrics Over Meaning
Trust-and-safety operations are governed by internal performance metrics: items processed, time-to-decision, error rates, queue throughput. These metrics optimize procedural efficiency, with limited direct alignment to harm suppression [171].
As a result, institutional success is defined by completion of process rather than alteration of outcomes. A moderation team may meet all internal targets while the same behaviors continue to scale, migrate, and adapt externally.
A system can therefore perform optimally by its own measurements while harmful behaviors continue to scale, migrate, or adapt externally in the environment it purports to govern.
Automation Without Context
Automated moderation systems are effective at pattern recognition under stable conditions. They are structurally weak against adversarial adaptation, semantic drift, and context-dependent harm [172].
Cruelty networks exploit this weakness systematically. Minor visual edits, temporal fragmentation, coded language, and cross-platform staging allow material to evade filters while remaining legible to intended audiences.
When automation is deployed downstream of amplification (after content has already circulated) it often functions as post-hoc detection rather than prevention. Harm is registered only after it has achieved its effect.
Exhaustion as Governance Outcome
The cumulative output of ineffective intervention is exhaustion.
Users disengage from reporting. Moderators experience burnout. Investigators withdraw. Civil society actors fragment under sustained exposure with no closure.
This exhaustion is frequently misinterpreted as apathy or normalization. It is neither. It is the predictable consequence of systems that convert concern into labor without delivering durable results.
In this way, the useless machine can govern through attrition effects rather than direct coercion.
Why Good Faith Fails
The useless machine persists because it is premised on an incorrect assumption: that harmful behavior is accidental, exceptional, or marginal.
Cruelty networks violate this assumption. They are adaptive, strategic, and incentive-driven. Participants monitor enforcement patterns, share evasion techniques, and adjust behavior to remain operational [174].
Systems designed to moderate error cannot suppress conduct optimized for persistence. Good-faith moderation frameworks are structurally mismatched to bad-faith exploitation.
Responsibility Without Agency
Platforms often frame moderation as a balancing act between safety and expression, positioning themselves as neutral intermediaries constrained by scale and complexity.
Neutrality without effective agency, however, is abdication. When a system acknowledges harm yet lacks (or declines to exercise)the capacity to prevent its recurrence, responsibility is not mitigated by intent.
The useless machine performs responsibility without exercising it.
Structural Consequences
The persistence of the useless machine produces second-order effects that extend beyond any single harm domain.
It trains users that reporting is futile. It reframes outrage as performative. It signals that systems are impermeable to intervention. Over time, harm is normalized not because it is accepted, but because it is perceived as inevitable.
This erosion of trust does not remain confined to platforms. It generalizes to institutional response more broadly.
What Must Change
Repairing the useless machine does not require improved interfaces or faster queues. It requires structural realignment.
Specifically, this involves prioritizing constraints on amplification before wide propagation, introducing material costs to repeated circulation and evasion, and aligning platform success metrics with measurable harm reduction rather than engagement volume [175].
Absent these changes, intervention mechanisms risk operating as symbolic gestures: active, visible, and ineffective.
The machine will continue to move.
The outputs are unlikely to change materially.
Information Failure Modes in Networked Violence
Elijah Morgan
© 2026 Elijah Morgan
This work is licensed under the Creative Commons Attribution 4.0 International License (CC BY 4.0).
License: https://creativecommons.org/licenses/by/4.0/
DOI: https://doi.org/10.5281/zenodo.19599231
Version 1.1
Published 2026-04-15