I. The Problem
Some harmful narratives about people stop when the person spreading them stops. Others don't.
This is the central problem the existing frameworks for narrative inversion have not adequately solved. The DARVO model, institutional betrayal theory, and related frameworks are primarily actor-focused: they describe what perpetrators and institutions do during the active phase of a campaign. They assume, at least implicitly, that the harm engine requires ongoing input from an identifiable actor. Remove the actor — hold them accountable, expose the pattern, exit the relationship — and the harm should stop.
It frequently doesn't.
In a significant and recognizable class of cases, the original actor disengages entirely. They stop pressing the narrative. They step away from the institutions that had been processing the harm. They move on. And the narrative continues. Communities maintain it without being instructed to. Institutions filter new information through it automatically. Social networks sustain it through ordinary interaction. The target's attempts to correct the record are processed through the lens of the narrative they are trying to correct, and those attempts become, in a perverse feedback loop, additional evidence for the narrative's accuracy.
This paper calls that phenomenon Distributed Maintenance. The central claim is:
Narrative formation and narrative maintenance are analytically distinct processes. Narratives can become self-sustaining after the original actor disengages, driven by identifiable mechanisms within social systems rather than by continued active maintenance from any originating actor. These mechanisms operate similarly regardless of whether the narrative being maintained is true, false, malicious, accidental, or politically constructed.
This is not a claim about DARVO specifically. DARVO is one mechanism by which a damaging narrative can be loaded into a social system. It is not the only one. Distributed Maintenance is a general feature of how social information systems behave once they have absorbed a narrative with sufficient institutional and social embedding. The analysis begins with narrative inversion because that is the context in which the phenomenon is most dramatically visible. It does not end there.
II. What the Existing Framework Provides and Where It Stops
Jennifer Freyd's DARVO model gave the field a precise vocabulary for a specific pattern: the way accused parties deny allegations, attack the credibility of accusers, and reverse the positions of victim and offender. The framework was built on careful clinical observation and has been productively applied across therapeutic, organizational, legal, and advocacy contexts. Its contributions are real and the intent here is to extend them, not to revise them.
The limitation is architectural rather than empirical. DARVO is actor-focused. The harm is produced by the actor's behavior. The intervention is directed at the actor. Even Freyd's institutional betrayal extension — which showed that organizations produce DARVO-like outcomes through structural risk-aversion rather than personal malice — retains the institution as the primary engine. The institution processes the complaint. The institution makes a decision. The institution's continued operation sustains the harm.
What neither framework describes is the phase that follows all active processing: the phase in which no identifiable actor is doing anything, and the inverted narrative sustains itself. This phase is common, consequential, and poorly described in the existing literature. Targets who have successfully held original actors accountable and still find the inverted narrative stubbornly intact are not experiencing a mystery. They are experiencing a phenomenon the frameworks were not designed to address.
The relevant theoretical precedents exist in adjacent fields. Path dependence in economics describes how initial conditions constrain subsequent options through self-reinforcing feedback. Labeling theory in sociology describes how applied classifications persist and shape subsequent interpretation of behavior. Information cascade theory describes how individuals adopt beliefs based on observed behavior of others rather than independent evaluation. Stigma research describes how reputational damage compounds through reputational inertia. Institutional momentum describes how organizations continue processing interactions through established frameworks regardless of their accuracy.
None of these is Distributed Maintenance. The contribution this paper is attempting is not to claim that any of these mechanisms is novel. It is to claim that they operate together as a unified maintenance system — that their combination produces a dynamic whose behavior is qualitatively distinct from any of its components, and that this system-level behavior has practical implications for both targets and researchers that the component-level frameworks do not fully capture.
The system-level behavior is this: a damaging narrative, once embedded in sufficient institutional and social infrastructure, becomes self-sustaining through mechanisms that require no continued active input from any originating actor, and the target's attempts to correct the narrative are automatically processed by the system as confirmation of it. That feedback loop — in which correction attempts reinforce the narrative being corrected — is not describable through path dependence, labeling theory, information cascades, or stigma research individually. It emerges from their combination. That emergent behavior is what this paper is attempting to describe.
III. The Five Mechanisms of Distributed Maintenance
The self-sustaining behavior of embedded narratives is produced by five mechanisms operating simultaneously. They are individually familiar from the literatures cited above. Their combined operation produces a system whose persistence and self-reinforcing character are more than the sum of the parts.
Cognitive ease. The brain processes established narratives more efficiently than contested ones. Daniel Kahneman's distinction between fast and slow cognitive processing is relevant here: a narrative already in place can be processed through System 1 — fast, automatic, associative. Evaluating a counter-narrative requires System 2 — slow, deliberate, effortful. The community is not malicious for maintaining the established narrative. It is rational. Accepting the pre-assembled account is cheaper than evaluating the contested alternative. This creates a persistent asymmetry between the cost of maintenance and the cost of updating that favors the established narrative regardless of its accuracy.
Conflict avoidance. Engaging with a target's counter-narrative requires a bystander to take a position in a contested dispute. Most people will not pay this social cost without strong incentive. Passivity defaults to the narrative with institutional backing. The silence that looks like neutrality is functionally indistinguishable from maintenance — it reproduces the narrative's social hold without requiring active endorsement. This is why socially expensive narratives are more durable than their accuracy warrants: challenging them costs more than accepting them, and silence sustains them.
Social centrality preference. When parties to a dispute have different social embeddedness — one more connected to the community's institutional and relational networks — bystanders default toward the more central party. This is a social cognition heuristic: people who are institutionally and relationally embedded are treated as more credible by default. The heuristic is generally useful. In the context of a loaded social system, it is exploited — either deliberately by actors who cultivate social centrality before deploying a narrative inversion, or inadvertently by institutional processes that weight the testimony of connected parties more heavily than isolated ones.
Reputational inertia. Once a person's reputation has been affected by a loaded narrative, subsequent behavior is interpreted through the lens of the damaged reputation. Behavior that would be evaluated neutrally in an undamaged reputation is evaluated as confirmation of the damaged one. A person whose reputation includes "aggressive" will have assertive behavior read as aggression. A person whose reputation includes "unstable" will have emotional expression read as instability. The inversion compounds through time without active maintenance. The target cannot escape it through behavioral change because the behavioral change is being evaluated through the frame the inversion established. This is not cognitive bias in the pejorative sense — it is the normal operation of social schema. Schemas are updated but updating is expensive, and reputational inertia is the cost of that expense.
Institutional momentum. Organizations process subsequent interactions with a person through the record established by prior interactions. A police report creates a framework for subsequent police interactions. A published article creates a reference point for subsequent editorial decisions. A committee decision creates precedent for subsequent committee behavior. The institution does not make a new decision to harm the target. It makes no decision at all. It continues operating according to its existing records. The harm is automatic. This is the mechanism that makes institutional correction so difficult: it requires an active decision to update a record, while maintenance of the record requires no decision at all. Active correction and passive maintenance are asymmetric in their cognitive and organizational costs.
These five mechanisms combine into a system with one distinguishing behavioral property that is not present in any of them individually: the target's correction attempts are automatically processed as confirmation of the narrative being corrected.
Here is how that feedback loop works. The community has been given a narrative: this person is unstable, aggressive, obsessive, unreliable. The person attempts to correct the record — persistently, because the harm is ongoing, because the institutions filtering their subsequent interactions are still running the corrupted data, because the correction is not yet reaching sufficient nodes in the maintenance system. The community observes the persistence. The persistence is evaluated through the lens of the established narrative. The established narrative described this person as obsessive and unable to let go. The persistence confirms the description. The correction attempt reinforces the narrative it is attempting to challenge.
This feedback loop is the most disorienting feature of high-maintenance-load cases for the people inside them. From outside the system, the situation appears static: the original incident is over, the original actor has often stepped away, nothing is visibly happening. From inside the system, the harm is continuous: every institutional interaction filtered through the corrupted record, every community response defaulting to the maintained narrative, every correction attempt recoded as confirmation. Targets who report that they cannot move on when the situation appears resolved from outside are accurately perceiving an ongoing process. The gap between that accurate perception and its appearance from outside is where much of the secondary psychological harm of these cases occurs.
IV. A Concrete Illustration
Consider a hypothetical that strips away all the particulars of any specific case.
A member of a professional community is accused of misconduct in a dispute that is contested. The accusation is publicized through institutional channels before investigation is complete. The target is temporarily suspended pending review. The review concludes that the accusation was not substantiated. The suspension is lifted. The original accuser does not retract publicly.
Twelve months later, the target applies for a position within the same professional community. Several committee members remember the suspension. None of them remember the outcome of the review. The institutional record of the accusation is accessible; the institutional record of the exoneration is filed in a different system and requires a specific request to retrieve. The committee defaults to the more accessible record. The target is not selected.
No one on the committee is malicious. No one is in contact with the original accuser. No one is actively maintaining the narrative of misconduct. The committee is doing its job with available information processed through the normal cognitive economy of a busy organization. Cognitive ease, conflict avoidance, reputational inertia, and institutional momentum are all operating normally. The target is experiencing Distributed Maintenance without any actor sustaining it.
This scenario is not exotic. Versions of it occur routinely in organizational, professional, civic, and legal contexts. Its commonness is precisely what the existing frameworks are not equipped to describe: it involves no ongoing malice, no identifiable villain at the stage when the harm is occurring, and no mechanism that can be addressed by holding anyone accountable. The people on the committee are not doing anything wrong. The institution is operating normally. The maintenance system is running on default settings.
V. The Walkaway Coefficient: From Active Harm to Maintained Harm
The transition from active narrative inversion to Distributed Maintenance is not automatic. It depends on whether the original actor strategically disengages — what this framework calls a high Walkaway Coefficient — or whether the actor remains actively engaged.
An actor who remains actively engaged can be challenged, documented, confronted. The active actor is visible. Their claims can be evaluated, their behavior can be documented, their institutional access can be addressed through formal process. The harm they produce requires their continued effort to sustain.
An actor who walks away converts a bounded harm into a potentially unbounded one. The walkaway hands the maintenance operation to a system that cannot be confronted because it has no face and requires no instruction. There is no one to challenge. There is only inertia.
This is why, counterintuitively, high-investment actors who are also high-walkaway actors are more dangerous in the long run than high-investment actors who remain engaged. The persistent actor can be addressed. The actor who loads the system and disengages has created a harm that does not depend on their continued operation.
The diagnostic implication is significant: when the original actor is no longer actively engaged but the harm continues, the appropriate analysis shifts from actor-focused frameworks to system-focused ones. Accountability processes directed at the original actor may be warranted on independent grounds, but they will not stop the maintenance system. The maintenance system requires a different intervention.
VI. The Soft Variant: A Subsection
Before describing the counter-intervention, one important variant of narrative inversion deserves acknowledgment, though a full treatment of it belongs in a separate paper.
The frameworks described here assume that narrative inversion requires some degree of self-protective motivation — malice, institutional risk-aversion, the need to sustain a particular account. This assumption fails in one identifiable class of cases: instances where the structural move is made by someone motivated by care rather than self-protection, with near-zero awareness of the mechanism, and with no investment in the inversion persisting beyond the immediate moment.
This occurs most visibly in intimate relationships under sustained external stress, where compassion fatigue — the depletion of the capacity for sustained empathic engagement with specific suffering — can produce structural deflection without any intent to harm. When a target of sustained narrative inversion presents a specific disclosure to a partner whose capacity for engagement has been depleted by years of proximity to that same campaign, the response may be generalization: everyone experiences this, you are not uniquely targeted. The specific disappears into the universal. The structural move is made. The motivation is care.
This variant is worth naming — structural deflection through compassion fatigue — because the appropriate response to it is completely different from the appropriate response to adversarial configurations. Relational repair is available. The correct tool is naming the structural effect without attributing malice, and making a specific request for the particular engagement that was deflected. Deploying the adversarial tools appropriate for high-investment, low-remorse configurations against a caring, single-instance deflection damages a repairable relationship without addressing the mechanism. The multidimensional model prevents this misapplication by requiring configuration reading before response selection.
A full treatment of this variant — its relationship to compassion fatigue theory, its distinguishing features, and its clinical implications — warrants its own paper. It is noted here because it completes the spectrum from fully malicious to fully caring and prevents the framework from implying that all narrative inversion is adversarial.
VII. The Counter-Record: Constructive Narrative Maintenance
The intervention logic follows directly from the maintenance mechanism.
Distributed Maintenance sustains itself by providing the community with a cognitively efficient default: a pre-assembled narrative with institutional backing that requires no active evaluation to accept, and that frames any challenge as confirmation of its own premises. The community accepts this default not through malice but through rational cognitive economy.
The counter-record intervention works — when it works — not by confronting the maintenance system directly but by competing with it on cognitive efficiency. The goal is to reduce the cost of updating the established narrative to something comparable to the cost of maintaining it. If the counter-record is accessible, organized, institutionally grounded where possible, and provides a coherent alternative that is easier to evaluate than the original inversion required, then the community has a genuine choice between two cognitively accessible narratives rather than between one easy default and one expensive challenge.
This is different from moral outrage, demands for acknowledgment from parties whose investment in the inversion makes acknowledgment unavailable, or confrontation with the original actor after the walkaway has occurred. All of those interventions address the origin of the narrative rather than its persistence mechanism. They may be appropriate on independent grounds. They do not stop Distributed Maintenance.
The counter-record has specific requirements to be effective as a maintenance competitor. It must be accessible — findable by the people whose updating matters. It must be organized rather than raw — a coherent account, not a collection of documents. It must be institutionally grounded where possible — because the community's credibility heuristics weight institutional backing, and a counter-record that can document judicial findings, official records, and neutral-witness testimony competes more effectively than a counter-record based solely on the target's assertion. It must be updated continuously — because the maintenance system continues running, and the counter-record must address the current state of the narrative, not only its original deployment.
The counter-record is not a guarantee of correction. Distributed Maintenance, once established with sufficient embedding, can sustain narratives for years or decades against competent counter-records. Historical examples are numerous. The counter-record is the only tool specifically designed to address the maintenance mechanism rather than the originating event, and it is the only available intervention once the actor has disengaged.
The people doing this work are not protesting or failing to move on. They are doing information design against a self-sustaining system. That is a precise description of a specific activity that requires specific tools. This framework is an attempt to make those tools legible.
VIII. Limitations and the Causal Attribution Problem
The most serious limitation of this framework is not empirical validation, though that limitation is real. The most serious limitation is causal attribution.
The framework describes a mechanism by which a narrative becomes self-sustaining after sufficient institutional and social embedding. It does not, as stated, provide reliable criteria for distinguishing a genuine maintenance system from a situation in which the original narrative was substantially accurate and the social system is correctly processing accurate information.
This is the question every critic will ask and should ask: how does one tell the difference between a target of Distributed Maintenance and a person who behaved badly, was accurately described, and is experiencing the normal social consequences of that accurate description?
The framework's partial answer is behavioral rather than evaluative. The diagnostic signatures of Distributed Maintenance are identifiable regardless of the narrative's accuracy:
The automatic recoding of correction attempts as confirmation of the narrative. This signature does not depend on the narrative's accuracy — it depends on whether correction attempts are evaluated on their own merits or processed through the lens of the established narrative.
The persistence of institutional filtering through corrupted records after the original actor has disengaged. This signature is identifiable through record analysis regardless of whether the record is accurate.
The asymmetry between the cognitive cost of updating and the cognitive cost of maintaining the established narrative. This asymmetry exists regardless of accuracy. True narratives are harder to update than false ones for the same reasons that false ones are.
These behavioral signatures help identify when the maintenance mechanism is operating but do not certify that the narrative being maintained is false. The framework is a tool for analyzing the mechanism, not for evaluating the content. Users of the framework should resist the temptation to treat Distributed Maintenance as evidence that the maintained narrative is inaccurate. It is evidence that the maintenance system is running. Whether what it is maintaining is true or false requires independent evaluation on the merits.
This limitation is not eliminable. Any framework that describes the persistence of narratives will be used by people who believe their narrative is the inaccurate one being maintained. Some of them will be right. Some of them will not be. The framework cannot adjudicate this. It can only describe the mechanism and recommend that both targets and observers evaluate the content of the narrative independently of the persistence mechanism — because persistence is not evidence of accuracy or inaccuracy, and the maintenance system operates similarly in both cases.
The second significant limitation is the multidimensional model. The nine dimensions proposed in this paper are derived from pattern analysis and existing theoretical literature, not from controlled empirical research. They are a structured analytical heuristic, not a validated clinical instrument. The dimensions may be incorrect, incomplete, or configured in ways that do not reflect how the underlying phenomena actually vary. The model should be used to generate better questions about specific instances of narrative inversion, not to certify configurations or diagnose individuals.
The third limitation is scope. The claim that Distributed Maintenance is a general theory of narrative stabilization rather than a DARVO-specific phenomenon is an extension that this paper does not fully demonstrate. The observation that the five maintenance mechanisms operate similarly across different narrative origins — deliberate inversion, institutional error, political construction, bureaucratic accident, historical myth — is plausible and worth investigating. Systematic empirical comparison across multiple narrative types and social contexts is required to ground the claim. This paper proposes the framework. It does not validate it.
IX. Conclusion: The Theory and What Follows From It
The central proposition is simple to state, though complex in its implications:
Narrative formation and narrative maintenance are analytically distinct processes, and the mechanisms that sustain an embedded narrative are largely independent of both the original actor and the original narrative's accuracy.
From this proposition, several things follow.
Holding the original actor accountable may be warranted on independent grounds, but it does not stop Distributed Maintenance. The maintenance system does not depend on the actor's continued operation.
The target's persistence in seeking correction is not evidence of pathology, inability to move on, or obsession with the past. It is a rational response to an ongoing process that is invisible to observers who cannot see the maintenance system running.
The appropriate intervention against Distributed Maintenance is not confrontation of the original actor but construction of a counter-record that competes with the established narrative on cognitive efficiency — reducing the community's cost of updating to something comparable to the cost of maintaining.
And the framework applies broadly. False accusations, true accusations, political narratives, bureaucratic errors, institutional records, and historical myths can all achieve Distributed Maintenance once sufficient social and institutional embedding has occurred. The maintenance mechanisms are features of social systems operating normally, not evidence of conspiracy, malice, or even error. Understanding them as system properties rather than as evidence of bad actors is what makes the framework potentially generalizable beyond any specific context.
The paper that follows from this one would attempt to test these propositions against systematic empirical data across multiple narrative types and social contexts. That is the work this framework is attempting to justify and motivate. Whether the framework survives that testing will determine whether it is describing something real or merely describing what it feels like to be inside a self-sustaining narrative.
That is the correct question to ask.
This paper is an attempt to make it askable with precision.
John F. Sendelbach is the originator of the Cold Cruel Sidestep behavioral framework and lead designer of the Pocumtuck State Park proposal. He is a metalworker, sculptor, and landscape designer based in Shelburne Falls, Massachusetts.