This article analyses workplace bullying as a form of antagonistic interrecursivity under asymmetrical interlegibility. Personal hostility becomes institutionally sustained when an authorised line manager converts private antagonism into formally legitimate action. Material and symbolic deposits then circulate beyond the original encounter, taken up by actors who need not share the originating intention. The resulting field remains largely occluded to the target, who must model a highly interrecursive network while being constrained by the bully’s procedural framing. Closure is institutionally blocked, producing recursive overload and maximal stress. Universities are particularly prone to this form of abuse because bullies can mobilise different types of actants (faculty with and without job security, administrators, and students on different levels). The way that universities are structured make certain forms of bullying practically invisible.
Part I: The Structure of the Problem
I. Introduction: Living Inside Unresolved Antagonism
A physical confrontation may be terrifying, but it usually has a relatively clear structure. The source of danger is present. The organism can flee, fight, submit, hide, or wait for the danger to pass. Even prolonged animal dominance relations involve comparatively direct antagonistic centres and recurrent behavioural signals: a rival is seen, assessed, and responded to, and the relationship, however tense, is conducted between bodies that are at least periodically co-present.
Human beings can instead remain caught for years in situations with a different shape entirely. The original antagonist may be perfectly identifiable, and yet the antagonist’s motives remain institutionally deniable. The distinction between personal hostility and authorised action cannot be resolved. Multiple institutional actors become involved, each with a partial and only loosely coordinated view of events. Each attempt to obtain clarification produces further uncertainty rather than less. The conflict persists through documents, meetings, records, rumours, and procedures even when the antagonist is physically absent, sometimes for weeks or months at a stretch.
Consider, in compressed form, a case that will recur throughout this article: a line manager in a university appears to pursue a personally motivated campaign against a member of staff, while acting at every point through powers formally delegated by the institution. The target may know exactly who the antagonist is. That is not, in fact, the difficult question. The difficult question is this: is this person acting as himself, as the institution, or as himself through the institution? The institution treats the manager’s conduct as authorised role performance. The target experiences the same conduct as personal antagonism expressed through institutional authority. Neither side, as a result, is modelling the same object, even though both are looking at the same sequence of emails, meetings, and decisions.
The full mechanism this article sets out to demonstrate can be stated as a single sequence, which the remainder of the article exists to substantiate step by step. A personal antagonism becomes an institutionally sustained hostile world when an authorised roleholder converts a private hostility into formally legitimate action; material and symbolic deposits carry that action beyond the original encounter; other institutional actors take up those deposits without sharing the intention that produced them; a radical asymmetry in what each side can understand about the other prevents the target from seeing how the institution is modelling them; and the resulting absence of usable feedback blocks the possibility of closure, forcing the target into continuing, metabolically costly work of trying to understand a situation that will not resolve. A compressed version of the same sequence will recur through the article: role capture, poietic extension, distributed uptake, interlegibility asymmetry, recursive blockage, and mesocosmic capture.
Workplace bullying in universities provides an especially clear illustration of this mechanism, and it will serve as the article’s main worked example throughout. But it is an illustration, not the article’s true subject. The deeper claim is anthropological: institutions make available a distinctively human possibility, the construction of antagonistic situations that are institutionally distributed, materially extended, symbolically amplified, and capable of organising large portions of a person’s life for years at a stretch. Other animals experience aggression, predation, dominance, expulsion, rivalry, deception, and sustained vigilance. No currently documented nonhuman species inhabits a hostile field composed of roles, allegations, records, procedures, reputations, committees, appeals, unanswered correspondence, hidden deliberations, delegated powers, and competing institutional jurisdictions.
The argument unfolds in five parts. Part I sets out the basic structure of the problem: what makes interrecursivity antagonistic, why some antagonism stays tractable while other antagonism does not, and what it means for two parties to be radically unequal in what they can understand about each other. Part II follows the antagonism from its origin in a personal relation into its first institutional form, as a person’s private project is converted into authorised action and then into a durable record. Part III follows that record as it is taken up by an institution that was not present at its creation, tracing how classification, evidence, and silence propagate the antagonism far beyond the person who began it. Part IV describes what this does to the person who has to live inside it: a mind forced to keep modelling a field that will not settle, and a body that pays the cost. Part V asks where this mechanism is most likely to occur, what makes it distinctively human, what should not be mistaken for it, and what a fairer process would actually require.
II. Antagonistic Interrecursivity
The vocabulary ordinarily available for this territory — aggression, competition, dominance, coercion, conflict, hostility, cruelty, bullying — describes actions, relationships, motives, or outcomes. None of these terms identifies the specific recursive structure that makes certain antagonistic situations uniquely difficult to escape or even to understand while one is inside them. Aggression can be nonrecursive or only minimally recursive: a predator strikes prey without needing to model what the prey believes about the predator’s own intentions. Even strategic competition does not necessarily involve the sustained mutual modelling across several distinct centres that is this article’s actual subject.
A recursive centre, in the sense meant here, is anything whose next action depends on an internal process sensitive to what it expects from another such centre. Interrecursivity is what occurs when two or more such centres model each other: each is asking, more or less continuously, what the other believes, what the other thinks it believes, what the other is trying to make it do, and what model of it the other is forming and communicating onward.
Not all interrecursivity under opposed interests deserves the name antagonistic. Negotiation, litigation, sport, and ordinary bargaining all involve parties pursuing incompatible outcomes, and none of them is what this article is about. The distinction that matters is this: antagonistic interrecursivity occurs when at least one recursively responsive centre organises its conduct around obstructing, subordinating, discrediting, extracting from, excluding, or otherwise damaging another centre’s capacity to coordinate — not merely around winning within a shared field, but around impairing the other party’s standing, agency, credibility, livelihood, or ability to sustain an ordinary life. This can be distinguished from two adjacent forms. Competitive interrecursivity involves parties pursuing incompatible outcomes within a field both accept as legitimate. Adversarial interrecursivity involves conflict that is explicit and bounded by agreed rules, as in litigation or formal debate. Antagonistic interrecursivity is neither: the damage sought is not incidental to winning, and the rules bounding the encounter are, as the following sections show, exactly what becomes unstable.
It helps to have a compact vocabulary for how such a situation typically develops, since the article will return to each stage repeatedly. A primary antagonism is the originating opposition between identifiable parties — in the recurring example, a manager’s project against a colleague. An antagonistic intervention is the use of a role, category, procedure, document, or institutional pathway to advance that project. Institutional propagation is the uptake and continuation of that intervention by actors who may not share, or even recognise, the original antagonism. An antagonistic field is the resulting distributed configuration of actors, records, procedures, interpretations, and unresolved questions. A hostile mesocosm, finally, is the condition in which that field reorganises the target’s embodiment, relationships, surroundings, and symbolic environment — the whole mediated world within which an ordinary life is conducted. The institution as a whole need not become antagonistic in any intentional sense for its ordinary operations to become, cumulatively, antagonistically consequential. That distinction does a great deal of work in what follows.
It would be a mistake to present this territory as though it had gone unstudied. Workplace mobbing — the systematic, prolonged, health-damaging harassment of a target by one or more colleagues — was named and first studied in detail by the German-Swedish psychologist Heinz Leymann in the 1980s, who distinguished it from ordinary conflict precisely by its duration, its patterning, and its capacity to erode a target’s standing and wellbeing over months or years. Kenneth Westhues, a sociologist who has spent more than two decades researching the university case specifically, has documented dozens of instances of academic mobbing and has argued that it disproportionately involves colleagues who are unusually productive, idealistic, or independent-minded — precisely the people whose success makes them locally threatening. This is real, well-documented territory, and nothing here claims to have discovered the phenomenon of prolonged, institutionally embedded hostility at work.
Westhues’s finding fits the mechanism developed here, though it should be stated as a pattern the mechanism helps explain rather than as something the mechanism directly predicts. A target whose competence, popularity, or independence gives them a strong claim to credibility is precisely the target whose case most acutely strains the institution’s ordinary presumption that roleholders act in good faith, a presumption discussed at length in Part II. An unremarkable colleague can be quietly managed out with comparatively little institutional strain. A colleague with a strong record generates a harder recursive problem: the institution must either accept that a well-regarded manager is targeting a well-regarded colleague for reasons unrelated to either person’s merit, which is difficult to hold in mind, or resolve the discomfort by finding some way to make the target’s own conduct the explanation instead. The harder a case is to explain through ordinary managerial logic, the more institutional pressure exists to explain it through some fault in the target.
What the existing literature on mobbing has not yet supplied is a precise account of the recursive mechanism that makes this particular form of antagonism so much harder to resolve, and so much more costly to sustain, than an ordinary dispute between two people who can see each other clearly. That mechanism turns on a second concept, distinct from antagonistic interrecursivity itself: asymmetrical interlegibility.
III. Tractable and Intractable Antagonism
Antagonistic interrecursivity comes in two forms that matter enormously and are easy to conflate. In tractable antagonism, the opponent is identifiable, the rules governing the encounter are sufficiently legible to both sides, feedback is available, and new information progressively improves each party’s model of the other rather than multiplying the number of possible interpretations. A chess match, a boxing match, a courtroom exchange, a formal debate: the parties may be fiercely opposed, sometimes for years, but the field of antagonism remains bounded, and each party can, in principle, arrive at a stable and testable understanding of what the other is doing and why.
In intractable antagonism, none of this holds: This distinction is the organising axis of the whole article. Parts II through IV can be read as an account of the specific mechanisms — role capture, material and symbolic extension, institutional propagation, occlusion, evidential asymmetry — by which an antagonism that begins tractable, a disagreement between two people who can in principle be checked against each other’s accounts, is converted into something recursively intractable.
It is worth being honest about the limits of this distinction before relying on it further. Recursive intractability, on its own, does not prove that a hostile intention originated the process. Institutional confusion, ordinary delay, fragmented responsibility, incompatible policies, genuine disagreement, administrative incompetence, a target’s own misreading of ambiguous events, and mutual escalation between two people who are each acting in good faith can all produce something that looks, from the inside, very much like the pattern described here. The theory developed in this article should not be read as licence to assume antagonistic intent merely because occlusion and uncertainty are present. The diagnostic question that has to be asked in any actual case is: what evidence distinguishes an institutionally propagated antagonism originating in selective hostile action from an impersonal coordination failure that produces similar experiential effects? Relevant indicators include patterned selectivity rather than isolated lapses; departures from how comparable cases are handled; asymmetric application of the same rule to different people; timing that tracks the antagonist’s interests rather than institutional need; repeated benefit accruing to the same actor; suppression or discouragement of independent review; evidence that contrary information was known and excluded rather than simply missed; and anticipatory framing of the target that precedes any documented wrongdoing on their part. Where these indicators are largely absent, what remains may still be worth fixing, but it is a different problem from the one this article is describing.
IV. Interlegibility and Its Asymmetries
Interlegibility is not simply visibility, information, or surveillance. It is the degree to which recursive centres can understand one another’s intentions, one another’s interpretations of events, the models others hold of them, how those models are being communicated onward, how those models are likely to affect future action, and which centres are connected to which others. A person may be highly visible to an institution, in the sense that a great deal of factual information about them is recorded and available, while the institution itself remains deeply illegible to that person, in the sense that how it is interpreting, weighing, and circulating that information cannot be observed at all.
It is worth being precise about why legibility alone — one party simply knowing more facts about the other — is not sufficient to generate the phenomenon this article is concerned with. A surveillance system can make someone highly legible, in the sense of recording everything they do, without any interrecursivity being involved at all; a camera does not model the person it records, any more than a river models the fish swimming in it. Interlegibility is the operative concept precisely because the problem concerns not merely what an institution knows about a person, but what it thinks the evidence means, whom it considers credible, how its internal actors understand one another, whether they are coordinating, whether they are protecting one another, and whether a given silence is accidental, procedural, strategic, or itself a form of hostility.
Interlegibility is also not a single quantity. It has several dimensions that can vary independently of one another: intentional legibility, whether motives can be inferred; interpretive legibility, whether one can see how the other understands the event; network legibility, whether one can see which actors are connected to which; procedural legibility, whether one can understand how the process moves; evidentiary legibility, whether one can see how evidence is being weighed; futural legibility, whether one can anticipate likely next actions; and reflexive legibility, whether one can know what model others hold of oneself. An institution may disclose its procedure while concealing its interpretation; disclose evidence while concealing how that evidence is weighted; name the decision-makers while concealing the informal influence acting on them. Asymmetrical interlegibility rarely means total darkness on one side. It usually means partial and uneven light, arranged so that the party with less power is denied exactly the dimensions that would let them predict what happens next.
A further asymmetry compounds this. Because the target is one node facing a network, while the institution can model the target through a single portable category, the two sides face radically different modelling burdens. The target is expected to model the manager, human resources, senior management, colleagues, any relevant committee, legal risk, and the reactions of people who will only ever encounter a file. The institution, by contrast, may model the target through one label: difficult, a risk, a complainant. One party must model an entire network; the network models the party through a label. Call this interlegibility bandwidth: the amount of another recursive centre, or field of centres, that a party can hold in view at once. The institution achieves a low-cost symbolic compression of the target at the price of the target’s high-cost recursive expansion of the institution. This asymmetry of bandwidth, developed further as the article proceeds, is one of the central mechanisms by which formally equal parties end up in a radically unequal relation.
The experience this article is describing has a serious literary and philosophical history, even where that history has not supplied a mechanism for it. Kafka’s fiction, and The Trial above all, is substantially an extended depiction of exactly the target’s side of this condition: an institution whose structure, and even whose coherence as a single deciding agent, cannot be established by the person subject to it. Hannah Arendt’s account of Adolf Eichmann illuminates a different side of the same structure, asking how personal moral responsibility can be traced through a bureaucratic chain in which everyone involved can plausibly describe themselves as merely following procedure — the actor’s side, where role performance appears to dissolve personal responsibility, rather than the target’s side, where role performance conceals it. Neither offers a mechanism in the sense this article attempts to supply, but a theoretical account of this territory that did not engage them would be missing serious existing company.
Part II: How Personal Antagonism Becomes Institutional
V. Person, Role, and Role Capture
Every officeholder occupies at least two positions simultaneously. A line manager is a person, with preferences, resentments, ambitions, friendships, rivalries, and private interpretations of the people around them, and is also an authorised institutional role, carrying delegated powers and duties that exist independently of who happens to occupy the position at a given time. Ordinary institutional functioning depends on treating the roleholder’s actions as expressions of the role rather than of the person; this is a necessary practical shortcut, since an organisation could not function if every managerial act had to be individually audited for hidden personal motive before anyone could rely on it.
It helps to separate three levels that ordinarily run together without conflict. Personal recursion is what the roleholder wants, fears, resents, or seeks, as an individual with a private history and private stakes. Role recursion is the authorised expectations and powers attached to the office itself, independent of who fills it. Institutional recursion is how the various offices around that role respond to one another and to the accumulated record over time. The antagonistic possibility this article is concerned with arises because these three levels can diverge sharply while continuing to appear aligned: a person may use role recursion to advance a personal project, while downstream institutional recursion treats the resulting deposits as ordinary outputs of the role. The personal motive disappears from view; its consequences are institutionally multiplied.
This dual position becomes a real problem when a person uses authorised institutional powers to pursue what is, at bottom, a personal antagonism — call this role capture. The conduct involved can include initiating procedures, escalating complaints, interpreting ambiguous rules in a consistently unfavourable direction, setting meetings at difficult times, recording concerns in a file, questioning performance repeatedly, invoking the language of welfare or risk, or describing ordinary conduct through institutionally loaded terms. Each of these actions, considered individually, may fall entirely within the manager’s formal powers. The issue is not necessarily rule-breaking in any narrow sense. It is the recruitment of legitimate authority into a private project, a recruitment that is difficult to detect precisely because each individual act looks exactly like the ordinary exercise of a job.
The result is that the two parties come to hold two fundamentally incompatible objects in mind, not merely two different opinions about the same object. The target experiences a person using an office. The institution, reviewing the same sequence of events, experiences an office occupied by a person. The target sees a pattern of selective or strategically coordinated actions, visible only when the individual acts are considered together over time. The institution sees a sequence of individually authorised acts, each defensible on its own terms and reviewed, if at all, one at a time. This is why the two sides can disagree completely without ever disagreeing, in the ordinary sense, about whether specific events occurred. They agree on the events. They cannot agree on what kind of thing the sequence of events is.
It follows that the institutional role does not only supply the power required to act. It also renders the underlying motive illegible, and does so more effectively the more closely the hostile action resembles legitimate managerial conduct. One consequence of this, worth stating plainly because it runs against intuition, is that the most institutionally effective way of pursuing a private antagonism may be to violate the fewest possible explicit rules, since every rule violation is a potential point of independent scrutiny, while conduct that stays within the letter of managerial discretion generates no scrutiny at all. This is the first and most important instance of a principle that recurs throughout the article: antagonistic consequence can become autonomous from antagonistic intention, propagating through actors who neither share nor recognise the motive that produced it.
VI. The Presumption of Role Fidelity
An organisation cannot operate by treating every roleholder as a potential rogue actor at every moment. It therefore begins from a necessary practical assumption: people occupying institutional roles ordinarily exercise their powers for institutional purposes. This assumption is not a failure of imagination or a design flaw to be regretted. It is an indispensable muting of institutional workload, a settlement that allows the overwhelming majority of an institution’s daily functioning to proceed without effortful re-examination of every decision. Without it, no organisation of any size could function at all.
The pathology begins not with the presumption itself but with what happens when accumulating evidence should reopen it and cannot. The institution does not merely conclude, on the balance of evidence, that a particular manager is innocent of wrongdoing. It frequently lacks a ready category into which the alternative possibility — a manager using formally legitimate powers primarily to pursue a personal hostility — can even be placed for consideration. Evidence is processed under a framework with no ready slot for the very phenomenon it would need to detect, so that even clear evidence tends to be reinterpreted to fit the available categories rather than treated as grounds for expanding them.
A recurring feature of these situations is worth naming directly: the judgement that a manager is, in the words commonly used, a nice guy, performs a kind of premature settlement. It converts a difficult and effortful question — could this specific, currently agreeable person nonetheless be doing this? — into an already-stabilised reputational category, letting the institution stop modelling the manager’s possible motives and rest on a prior impression instead. The institution substitutes a sedimented model of the person for renewed attention to the present situation. This is not to say reputation is irrelevant; it is that a prior attribution is allowed to override current coordinative evidence, converting reputation from one piece of evidence among others into something closer to an ontological defence.
This asymmetry is not an accident of any particular institution’s culture. Institutions are generally well designed to detect unauthorised actions, explicit procedural breaches, prohibited conduct, failures to document properly, and discrimination expressed in recognisable forms, because these are precisely the categories institutional oversight mechanisms were built to catch. They are considerably less well equipped to detect the selective and antagonistic deployment of powers that are, individually, entirely legitimate. There is a meaningful difference between violating a procedure and weaponising one, and most institutional machinery is built to notice only the first.
Something similar operates at the level of institutional bandwidth more generally, and it is worth naming before the article turns to specific mechanisms. Formal processes are built to register certain kinds of object well: a specific allegation, an applicable policy, a procedural decision, a reputational classification. They are far less able to register a slow deterioration of ordinary working coordination before any single event is provable; an unease that precedes proof; a case-specific account that does not map cleanly onto an existing category; or a question about whether the institution’s own machinery is producing the harm rather than merely failing to catch it. Nor can such processes easily hold together, in a single assessment, the bodily exhaustion, the damaged working relationships, the loss of an ordinary workplace, threatening correspondence, and reputational classification that together constitute what has actually happened to a person. The problem here is not simply negligence. The institution’s machinery has insufficient bandwidth to perceive the object the complainant is describing, and this bandwidth limit, rather than any single actor’s bad faith, explains a great deal of what follows.
VII. Recursive Single Points of Failure
Line managers often serve, in practice, as the first interpreters of staff conduct, the first source of performance information reaching higher levels of the institution, the initiators of formal procedures, institutional witnesses to events nobody else observed directly, the channel through which concerns first reach human resources, and the interpreters of local context that more distant offices have no independent means of assessing. The institution, in other words, delegates a substantial part of its own perception to the manager, trusting the manager’s account as a proxy for direct observation.
When the manager is, in fact, acting antagonistically, the same person who is producing the distortion is also the institution’s authorised source of information about the target. This creates a recursive single point of failure: a node in the institution’s perceptual apparatus whose corruption is not visible to the downstream actors relying on it. Four conditions together make a node this kind of failure point rather than a mere informational bottleneck: interpretive priority, meaning the node frames the case first, before anyone else has an independent account; authorised credibility, meaning its account enters the system as rolelegitimate rather than as one claim among others; downstream dependence, meaning later actors rely on that framing rather than building their own; and absence of independent feedback, meaning the target has no channel by which to correct the model before it sediments into institutional fact.
Human resources trusts the authorised manager’s account, because that is what the role is for. Senior leadership trusts human resources, because that is what the escalation structure is for. Committees trust the case file assembled for them, because assembling a case file is precisely what a committee’s supporting staff exist to do. The distortion, once introduced at the point of origin, is not merely preserved as it moves through this chain. It is amplified, because every subsequent actor is behaving rationally according to the ordinary logic of institutional trust. This framing matters because it allows the phenomenon to be described without requiring that every actor in the chain share the original hostile intention, or even be aware that anything is wrong. The system can reproduce and extend an antagonism that most of the people implementing it neither intended nor recognise.
A recursive single point of failure does more than transmit a distorted first account, however. It also tells later actors how to interpret anything the target does to contest it, and this is what makes it self-protecting rather than merely self-perpetuating. Call this recursive inoculation: an early authorised framing that determines in advance how future attempts at correction will themselves be read, thereby protecting the original model from revision. Once a target has been framed as difficult, their later insistence becomes further evidence of difficulty. Concern about bias becomes evidence of paranoia. Careful documentation becomes evidence of obsession. Visible distress becomes evidence of instability. The target’s most reasonable responses to an unfair process are recoded, by the very framing that made the process unfair, as further confirmation of the framing.
VIII. Poietic Extension: Documents, Categories, and Recursive Decontextualisation
Most non-human antagonism depends heavily on bodies sharing space at the same time: attack, pursuit, display, exclusion, submission, territorial defence. The antagonistic relation may recur across many separate encounters, but it is not, in any non-human case currently documented, extensively externalised into durable symbolic and material infrastructure that persists and continues to act independently of the antagonist’s own physical presence.
Human beings, distinctively, can convert a live, mutually responsive episode into a relatively fixed deposit — an email, a file note, a classification, a decision — that can later be taken up by people who were nowhere near the original encounter. This conversion does something specific: it does not preserve the original interaction. It preserves constraints, traces, and orientations capable of starting new interpretive processes later, while stripping away the tone, sequence, prior provocation, and reciprocal adjustment through which the original meaning was actually generated. In institutional conditions, antagonism is carried by emails, letters, policies, complaint forms, meeting invitations, personnel files, case-management systems, calendars, minutes, office architecture, access permissions, reporting lines, and archived correspondence. The antagonist no longer needs to be present for the antagonism to continue exerting its effect. A message received at home in the evening can reactivate an entire antagonistic field that had, moments before, been quiet. A calendar notification can produce anticipatory distress days before the meeting it announces has even occurred.
Call the specific loss involved in this conversion recursive decontextualisation: the reduction of a living, mutually responsive process into a durable deposit that can circulate without the field of responses through which its significance was originally generated. The document does not preserve the manager’s motive; it preserves an allegation or a classification. Later actors encounter the deposit without access to the recursion that produced it, and generate new interpretations from the deposit alone. The document can therefore continue the antagonistic consequences while concealing the original antagonism. This is why documents are simultaneously necessary and dangerous to institutional fairness: they permit institutional memory, but they preserve an asymmetric cross-section of the interaction, discarding almost everything that would let a later reader judge it fairly.
It would also overstate the case to say, as it is tempting to say, that institutions remember through artefacts, full stop. An artefact does nothing on its own. Each act of institutional remembering is a fresh act of uptake — the document must be read, interpreted, relied upon, or reactivated by someone, at some later point, doing something that looks entirely ordinary from the inside. Antagonistic continuation is not automatic. It is repeatedly reenacted through apparently unremarkable acts of reading, forwarding, classifying, and relying on records that most of the people performing them will never think of as taking a side.
One further consequence of this materialisation deserves emphasis, because it complicates any simple story about individual malice. Once an antagonistic intervention has been materialised — in a file, a formal record, a reputational label attached to a case — its effects no longer require the continuing active hostility of the original antagonist to persist. The file itself can continue organising institutional behaviour toward the target regardless of whether its author still wishes to pursue the matter, has moved to another role, or has left the institution altogether. Human institutions, in other words, can continue a hostile process that no single actor is any longer actively willing.
Part III: How Institutions Propagate Antagonism
IX. Categories, Symbolic Surplus, and the Reallocation of Relevance
Institutional categories do not merely describe an event after the fact. Terms such as safeguarding, inappropriate conduct, unprofessional behaviour, risk, harassment, or fitness to practise place an event onto a specific institutional pathway the moment they are applied to it, and different pathways carry very different consequences. Classification does more than raise the amount of attention an event receives, though it does that too. It reorganises the field itself, in at least four ways at once: it confers relevance on the allegation; it confers relevance on some institutional actors and not others; it withdraws relevance from alternative descriptions of the same events; and it changes the future relevance of prior events, which are now reread through the new category. Earlier ambiguous incidents may acquire new significance retrospectively, while contextual information that does not fit the activated pathway may lose relevance altogether, however important it seemed a week before.
Certain classifications trigger rapid escalation, wider internal circulation, precautionary reasoning that treats ordinary ambiguity as potential risk rather than an open question, additional documentation requirements, and the involvement of specialist offices that would otherwise have no role in the matter at all. A managerial conflict that began as an ordinary disagreement can be transformed by associating it, even loosely, with welfare concerns or professional impropriety. It is not necessary to claim that all such reframing is deliberately manipulative in order to make the structural point: institutional systems respond to the category applied to an event, not to the event itself prior to categorisation, which means the act of categorisation is doing real causal work regardless of the intentions behind it.
Terms such as inappropriate and unprofessional also carry a symbolic surplus, a field of implication considerably larger than their literal content. Such phrases can suggest, without stating outright, boundary violation, danger to others, poor judgement, or unfitness for professional trust, and the indirectness of the language is part of what makes it effective: it does antagonistic work while remaining vague enough to resist precise rebuttal. This works because ordinary symbolisation hides, on its surface, whether a term is naming a specific event, coordinating an institutional response, gesturing at an absent pattern, or simply imposing an institutional classification. A portable category falls through this ambiguity and appears to name something directly present: inappropriate conduct is received not as one possible abstraction from contested events but as the thing that happened. Once a negatively loaded category has been attached to someone, they must respond not only to whatever factual allegation lies underneath it, but to the entire field of possible meanings the category itself has activated. The party making the accusation can remain comparatively noncommittal about which meaning is intended. The accused party must attempt to disprove an expanding cloud of implication that no single piece of evidence can fully dispel.
X. Nested Institutional Interrecursivity and the Accumulation of Stakes
The university, considered as a single recursive centre, is a convenient simplification that does not survive close examination. It consists instead of numerous partially connected centres: line management, school leadership, human resources at several levels, senior management, legal advisers, trade unions, committees, and specialist safeguarding personnel. Each of these centres models the case under review, and each also models the other centres involved in reviewing it, which is where the real complexity begins.
At the outset of a case, the stakes are unevenly distributed: the target carries high livelihood, reputational, relational, and bodily stakes, the institution carries relatively modest procedural stakes, and the manager may carry personal and reputational stakes of their own. But once a complaint is escalated beyond its point of origin, new parties acquire new stakes that have nothing to do with the original dispute. Human resources acquires a stake in its own competence. Senior management acquires a stake in institutional consistency. Legal services acquires a stake in liability. The institution as a whole acquires a stake in precedent. And at every level, once a judgement has been made, that level acquires a stake in the correctness of its own prior judgement. This produces a face-saving cascade: a higher level reviewing a lower level’s handling of a case is rarely independent of that lower level in any meaningful sense, since acknowledging an original failure now threatens an entire chain of institutional relationships rather than a single decision. Every additional level of review can increase, rather than decrease, the number of actors whose own coordination now depends on the original interpretation remaining defensible. More review does not automatically produce more impartiality; it produces more impartiality only when the reviewing levels are genuinely independent of the decisions, reputations, and relationships already invested in the outcome.
None of this requires that the actors involved share the original antagonistic intention, and it is worth separating two things that are easy to run together. Distributed intention would mean that multiple actors consciously share an antagonistic project. Distributed consequence means that their independently motivated, individually reasonable actions jointly reproduce the antagonistic field. The case developed here generally requires only the second, and this matters for reasons beyond fairness to the individuals involved: it means the process is not a conspiracy that could be exposed by finding the right hidden agreement. The more disturbing possibility is that ordinary institutional rationality, exercised by people acting in good faith according to their own local incentives, is sufficient on its own to reproduce what began as one person’s antagonism.
XI. Recursive Occlusion and the Proliferation of Hypotheses
The target of an antagonistic institutional process typically does not know who discussed the case internally, which documents were circulated to whom, whether informal advice was given behind the scenes, why one level of the institution referred the matter to another, or whether different actors within the institution hold meaningfully different interpretations of the same events. It is worth being precise about what is actually missing, because it is not information in general. The target may know exactly what decision was issued, which policy was invoked, and what allegation was formally made. What remains occluded is who interpreted what, how models circulated, whose judgement influenced whom, whether the relevant actors agreed with one another, what future action is being considered, and how the target’s own responses are being read. Secrecy withholds content. Recursive occlusion withholds the organisation of modelling itself — the far more consequential thing.
Silence, in this setting, is not the absence of recursive effect; it is a specific and asymmetric way of allocating the labour of interpretation. The institution can pause its attention to a case at essentially no cost, simply by not responding. The target cannot pause, because their livelihood, reputation, and future action remain dependent on a process they cannot see. An unanswered appeal does not suspend interpretation on the part of the person waiting for a response; it forces them to generate an entire set of possible explanations to fill the gap. Perhaps the appeal was lost administratively. Perhaps it is being actively considered. Perhaps the recipient privately disagrees but has not said so. Perhaps legal advice is being sought. Perhaps a lower level is being protected from further scrutiny. Because no feedback arrives to eliminate any of these possibilities, silence produces proliferation rather than settlement, and it does so by transferring attention, uncertainty, and the entire burden of anticipation onto the person least equipped to carry it at low cost.
Each failed attempt at obtaining clarity or repair tends to generate a further antagonistic centre that must itself now be modelled, and the expansion this produces is not linear but closer to combinatorial. The original question may be no more complicated than why a line manager is behaving in a particular way. After an appeal goes unanswered, a second question appears: why is human resources not responding. After a senior leader dismisses the concern without apparent consideration, a third: why does this person refuse to consider the possibility that the manager’s conduct was not simply ordinary management. The target attempting to make sense of the situation must eventually try to model human resources’ view of the manager, the department head’s view of human resources, whether colleagues are quietly influencing the department head’s thinking, and whether an extended silence reflects real alignment among these actors or simple confusion among people who have not actually spoken to one another. The recursive field the target must hold in mind does not grow by adding one actor at a time. It grows by adding relationships between actors, each of which is itself uncertain.
XII. Asymmetrical Evidencing, Pattern Destruction, and Forced Misarticulation
In ordinary inquiry, a relevant new fact eliminates some hypotheses and lends support to others, which is precisely why gathering more information is usually the correct response to uncertainty. In antagonistic institutional situations, evidence can fail to perform this function, and can sometimes do the opposite. The asymmetry involved is worth disaggregating rather than describing as a single vague unfairness, because its component parts recur across very different institutions. There is an asymmetry of admissibility: what is allowed to enter the process at all. An asymmetry of burden: who must prove what, and to what standard. An asymmetry of granularity: whether individual acts or longitudinal patterns are what actually gets assessed. An asymmetry of amplification: what kind of evidence triggers escalation and what kind does not. An asymmetry of credibility: whose account enters the record as an authorised interpretation and whose enters as a contested claim requiring separate verification. An asymmetry of response: which evidence requires visible institutional action and which can simply be noted. And an asymmetry of explanation: whether evidence that is set aside must be accounted for, or can simply be left unaddressed.
A minor alleged infraction on the target’s part can receive substantial institutional amplification, while a significant apparent conflict of interest or rule violation on the manager’s part produces comparatively little visible response, and this differential amplification does not require deliberate bias to occur; it follows directly from the role asymmetry described in Part II, since the manager’s account enters the record as an authorised interpretation while the target’s account enters it as a contested claim. Faced with this pattern, the target is eventually forced to ask a more corrosive question: not only what happened, but what counts as evidence here, and why evidence appears to have radically different effects depending on whom it concerns.
One specific evidentiary operation deserves to be named as its own mechanism, because it is doing more work than mere bandwidth limitation. Pattern destruction is the procedural decomposition of a longitudinal coordination into isolated events, each of which becomes explainable in ordinary terms once severed from the sequence that gave it its antagonistic significance. A hostile campaign often consists precisely of acts that are innocent singly and diagnostic cumulatively — one meeting here, one complaint there, one email, one performance concern, each reviewed largely in isolation. An investigative process that insists on atomisation is structurally incapable of seeing the phenomenon it is meant to detect, however carefully each individual act is examined.
A further operation compounds this. The target is not merely denied a voice; they are usually required to speak through forms that systematically distort what they are trying to describe. A personal campaign must be redescribed as a series of individual procedural errors. A longitudinal pattern must be divided into discrete, dated incidents. Motive may be declared irrelevant to the process. Cumulative harm may have to be rendered as a checklist. Call this forced misarticulation: the target is offered participation on condition that the phenomenon be redescribed in the institution’s own terms. They may speak, but not of a private antagonism operating through authorised discretion, if the only available form recognises discrete procedural breaches. The account is admitted only after being converted into a shape that removes the very pattern it was meant to reveal, so that participation and distortion become, in practice, the same thing.
Part IV: How the Antagonistic Field Captures a Life
XIII. Recursive Blockage and Recursive Overproduction
A disturbance in ordinary working coordination is usually felt before it can be fully explained, and explained before it can be formally submitted anywhere. In a healthy process, that submission returns, eventually, to repaired coordination: the disturbance is named, addressed, and the working relationship is restored to something workable, even if not to what it was before. In the situations this article is describing, that return does not happen. The felt disturbance is articulated, submitted into an institutional process, and then the institutional process either mistranslates it, fragments it, forces it into a distorting category, or simply treats its own existing categories as more authoritative than the lived coordination the complaint arose from. Instead of returning to repaired working relations, the process becomes stuck: the unresolved case falls back into renewed disturbance and further attempts at articulation, which are met, again, with distortion or refusal. Call this recursive blockage: a felt problem that can be sensed and put into words, but that cannot move through the institutional stages required for repair.
Under blockage, the target is left generating more possible explanations than any available feedback can test. Recursive overproduction is the proliferation of models beyond what the situation supplies the means to eliminate. This is not excessive cognition in any psychological sense requiring correction. It is a structurally induced surplus of interpretation, produced jointly by occlusion, silence, and blockage, and it connects directly to the metabolic cost discussed below: a mind that cannot stop generating hypotheses it also cannot test is a mind doing real, continuous, unpaid work.
It is worth stating a further qualification precisely, because it is easy to collapse into either of two false comforts. The modelling activity forced on the target by recursive occlusion can be entirely rational and necessary, even though many of the specific hypotheses it generates turn out to be false. The tragedy here is double, not single: the person cannot safely stop modelling, because their livelihood and standing genuinely depend on getting some of it right, and at the same time the conditions available to them make it structurally impossible for the modelling to become reliably accurate. Recognising the necessity of the modelling should not be mistaken for endorsing the accuracy of any particular thing it produces.
XIV. Mixed Futurity, Recursive Capture, and the Cost the Body Bears
Recursive closure, in the sense meant here, does not require agreement, reconciliation, or certainty about every disputed detail. It requires two things together: a model settlement, meaning a sufficiently stable understanding of what happened and what the parties are likely to do, such that further encounters confirm the model more often than they overturn it; and a coordinative settlement, meaning a sufficient alteration of the relationship, the power balance, the pathway, or the future stakes involved, such that continual remodelling is no longer required to protect what matters. A person can, in principle, reach a stable and accurate understanding of an antagonist’s behaviour and still be unable to disengage, because that antagonist retains ongoing power over their future. Closure, on this account, does not require reconciliation, confession, or full agreement; it requires that both understanding and consequence have settled enough that constant vigilance is no longer necessary.
Chronic antagonistic interrecursivity is characterised precisely by the structural prevention of this double settlement: every new encounter reopens the model rather than confirming it, and every new procedural step leaves the underlying stakes exactly as unresolved as before. A person caught in this situation will typically be found replaying meetings, rereading emails, reconstructing timelines, modelling the likely motives of several people at once, examining the exact wording of a message for implications that may or may not be intended, anticipating future moves several steps ahead, and trying to identify alliances that may or may not actually exist. Recursive capture is the condition in which this unresolved modelling becomes sufficiently tied to the person’s future that attention cannot safely disengage from it, so that the conflict comes to organise an expanding share of everything the person does, whether or not they are currently at work. This is not the same thing as rumination, obsession, hypervigilance, anxiety, or trauma, though it can produce experiences that overlap with all of these; those terms describe possible effects, while recursive capture names the relational and institutional mechanism that keeps the modelling active in the first place.
Part of what makes this so exhausting is that the target is not modelling a single uncertain future but several incompatible ones at once. Dismissal, vindication, procedural delay, further escalation, quiet settlement, career departure, exposure of the antagonist, further accusations: each of these possible outcomes demands a different present response. Should one respond in writing or stay silent, gather further evidence or let the matter rest, contact a union representative or wait, seek legal advice or hope the process resolves itself, warn colleagues or say nothing, resign or hold on. Call this mixed futurity: several incompatible anticipated futures organising the same present, none of them ruled out by the information actually available. The person must remain prepared for several mutually exclusive futures at once, without knowing which one the institution is actually producing, and this, more than uncertainty considered in the abstract, is what explains the specific character of the exhaustion involved.
That exhaustion should be understood as proportionate rather than excessive once the underlying stake asymmetry is made explicit. For the institution, the case may be one file among hundreds, reviewed intermittently by people with many other responsibilities. For the target, it may organise livelihood, professional identity, income, friendships, daily routine, sleep, family life, and future employability all at once. The two parties cannot be expected to devote comparable attention to the same file, and what can look, from the institutional side, like disproportionate fixation is very often exactly proportionate to how the stakes are actually distributed between the two sides. This has embodied consequences that should be named plainly, without reducing the phenomenon to individual psychology: vigilance that does not switch off, exhaustion that rest does not resolve, disrupted sleep, bodily tension, changes in appetite, difficulty disengaging attention even during unrelated activity, and a specific dread attached to notifications, particular buildings, or particular meetings on a calendar. The body bears the metabolic cost of an interrecursive situation that has not been permitted to close.
It follows that the common advice simply to stop thinking about it tends to fail for a specific, structural reason rather than because the advice is badly delivered. That advice presupposes the ongoing modelling is a habit that could, in principle, be dropped without cost. But a person’s continued livelihood, professional standing, and sometimes physical safety may actually depend on correctly anticipating what happens next, and the underlying problem is not excessive cognition in need of correction. It is an institutionally sustained absence of the conditions under which settlement would even be possible.
That absence should also not be confused with formal quiet. A case can be closed, in the sense that correspondence stops and a file is marked complete, without the target’s recursive workload dropping at all — because nothing about the underlying model settlement or coordinative settlement has actually occurred. A conflict has been successfully muted only when the conditions requiring continued modelling have genuinely been removed, not merely when institutional communication has ceased. Where a process ends without either kind of settlement, the target may reach a further, distinct requirement that a merely procedural ending cannot supply: recognitional closure, in which what happened is named in a form the person actually recognises, the evidentiary pattern is acknowledged rather than dissolved into isolated incidents, responsibility is allocated with adequate precision, and the institution’s account no longer forces the target to inhabit a description of events they cannot recognise as their own. This does not mean every complainant must receive the interpretation they personally favour. It means a fair process must explain why its account differs from the target’s, and must show that the original articulation was genuinely understood before it was set aside.
XV. Antagonism as a Hostile Mesocosm
Drawing the preceding sections together: a person in this situation is no longer, in any straightforward sense, involved in a conflict with one other person. They inhabit a hostile mesocosm — the whole mediated field of body, place, relation, material, and symbol within which an ordinary life is sustained — whose usual arrangements have become reorganised around the unresolved conflict. Embodiment shows up as vigilance, exhaustion, and disrupted bodily rhythms. Relationships with colleagues and institutional actors become difficult to read with any confidence: a friendly greeting, an unusual silence, a change in someone’s manner can all now be read, rightly or wrongly, as data about where that person stands. The wider temporal and spatial orientation of ordinary life narrows and destabilises, as day-night rhythm, travel, and a reliable sense of where one belongs in time and place become harder to sustain. Files, devices, and communications preserve the conflict and can reactivate it at any moment; a message received at home in the evening can undo an entire evening’s rest. And allegations, labels, and institutional categories determine, before anyone has had the chance to explain themselves, what a given action is taken to mean.
Some of the ground on which this hostile mesocosm is built long predates the conflict itself. The office, the reporting line, the complaints procedure, the welfare category, the email system: none of these was created by the antagonist, and none of them was created for this purpose. They are, from the target’s point of view, simply there — inherited institutional arrangements that feel as fixed and unchangeable as terrain, even though every one of them was, at some point, deliberately built. This matters because the antagonistic actor does not need to manufacture an entire hostile world from scratch. They can recruit ground that already exists, and whose apparent neutrality conceals exactly how selectively it is now being activated. The office, the policy, the reporting line, and the welfare category all preexist the conflict; their recruitment is what gives one person’s private hostility a reach far beyond anything they could have built alone.
This hostile mesocosm is not simply located in the person’s head, and describing it that way would badly misdescribe what is actually happening. It is also not a straightforwardly objective fact existing independently of how it is experienced. It is a real, mediated environment, produced by the interaction of roles, artefacts, symbols, bodies, and recursive centres, no less real for being difficult to point to as a single object. This gives the analysis a substantive alternative to purely individual psychological explanation. What might appear, from the outside, as rumination or hypervigilance can instead be the embodied continuation of a structurally unclosed situation — a rational response to a real and ongoing informational asymmetry, not a distortion imposed on top of an otherwise resolved one.
Part V: Scope and Consequences
XVI. Why Universities Are Especially Vulnerable
Universities combine decentralised academic authority with weakly defined managerial discretion, collegial and hierarchical structures operating side by side, overlapping academic and administrative jurisdictions, strong reputational dependence, considerable informal influence alongside formal complaints machinery, and, characteristically, multiple layers of human resources provision at department, school, college, and university level. This combination creates unusually abundant opportunities for the kind of role ambiguity discussed in Part II.
Universities are also, properly, highly attentive to welfare, complaints, and safeguarding, and this appropriate attentiveness creates a structural affordance that can be misused. A manager pursuing a personal antagonism against a colleague can move the underlying issue from the comparatively low-consequence category of ordinary staff conflict into the considerably higher-consequence categories of formal complaint, professional misconduct, safeguarding concern, or inappropriate conduct. Once an issue has been reframed in these terms, the ordinary restraint that might otherwise apply to an unproven allegation can disappear almost entirely, because the institution’s own risk calculus now points toward precaution rather than patience. This tendency toward precautionary escalation is reinforced by realistic institutional fears: legal liability, reputational damage, regulatory scrutiny, unfavourable coverage, and effects on recruitment and standing. None of these fears is irrational on the institution’s own terms. Together, they create strong incentives to treat any allegation touching a high-relevance category as something to be managed defensively rather than investigated openly.
It might seem that the multiple layers of review universities characteristically provide should offer real protection against exactly this kind of capture. In practice, this protection is often weaker than it looks, because the apparent redundancy is frequently only apparent. Human resources relies on the manager’s account. Senior management relies on human resources. A committee relies on the file assembled for it. Formal review relies on the record produced by prior review. Multiple institutional nodes seem to provide independent scrutiny, but if they all ultimately inherit the same originating interpretation, they will tend to reproduce, rather than catch, one another’s failure — a kind of pseudo-redundancy in which the appearance of multiple checks conceals a single point of dependency running underneath all of them. At the same time, the protective redundancy that would ordinarily help a target withstand this — trusted colleagues, union support, a stable professional identity, a home life untouched by work — tends to collapse under sustained pressure of exactly the kind described in Part IV: colleagues grow cautious, home becomes reachable by work correspondence, professional identity narrows around the case, and alternative paths forward come to seem threatened rather than available. The institution’s redundancy weakens at precisely the moment the target’s own redundancy is most needed and least available.
Academic institutions also rely heavily on informal reputational shortcuts in day-to-day functioning. Judgements such as he is a nice guy, she is difficult, he is reliable, or she often complains function as pre-existing models that shape how new evidence is received, often before any formal process has begun. And because different actions are frequently reviewed separately — one meeting, one complaint, one email, one performance concern, each considered largely in isolation — a real pattern of conduct can remain invisible at every individual point of review even when it would be unmistakable if the same events were considered together. The antagonism becomes visible only at the level of the pattern, while the institution’s own way of processing complaints repeatedly decomposes that pattern back into a series of isolated, individually defensible acts.
XVII. The Human Distinctiveness of Institutionally Extended Antagonism
It is important to be precise about what is and is not being claimed here, because the weaker version of this claim is not interesting and the stronger version needs to be stated carefully enough to be tested. The claim is not that prolonged antagonistic stress is uniquely human. Many animals experience chronic predation risk, unstable dominance relations, social exclusion, defeat in repeated contests, territorial pressure, competition for mates or food, and, in captivity, human-inflicted distress that can last a lifetime. None of this is being denied or minimised.
What appears distinctively human is a specific capacity that underlies everything described in this article: the capacity to sediment recursive coordination into durable material and symbolic deposits that can be taken up by new recursive centres across absence, distance, and generations. Institutionally extended antagonism is one dark consequence of this capacity, not a separate phenomenon requiring its own explanation. It is what makes possible antagonistic situations that are simultaneously distributed across multiple centres who need not share a single hostile intention; preserved in artefacts that persist independently of any person’s continuing attention; organised through categories that reallocate institutional relevance; enacted through roles that separate authority from the person exercising it; extended well beyond any face-to-face encounter; recursively opaque to the person most affected by them; capable of persisting for years without reaching closure; and partially autonomous from the continuing intentions of any single antagonist.
The claim deserves to be stated in a way that could, in principle, be shown wrong, and it is worth doing so directly rather than relying on the vaguer phrase no other animal. It would be weakened, or falsified outright, by evidence that some other species constructs antagonistic situations involving durable external records of allegations or status; institutionally differentiated roles occupied by different individuals over time; delegated authority exercised on behalf of a collective rather than an individual; distributed enforcement carried out by individuals who were not present at the original conflict; symbolic reclassification of a target through something functioning like a category system; prolonged recursive uncertainty about deliberation the target cannot observe; and material or social continuation of the antagonism across a timescale of years. No currently documented nonhuman species appears to produce this specific configuration.
XVIII. Implications for Institutional Design
The theoretical account developed above carries practical implications, worth stating briefly and in conceptual rather than prescriptive form, without turning this article into a policy document. Five principles follow fairly directly from the mechanism described.
Independent recursive access. Reviewers must have channels of information that do not depend on the actor whose conduct is under review. A process that can only see a case through the eyes of the person accused of misusing their authority has not actually achieved independent review, however many formal stages it passes through.
Pattern-preserving evidence. The process must be capable of examining sequences, comparison cases, timing, and selective application, not only discrete authorised acts considered one at a time. Given how pattern destruction operates, a process that insists on atomising every incident before assessing it has built its own blind spot into its method.
Interlegibility obligations. The institution must disclose enough of its own reasoning structure for the affected person to understand who considered what, and how interpretations were reached, even where full transparency is neither possible nor desirable. The aim is not to eliminate every asymmetry but to keep it within bounds a person can work with.
Category accountability. Movement into a high-relevance pathway, such as a safeguarding or misconduct process, must itself be reviewable. Who classified the case, from what evidence, and against what threshold, are questions a fair process needs to be able to answer, not because such categories are illegitimate but because their consequences are serious enough to warrant their own scrutiny.
Recursive closure as the actual goal. The purpose of a process is not merely to issue a final decision but to achieve sufficient epistemic and practical settlement that unnecessary continuing modelling by the target is no longer required. A letter closing a file is not the same achievement, and treating it as equivalent is where many otherwise well-intentioned processes fail.
It should be said plainly that maximal transparency is not, on its own, the remedy this diagnosis implies, even though the problem has been described throughout in terms of asymmetrical interlegibility. Institutions legitimately require privacy, confidential testimony, protected deliberation, legal privilege, safeguarding boundaries, and room for preliminary discussion that has not yet hardened into a position. The aim is not complete interlegibility but sufficiently reciprocal and contestable interlegibility: a process must disclose enough to make correction and closure possible without making every participant maximally visible to every other. Confusing these two things would trade one form of institutional dysfunction for another.
XIX. Conclusion: The Longest Human Conflicts Are Built
Human beings can certainly attack, threaten, dominate, and exclude one another in ways that are continuous with the behaviour of other social animals. Institutionalised human life produces something further, and the whole of this article has been an attempt to state precisely what that further thing consists of, and how it actually works, stage by stage. The conflict begins as an interrecursive relation between two people. An officeholder converts a personal antagonism into authorised action. Material and symbolic deposits preserve that action while obscuring the recursion that produced it. Later actors, understanding themselves to be doing nothing more than their jobs, take up those deposits through ordinary institutional trust. Categories redistribute relevance and compress the target into a portable label. The target, meanwhile, faces a network that remains recursively occluded to them at every level that matters. Evidence and appeals generate more hypotheses rather than fewer. Closure becomes structurally unavailable, not through anyone’s deliberate design, but through the accumulated, individually reasonable behaviour of an entire institution. And the target’s whole world is reorganised around anticipating an institution that can model them more cheaply, and more consequentially, than they can ever model it.
The distinctively human achievement here is not merely that hostility can persist after the antagonist has left the room. It is that recursive hostility can be deposited into fixed anchors, taken up by people who never shared it, concealed within entirely legitimate coordination, and returned to the target as an entire world they must continuously try to understand but can never adequately see.
The most devastating question in situations of this kind is, therefore, often not who is attacking me. The person at the centre of it may know the answer to that question perfectly well. The unanswerable question is where the person ends and the institution begins. Antagonistic interrecursivity under asymmetrical interlegibility is, in this sense, one of the darker achievements of human institutional life: human beings have become capable of constructing hostile worlds that no single individual fully controls, that no outsider can easily perceive from beyond their boundaries, and that no target can simply leave behind by walking away from the room where the conflict appeared to begin.