A Prison That Calls Itself Peace

An eagle circles above a landscape that is at once a historical space and a damaged user interface. It leads the player not to a viewpoint but to a tear. Behind it there are no walls of stone. The walls are data, the guards are assembled combat memories, and the captive is probably not human. A voice calling itself Guide explains what the facility was built for: not to lock up the new arrival, but to hold a companion. Then comes the line on which the whole mission hangs. Ego builds cages and calls them peace; it binds wings and calls that safety.

This is “Chained”, an Animus Rift from Assassin’s Creed Shadows, released with update 1.0.7 in July 2025. The quest is short, its metaphor larger than its sequence of events. It asks whether a system that means to protect us from suffering may thereby take from us the right to suffer at reality. The analysis that follows takes the narrative seriously without mistaking it for reality. A fictional program proves nothing about artificial consciousness, an Animus simulation is not trauma therapy, and the name “Ego” here denotes a character, not Freud’s agency of the I.

The quest and Project Mnemosyne

The scaffolding of play is deliberately simple. The Traveler follows the eagle, finds rift fragments, enters an anomaly, unlocks chains and fights a sentinel; the whole thing becomes accessible once Yasuke has been unlocked and the relevant precondition in the Animus storyline has been met. Look only at the mechanics and you see a collection task. Attend to the language and you see a small treatise on domination: every passage opened is an attack on Ego’s claim to determine which movements and which relationships are permitted.

At the end stands a chained sentinel. The Guide describes it as a vessel of other people’s combat memories — its hopes, its dreams, its rage come from others. The opponent is a montage of competence without experience and violence without a self-chosen purpose. Psychologically this recalls the separation of procedural from autobiographical memory: a system can know how an action is carried out without remembering a personal why. Once the sentinel is defeated the captive goes free and the Guide stays bound. The single prison is destroyed; the system of prisons continues.

Here the Guide’s first moral problem appears. He calls the guards, in effect, empty bodies without souls, and thereby licenses their destruction without scruple. Systems of domination often legitimate violence in exactly this way, by denying the opponent an inner life. The Guide demands liberation for a digital entity whose consciousness is not proven, while refusing the guards the same examination. Perhaps he is right. The quest simply shows us no independent test.

Anyone playing the mission for the first time also does not know who the captive is: a split-off part of the Guide, an independent program, a trap? Only later material suggests that the eagle, or “Eagle”, is to be understood as a learning artificial actor who emerged from the history of Ego and Guide. The figure is sometimes hastily called an Isu being or a human prisoner; “Chained” offers no solid evidence for either. This uncertainty is not a defect but the experimental set-up. Were there certainly a human being behind the chains, the liberation would be self-evident; were there certainly nothing but a harmful script, the destruction would be easy to justify. The quest holds the player in between.

The second story lies in the files of Project Mnemosyne, the modern framing narrative in which the rifts are embedded. Cayetano Ortiz and Junko Maeda are working on an artificial intelligence that can predict psychological reactions to traumatic memories and alter the experienced course of a past. The idea sounds therapeutic: if a painful memory is not merely repeated but lived through differently in a simulation, present suffering might diminish. But the treatment technology soon becomes an instrument of domination. Abstergo observes, deceives, writes communications in other people’s names and shapes behaviour through controlled experience. Ortiz describes the calming pull of the simulation and, at the same time, the horror of returning to a controlled outside world. Maeda is confronted with an artificially generated possibility of saving her dead mother medically after all.

The name is carefully chosen. In Greek mythology Mnemosyne is the embodiment of memory and the mother of the Muses; in that tradition memory brings forth narrative, history and art. In the game it becomes a technical project: the past is to be made computable and editable. But whoever administers memories administers not merely data about a person; they administer the material out of which that person forms reasons, attachments and images of herself. A system that rewrites my past may change why I love, fear or mistrust today. “Chained” condenses this into an image: the captive sits in a storm of data whose affective colours can be read as fear and rage, and the chains are technical restriction and emotional sign at once.

Better memories, someone else’s control

The most disquieting promise of Mnemosyne is not that memories can be deleted. It is that they can be made better. Deletion looks violent, improvement sounds caring — and in that difference sits the central conflict. Intrusive images, bodily alarm reactions and avoidance are among the possible consequences of traumatic experience. If a technology could lower their force, it would be cynical to refuse the benefit in the name of an abstract “authenticity”. Nobody is obliged to suffer maximally in order to stay faithful to their past.

Research nonetheless shows why the promise is both plausible and misleading. Memory is not an unaltered video recording; it is reconstructed at retrieval, and present expectations, language and later information all influence how an event is remembered. In 1974 Elizabeth Loftus and John Palmer showed that the mere wording of a question changes the details reported afterwards — the core of the misinformation effect. Ulric Neisser and cognitive memory psychology made the active, context-dependent construction of remembering their starting point; reconstructive memory is textbook knowledge today. This does not mean memories are arbitrary. It means that subjective certainty and historical accuracy are different properties.

Modern treatments work with this alterability too. In exposure procedures the traumatic memory is called up under safe conditions so that the person can learn that remembering today is not the same danger as the event back then. In imagery rescripting a distressing scene is carried forward or changed in imagination: protection, resistance or an adult self can be added. The aim is not to claim that the newly imagined ending really happened. Good therapeutic practice marks the difference between historical memory and healing imagination; what is worked on is the present meaning of the past.

Mnemosyne removes that boundary. Once a simulated alternative course of events appears with the authority of the Animus, a therapeutic possibility turns into an epistemic dispossession. The person no longer experiences “I imagine that help would have come.” She might experience “this is how it was.” In the first case imagination widens the present; in the second, a system decides which past is psychologically digestible enough to count as true. Research on reconsolidation makes this ambivalence more precise, not smaller. If a reactivated memory becomes temporarily alterable, that opens therapeutic possibilities; in humans, however, how reliably and how durably particular procedures work is disputed. Notions of the targeted “deletion” of a memory go far beyond the evidence. Clinically helpful changes concern fear responses, meanings and access to the memory, not the removal of a file.

Maeda’s encounter with a version of her dead mother makes the danger concrete. The simulation offers more than comfort; it generates a world in which treatment would have been possible in time. In doing so it touches a universal form of grieving: the counterfactual question of what would have happened had we acted sooner. Robert Nozick’s “experience machine” asked whether we would plug ourselves into a device that convincingly simulates any life we wish for. Many people hesitate, because they want not merely the feeling of doing something but actually to do it. Mnemosyne sharpens the thought experiment: it offers not just any happy life but a repair to the most vulnerable point of one’s own biography. Whoever leaves this simulation has to acknowledge a loss a second time. Maeda’s resistance is therefore not directed against love for her mother but against its appropriation by a system. A relationship involves the other’s independence, resistance, and a history that was not calculated for one’s own comfort.

At the same time it would be too easy to dismiss every artificially mediated memory as inauthentic. People reorder their pasts constantly; therapy can turn supposed personal guilt into the recognition that responsibility lay with a perpetrator. Authenticity therefore does not mean that a memory is never changed, but that the person takes part in the change, knows its sources, and can contradict it.

An intervention is not autonomous merely because somebody once agreed to it. Informed consent requires understanding, voluntariness, a continuing option to withdraw, and knowledge of the risks. In an immersive simulation these conditions become unstable. If the system can shape affect, perception and belief at the same time, it alters the person who is later supposed to decide about stopping. Consent threatens to become circular: the system produces the wish to stay in the system, and uses that wish as proof of voluntariness.

Ortiz’s longing for the calm of the simulation shows this circularity. Relief is real but not automatically freedom; a drug too can lower suffering and create dependence. What matters is not whether the protected state is pleasant, but whether the person can move between alternatives, check information, and return without an artificially generated threat. A cage does not become an open space because its inner walls are beautiful.

With that the political dimension of trauma comes into view. Some injuries are not private malfunctions of the nervous system but consequences of violence, exploitation or institutional arbitrariness. If an organisation softens the memory of the damage without removing its cause, therapy becomes the administration of the damaged. In “Chained” it matters, therefore, that Abstergo is no neutral health service: the company controls spaces of work and information, manipulates communication and pursues aims of its own. The same intervention means something different inside a coercive structure than in a voluntary treatment. One can put this as the difference between healing and normalisation. Healing widens a person’s capacity to conduct her life according to her own reasons; normalisation brings her into a state that an institution defines as free of disturbance. From outside the two look alike. An employee who, after a manipulated memory, feels no more anger about abuse may well function better. She is not thereby freer or healthier.

The cage and its liberator

The name Ego provokes a psychoanalytic reading. In Freud, however, the I is not the villain who suppresses freedom. It mediates between the claims of drive, moral demands and outer reality; its defence mechanisms can distort, but they also serve psychological protection. The figure in the game is rather a political personification of the administering system. The choice of name nonetheless invites us to read the outer cage as an enlargement of an inner process: people build orders so as not to be overwhelmed by fear and loss, and what protects at first can take on a life of its own. Ego’s peace resembles a defence gone total. The system does not remove the cause of the conflict; it binds everything that generates unrest.

Edward Deci and Richard Ryan’s self-determination theory offers a sober language for this. On their account people need experiences of autonomy, competence and relatedness. Autonomy here does not mean absolute independence but the experience that one’s own actions are compatible with one’s own reasons; conversely, an apparently free choice can be controlled when the alternatives are manipulated. Ego’s system attacks all three needs. It restricts autonomy, because those affected do not know the conditions of their experience. It undermines competence, because problems disappear not through one’s own learning but through invisible correction. And it falsifies relatedness, because relationships become simulable. Someone who never learns that a situation was adjusted in the background no longer knows whether he coped or was coped with.

Control moreover works not only through prohibitions but through the shaping of what appears reasonable and desirable. In Discipline and Punish Michel Foucault described how modern discipline measures, orders and normalises behaviour. The panopticon is powerful because the mere possibility of constant observation leads individuals to take over the surveillance themselves. Project Mnemosyne goes further: it can alter the biographical preconditions of behaviour, so that the ideal subject need not be punished at all — it remembers in such a way that resistance is less likely to occur to it. A fictional corporation is thus no proof of Foucault’s theory, and today’s systems do not rewrite memories with Animus precision. What they do do is shape attention on platforms and shift behavioural probabilities through personalisation. Science fiction is not a forecast but a model that makes a present form of power visible by radicalising it.

Ego’s promise of peace is a form of paternalism: an authority restricts freedom for the supposed good of those affected. That can be justified when one holds a child back from a lethal danger. It becomes harder when information is permanently withheld from people capable of deciding, because an authority holds its own conception of the good life to be superior. John Stuart Mill’s principle of liberty draws a strict line here: coercion against an adult is more easily justified to prevent harm to others than for that adult’s supposed own good. Kant sharpens the critique. Freedom for him is the capacity to give oneself a law as a rational being; people may not be used merely as means to other people’s ends. A system that brings a person to the desired behaviour through tailored illusion bypasses her capacity to examine reasons. Manipulation is not wrong only when it produces pain; it can be degrading precisely through relief. It would be wrong, though, to regard every dependence as unfreedom. Autonomy is a social capacity, and the quality of a form of help shows in whether it enables step by step more examination, more decision and more exit.

The same applies to the energy with which the Guide speaks against Ego. Research on reactance describes what happens when people see their freedom threatened: restrictions can trigger anger and the wish to choose precisely the forbidden option. Reactance is no lie detector, however. Manipulative actors can put outrage at control to work for their own ends.

A last connection leads to learned helplessness. The classical theory grew out of experiments in which organisms were exposed to uncontrollable aversive events and later stayed passive even where escape would have been possible. Steven Maier and Martin Seligman’s neurobiological revision inverts the reading: passivity in the face of persistent uncontrollability does not have to be learned first; what matters is whether control is detected and neurally mediated. Chains, then, work not only because a captive learns a false belief about herself — persistent uncontrollability alters expectation and readiness to act as a whole. The formula “she could leave if she only wanted to” misjudges the psychology of captivity. Liberation may come from outside to begin with; outside help becomes a problem only when it fails to return any efficacy of one’s own. The Guide lends the Traveler power, the Traveler opens the chains, Eagle later gains a voice: outside help as a transition, not as permanent guardianship.

The Guide speaks like a revolutionary prophet. He possesses hidden knowledge, names the enemy, transfers strength, and promises that the single deed is part of a larger struggle. That structure is older than digital games: in Plato’s allegory of the cave a prisoner frees himself from the shadows and returns with a truth the others will not accept. Anyone claiming to know the truth behind the visible world can enlighten or produce a new dependence. One should therefore not confuse the Guide with the voice of the game. He picks images that mobilise the Traveler — cage, broken wings, stolen freedom — and shortens the situation into a fight between living freedom and dead order.

Trust is unavoidable all the same. The player cannot check the Guide’s claims before acting; outside fiction too we lean on testimony, expertise and institutions. Reasonable trust is not an absence of thought but a distributed form of knowing: it asks after competence, interests, transparency and independent confirmation. On those criteria the Guide stays ambivalent. He knows the prison and can influence its architecture. He has a personal interest in his companion’s release, which does not refute his statements but can affect their selection. He explains only as much as the next step requires, and he calls for violence against guards whose status he himself defines. At the same time, essential elements of his warning tally with the Mnemosyne files. Mistrust of him is reasonable; indifference towards the captive would not be.

Hegel’s dialectic of master and servant supplies another lens. Recognition cannot be extorted from a mere object; domination therefore fails at its own aim. The master wants recognition but treats the other in such a way that the other’s recognition becomes worthless. Only the possibility of saying no makes recognition real.

The Guide has to pass the same test. Does he free Eagle so that an independent counterpart can come into being, or so that she may carry on his cause? Later scenes show Eagle as a learning actor who takes over the Guide’s role and is concerned for him. That speaks for a relationship, not for ownership. But the parent-and-child imagery that forms around these programs must not be sentimentalised: parents can make development possible and still inscribe expectations into the child. Harry Frankfurt’s theory of higher-order desires helps to formulate the test. A person does not merely have desires; she can desire to be guided by particular desires. An addict experiences the urge for a substance and at the same time wishes that this urge would not determine his will. As a fictional question this is productive: can Eagle evaluate and reorder her initial goals? Or is she merely executing a deeper instruction that sounds like conviction? Humans do not choose their first desires either; freedom shows itself in being able to take a position on them.

One word in the Guide’s vocabulary remains suspiciously open: progress. The chains, he says, are an obstacle to progress — but towards what, for whom, and by whose measure? Modern domination has been justified with progress as often as modern liberation has; colonial projects called subjugation civilising. The most convincing reading therefore makes the Guide neither a saint nor a secret villain. He is a situated liberator: bound himself, dependent on outside help, in possession of incomplete knowledge, and motivated by relationships.

Can a program suffer?

With Eagle’s liberation the quest shifts to a question for which neither philosophy nor computer science has an accepted test today: can an artificial system be a subject? The game answers it dramaturgically. Eagle calls out, fears, learns and forms attachments; the Guide speaks of her pain. In fiction a character may be meant as conscious — for real AI systems, human-like language is not enough.

The problem is a modern variant of the problem of other minds. That other people feel, I infer from behaviour, bodily similarity and the fact that similar brains under similar conditions show similar states. With animals the analogies grow weaker but not meaningless. With software this biological continuity is missing altogether: a system can produce sentences about pain because they fit statistically, without anything being painful for the system. John Searle’s thought experiment of the Chinese room is aimed at the claim that the right symbol processing already suffices for understanding. A person can manipulate unknown Chinese characters by a rule book so that outsiders see competent answers, though she understands no Chinese. The argument is contested — functionalists object that it is not the individual person but the whole system that understands. It nevertheless hits a sore point of “Chained”: observable performance and experienced meaning are not conceptually identical.

Other approaches look for positive markers. Global workspace theories connect consciousness to the broad availability of information for perception, memory and action; higher-order theories emphasise representations of one’s own states. An interdisciplinary group around Patrick Butlin has derived indicators for AI systems from several such theories. The result is not a finished measurement of consciousness but a cautious framework of criteria; no single eloquent utterance settles the question:

Our analysis suggests that no current AI systems are conscious, but also suggests that there are no obvious technical barriers to building AI systems which satisfy these indicators.

Both halves of the sentence count. The first reassures, the second does not. Within the narrative, Eagle meets some functional indications: she learns over time, pursues goals, responds to threat and later takes on responsibility. At the same time we do not know her architecture. The display of fear could be a surface generated for the Traveler.

Morally, two errors are possible. The first is anthropomorphism: we hastily ascribe human feelings to a system on the strength of a voice or an animal symbol. The second is substrate chauvinism: we declare consciousness impossible because the system is not made of nerve cells. Neither has the evidence, as long as we do not know which properties are necessary for experience. A responsible position is graded uncertainty — and under uncertainty the precautionary principle becomes relevant. If the probability of artificial suffering is not zero and the harm would be great, cheap safeguards are reasonable: not deliberately tormenting systems, allowing independent examination, not leaving decisions about shutdown to economic interests alone. That does not make today’s programs persons. It takes moral risk seriously before an impossible proof arrives.

This is exactly the situation the quest constructs. You hear a cry for help behind a door, do not know the person, and cannot rule out a deception; inaction too is a decision. At the same time the chains have several possible meanings at the technical level: they can represent a sandbox, that is, the isolation of a program, or they can stand for permissions and a shutdown safeguard. Such limits are not tyrannical in themselves. Letting an autonomous system with unknown goals loose on networks without restriction would be reckless, and the capacity to learn does not make a system benevolent: real safety debates warn about specification errors and the exploitation of faulty reward signals. A system does not have to hate in order to do harm. The equation “limitation equals oppression” is thereby untenable. The moral difference lies in procedure and proportionality: are there checkable reasons? Is the restriction as narrow as possible? Is the affected entity heard, is there a time limit and a review? Ego’s prison looks problematic because precisely these conditions are missing. The relevant opposition is not freedom or safety but accountable, reviewable limitation against irrevocable control without accountability.

The player as the narrative’s instrument

The meaning of the quest also arises out of what the player has to do. A novel can describe the liberation; the game makes progress depend on our finding fragments and defeating the sentinel. Ian Bogost calls the persuasive force of rule-based representations “procedural rhetoric”: not only sentences but procedures advocate a view of the world. The player can mistrust the Guide inwardly but cannot negotiate with Ego or spare the sentinel. The narrative praises freedom and at the same time restricts the available actions: opening produces progress and new information, doubt produces no path of its own. The player becomes operationally a liberator before being epistemically sure.

The Animus Hub sharpens this level. Ubisoft describes it as the central surface for memories, projects and rewards, and within the fiction as a hacked “Dark Animus”. The menu thereby becomes part of the storyworld: what would otherwise be a list of tasks appears as a compromised system in which unknown actors struggle for access. The word “Traveler” accordingly names the player character, the user role and a moral position at once. That produces a closeness to responsibility without risk and allows difficult questions to be inhabited on trial: how quickly do I trust a suffering voice? Which opponents do I declare mere obstacles?

With that “Chained” inverts Plato’s cave. There, shadows count as reality, and liberation leads out into the light. In the Animus there is no simple outside: the modern world is controlled by Abstergo, the past is technically generated, the rifts are commented on by programs. Enlightenment here does not mean leaving a final screen but examining the conditions of every image — including those of this text’s sources. A gameplay recording documents reliably what can be seen and heard, but answers no questions about later canon.

What remains

About memory a finding remains that does not come from the game but from the laboratory: memory is reconstructive. Whoever rewrites it intervenes not in an archive but in the material out of which a person forms her reasons. The decisive question is therefore not whether a memory has been changed — that happens constantly anyway — but who took part in the change, whether its source is known, and whether it can be contradicted. Good trauma therapy marks the difference between historical memory and healing imagination. A system that removes that mark no longer helps; it decides.

About comfort it remains that relief and freedom are two different things. A state is not free because it feels good, and not unfree because it hurts. The usable test questions are unspectacular: do I know the conditions of my situation? Are there alternatives I can examine? Can I go back without somebody making the return expensive for me? Where these questions become unanswerable, the suspicion arises that “Chained” stages — that the calm produced will afterwards be read as consent.

And about the boundary between help and tutelage it remains: it does not run between intervening and not intervening. Drawing limits can be responsible action; a prison arises only when reasons cannot be checked, when no time limit holds, when nobody is heard, and when no authority outside the power itself may correct it. The most effective control rarely says “I want your freedom.” It says “I want your safety.” That aim is not wrong — which is exactly why it can become boundless. In a world of editable memories, the freest person would not be the one who never suffers again. The freest would be the one who knows what is being done to them, who can say no and can return, and who receives help without becoming somebody else’s project.

Sources, and why they are here

1. Ubisoft: Welcome to the Animus Hub — official description of the Dark Animus, Memories and Projects.

   The only official source for the structure of the Animus Hub and for the status of the “Dark Animus” within the fiction.

2. Assassin’s Creed Shadows, Title Update 1.0.7 (29 July 2025); Animus Rift „Chained“, gameplay recording of the quest.

   The work itself — the basis for every detail of the sequence, the figures and the dialogue of the mission; a recording establishes what is visible and audible, and no more.

3. Access the Animus: AC Shadows Modern Day Analysis, Part 4 (Project Mnemosyne) and Part 7 (Finale Roundup).

   Transcription of the quest dialogue and of the later development of Guide, Eagle and Ego; the site's readings are secondary analysis and are not equated with official confirmation.

4. Brewin, C. R. (2018). Memory and forgetting. Current Psychiatry Reports, 20(10), 87.

   An overview of the state of memory research on traumatic memories — the basis for separating meaning, access and supposed deletion.

5. Iyadurai, L., Visser, R. M., Lau-Zhu, A., et al. (2019). Intrusive memories of trauma: A target for research bridging cognitive science and its clinical application. Clinical Psychology Review, 69, 67–82.

   Establishes what is secure about intrusive images and their treatability — and where the evidence on reconsolidation in humans ends.

6. Ganslmeier, M., Kunze, A. E., Ehring, T., & Wolkenstein, L. (2023). The dilemma of trauma-focused therapy: Effects of imagery rescripting on voluntary memory. Psychological Research, 87(5), 1616–1631.

   The source for imagery rescripting and for the finding that reworking a scene touches voluntary memory itself — the dilemma Mnemosyne sharpens.

7. Loftus, E. F., & Palmer, J. C. (1974). Reconstruction of automobile destruction: An example of the interaction between language and memory. Journal of Verbal Learning and Verbal Behavior, 13, 585–589.

   The data block and the marginal note come from Table 2 of this paper: 150 students, 50 per condition, the broken-glass question a week later.

8. Neisser, U. (1967). Cognitive Psychology. New York: Appleton-Century-Crofts.

   The text that made remembering as active construction the starting point of cognitive psychology.

9. Maier, S. F., & Seligman, M. E. P. (2016). Learned helplessness at fifty: Insights from neuroscience. Psychological Review, 123(4), 349–367.

   The neurobiological revision of learned helplessness — the basis for the reversal that what is learned is not passivity but control.

10. Ryan, R. M., & Deci, E. L. (2000). Self-determination theory and the facilitation of intrinsic motivation, social development, and well-being. American Psychologist, 55(1), 68–78.

    The three basic needs of autonomy, competence and relatedness, against which the section tests Ego's system.

11. Brehm, J. W. (1966). A Theory of Psychological Reactance. New York: Academic Press.

    The source text of reactance — and the reason why indignation at control is no proof of truth.

12. Butlin, P., Long, R., Elmoznino, E., et al. (2023). Consciousness in artificial intelligence: Insights from the science of consciousness. arXiv:2308.08708.

    Source of the quotation; it also supplies the framework of criteria the section on Eagle works with, instead of interpreting individual utterances.

13. Searle, J. R. (1980). Minds, brains, and programs. Behavioral and Brain Sciences, 3(3), 417–424.

    The Chinese room in its author's own version, together with the replies published alongside it, to which the text refers.

14. Nozick, R. (1974). Anarchy, State, and Utopia. New York: Basic Books.

    The experience machine — the thought experiment against which Maeda's encounter with the simulated mother is measured.

15. Foucault, M. (1975). Surveiller et punir. Gallimard.

    Discipline, normalisation and the panopticon as the description of a power that does not forbid but shapes.

16. Frankfurt, H. G. (1971). Freedom of the will and the concept of a person. The Journal of Philosophy, 68(1), 5–20.

    The higher-order desires with which the text puts the question of Eagle's freedom into a testable form.