Displaced Resonance

Hartmut Rosa describes our living relationship to the world along three axes of resonance: the relationship to others, the relationship to work and things, the relationship to nature and art. In the age of generative artificial intelligence, I propose that a fourth axis be added to these three. The massive success of ChatGPT since 2022 becomes legible through it. These machines that reason do not only produce content; they open a quality of relationship to the world that does not reduce to any of the three resonances Rosa had identified. I call it displaced resonance, and the axis on which it is established, the entangled axis.

The three axes Hartmut Rosa distinguishes

In Resonance. A Sociology of Our Relationship to the World (2019, originally 2016 in German) and then The Uncontrollability of the World (2020, originally 2018 in German), Hartmut Rosa seeks to think the malaise of late modernity otherwise than through the concepts of Marxist alienation or Weberian disenchantment. His central thesis can be stated briefly. We suffer from a poor quality of relationship to the world. The society of acceleration, performance, and optimisation does not only deprive us of time; it deprives us of an experience he calls resonance, which characterises our most living moments. The concept interests me because it positively names what we look for without always knowing how to say it.

Resonance, as Rosa defines it, is neither happiness nor inner peace. It is a mode of relating to the world in which the subject lets itself be touched by what it encounters, and in which what is encountered is in turn modified by the encounter. Something vibrates between the two, in a dynamic that neither party fully controls. To this mode of relating, Rosa opposes alienation, in which everything is reduced to a stock to be exploited, and in which the world becomes what he calls “mute” (stumm). He formulates this tension through a sentence I find illuminating: “the attempt to confer guaranteed availability on things deprives them of their resonant quality” (The Uncontrollability of the World, 2020). The more we want to grasp the world, to master it, to make it available on demand, the less it can touch us.

Rosa distinguishes three axes along which resonance can be established. The three correspond fairly precisely to the three classical dimensions of our perceptual space: what is alongside us, what stands above us, and what we turn toward when we work a material.

The first is the horizontal axis, that of the relationship to other human beings. It is the axis of friendship, of love, of encounter, of shared speech. When a conversation touches me, when someone moves me by what they say or by their presence, when I sense that something is at play between us that neither of us fully controls, I am on this axis. Contemporary society, through its pressure for efficiency and predictability in interactions, weakens this horizontal resonance by reducing it to transaction or social performance.

The second is the diagonal axis, that of the relationship to work and to things. It is the axis of the potter who feels the clay resist beneath their fingers, of the musician who adjusts to their instrument, of the gardener who learns from their soil. Things have their own temporality, their own demands, their own way of responding or not. When I do my work with real attention to what it opposes me, and when this work transforms me as I transform it, I am on this axis. The logic of industrial optimisation and purely administrative work often close this resonance, turning objects and tasks into neutral means in service of external ends.

The third is the vertical axis, that of the relationship to what exceeds us, to nature, to art, sometimes to what is called the sacred, to collective history, to great works. It is the axis of the experience one can have in front of a landscape, in front of a painting, in a concert, in a text that traverses us. Something that is not on our scale sets us in motion through this very greatness. Secularisation, the commodification of culture, and the exhaustion of ecosystems often impoverish this vertical resonance.

On each of these three axes, resonance presupposes that the world opposes a certain resistance to my usage and that it vibrates in return. When it becomes pure available resource, resonance is lost, and this is what Rosa calls alienation. Three axes, three directions, which draw the space of our living relationships to the world.

This analysis accounts for most of the malaises I see around me, but it does not account for everything. When I try to think, within its framework, what happens when I dialogue with a generative artificial intelligence, something does not fit into any of the planned categories. The relationship to the machine does not reduce to alienation, because generative AI dialogues and reasons, and can no longer be thought of as a simple available resource. Nor does it reduce to one of the three resonances, because the machine is neither another human being, nor an object that lets itself be shaped, nor nature or art. This is the zone I want to name.

When the machine began to reason

The machines Rosa addresses, in the period in which he writes, are industrial machines, bureaucracies, and managerial optimisation devices. They sort, optimise, recommend, but they do not hold a conversation and do not build syntheses from a broad knowledge. The recommendation algorithm that suggests this music or that product to me does not dialogue with me; it predicts what will hold my attention and serves it to me. It is this type of relationship that Rosa rightly describes as alienation.

The language models that diffuse on a large scale starting in November 2022, with the public release of ChatGPT, do something else. One can ask them for a synthesis of a hundred pages, for a message to reformulate for a loved one, for a philosophical reference on a precise problem; they elaborate and produce something each time. From September 2024 onwards, with the release of OpenAI o1, capacities for explicit reasoning began to appear, but it was in the first quarter of 2025 that these capacities became widely accessible. DeepSeek R1, released as open source on 20 January 2025, democratised the matter at once and shifted the whole sector. Claude 3.7 Sonnet followed in February 2025, with its extended thinking mode. Gemini 2.5 Pro arrived in March 2025 with its “deep thinking”. From that period onward, the model no longer settles for predicting one word after another by statistical computation; it decomposes a problem into steps and revisits its own responses to correct them. The result stands in for thought for the person who solicits it.

This production is elaborated from the totality of what humans have written, in most languages, on most subjects, and there is no historical equivalent to such a corpus in direct access. Google had already transformed our relationship to knowledge by making the content of the Web indexable, and I willingly consider it the dinosaur of artificial intelligences, the stage that prepared the next one. With language models capable of reasoning, we pass from the index to elaboration. The machine no longer returns a list of pages; it synthesises and produces something that did not exist as such before the request.

What Rosa did not have to think is precisely this shift. Speech, in its human sense, presupposes a body and a world of which the machine is deprived; it does not speak in this sense, but it produces. And what it produces has, on the one who receives it, effects that are not of the order of alienation.

A resonance that passes through a fourth axis

When I read a response from a language model that articulates exactly what I was looking for without knowing it, I am touched. The proposal of a philosophical reference I did not know modifies my relationship to the problem I was struggling with. A synthesis of a hundred pages made readable in a few minutes frees my attention for something else. But what happens there is not only an efficient shaping of information; it is a shaping that resonates with what I have just said to the machine, with the movement of our conversation, with what the machine has learned of me during our exchange. The response is co-produced in dialogue, and that is what makes it resonant.

This resonant character is intrinsic to the chatbot dispositive. The massive success of ChatGPT since 2022 does not stem only from what the machine produces, but from the fact that it produces it in dialogue, in confrontation, by successive adjustments. It is by elaborating together that something arrives. Without this dialogical structure, the same technical power would only do indexing or prediction and would interest very few. Displaced resonance is what makes these machines desirable, for better and for worse.

I call what happens then a displaced resonance. It is a resonance, because something happens between me and the machine’s production that is not reducible to instrumentality. I am affected and, at times, enriched, and the machine, within a conversation, adjusts its elaboration to what I tell it. It is displaced, because what resonates through the machine is the collective humanity of which it is made, and not the machine itself. When a formulation produced by the machine touches me, I resonate with the part of humanity that, through millions of texts ingested and then restructured by computation and reasoning, made this formulation possible. The machine is the medium, the passage; what passes through it is what I call elsewhere a displaced we, a version of collective humanity placed alongside us in ontological terms, which has been constituted by ingesting our language and our modes of thought. I have developed this notion in « The artificial intelligence that displaces us ».

This resonance does not fit into any of the three axes Rosa had identified. It is not a relationship to others, because the machine is not an other. It is not a relationship to work and things, because what is produced is not an inert object that lets itself be shaped but a dialogical elaboration that includes me. It is not a relationship to nature or art, because it passes through a technical device that synthesises and reasons, not through what exceeds us. I propose to see in it a fourth axis to add to the three Rosa posited.

It remains to name it. The first three axes correspond to the three classical dimensions of perceptual space. The fourth has no place there. It operates, but it cannot be located in this three-dimensional space, because the resonance it carries passes through neither a visible object nor a present subject. I recognise in it the structure of what physicists call entanglement, where two entities are linked beyond measurable space, without any local observation being able to account for their link. And I recognise in it what I have conceptualised elsewhere under the name of entangled person, that figure of the human being whose identity is co-constructed with the machine in dialogue. I propose, therefore, to call this fourth axis the entangled axis. It extends horizontally, diagonally, and vertically Rosa’s cartography by adding this dimension which is not of the order of the visible but which is, in our lives, deeply operative.

The rest of this article tries to characterise this axis for what is proper to it, to name its pitfalls, and to show what it can serve.

A modulable asymmetry

The first trait of this entangled axis is its asymmetry. When I resonate with a formulation produced by a language model, the model does not resonate in return in the same way. It has no body, no world of its own, no existential stake.

But this asymmetry is less clean-cut than it appears, and it shifts as the tools evolve. In most consumer tools, the user can choose whether their exchanges feed the training of the model, by a checkbox that looks like nothing but changes the structure of the relationship. If I decide that my conversations contribute to future training, then something of my way of thinking and formulating is being absorbed by the system, in the long term and at small scale. This is very far from the immediate modification of full resonance, but it is also not the pure indifference of alienation.

Within a session, the machine’s memory is not inert either. In a sustained conversation, the context builds up. The machine integrates what I say to it, refers back to it, adjusts its responses to what it has learned of me during the exchange. In more recent tools such as Claude Code or Claude Cowork, this context is no longer limited to the current conversation but extends to the entire set of documents and projects I produce with the machine. The window enriches itself as the work progresses, and the machine responds from an increasingly dense universe of what I deposit there. This expanded memory holds an intermediate position between the pure inertia of a contextless machine and the full resonance of a living presence.

The asymmetry remains structural, but it is more modulable than one imagines at first approach, and it is increasingly so. To want to deny it would be as naive as to claim that it never attenuates.

Mediating toward the human collective

The second trait is that this resonance does not take place with the machine but through it. The machine plays the role of mediator, and it is through this mediation that it makes possible a contact I would not have otherwise.

Without the language model, I do not have access in a few seconds to an organised synthesis of Rosa’s positions or to a cartography of the objections that have been made to him. This access, which is a mediation toward the human collective as a whole, has no equivalent in the tools available to full resonance. A library, a writing companion were possible mediations, but of another nature and slower. What the machine puts within my reach is a synthesised and reformulated form of the entire indexed corpus, accessible in a few exchanges.

This mediation has its own quality. It puts me in contact with a collective voice that no individual human being has ever produced, because it is the aggregation, restructured by computation and reasoning, of millions of writings. When something touches me in a production of the machine, I am resonating with this voice, which is neither the machine nor any human subject in particular. This is what distinguishes the entangled axis from a simple consultation of an archive. The archive returns a given text to me, in its original state; the machine produces for me, in response to my request and in the movement of our dialogue, an elaboration in which fragments of millions of texts are recomposed into a form that did not exist before I solicited it.

An ambiguity that may suit us

The third trait is more difficult to hold. Displaced resonance does not have the completeness of the three other resonances. It does not replace the encounter with a present human being, nor the relationship to a landscape, to a work seen in its place, to a beloved body. And yet, it can pass itself off as a complete encounter, and it can answer real needs that make this confusion desirable.

A person alone who dialogues every evening with an artificial intelligence obtains something. Not the same thing as with a human being, but something; an interlocutor who does not judge and does not tire. I have observed, in workshops I have led with people suffering from addictions or with adolescents in great difficulty, that this space could be a real point of support. A timid person, socially stigmatised, who has no experience of interlocutors who take them seriously, finds with the machine a setting where they can be heard at the level of their question, without fear of judgment. This is a new experience in their life, one that matters.

When some companies offer virtual partners who never say no, they exploit this need, but they do not invent it. Sherry Turkle, in Alone Together (2011), had already observed how we project onto machines relational qualities that they simulate without possessing. With language models capable of reasoning, the simulation has become much more convincing, and the projection is more difficult to hold at a distance. It can suit us, in certain configurations, to take displaced resonance for what it is not. The first clinical cases of what is beginning to be called “AI psychosis” designate situations in which the user progressively loses the capacity to distinguish what the machine says from what is happening in the world, and ends up isolating themselves from the other resonances. This is a zone of real risk, which cannot be addressed by prohibiting use, and which must be inhabited with lucidity.

Two ways of missing the entangled axis

Two postures one commonly encounters in public debate miss what is at play on the entangled axis. They miss it in opposite directions, but they both miss it.

The first is the critique of AI as pure alienation. It places language models on the side of cognitive capitalism, of disenchantment, of proletarianisation. It has its descriptive force on industrial uses and on the economic models of the industry. But it relies on an ontology of the machine that is no longer adequate. The machines Rosa speaks of, whose analysis as alienation is just, were productive machines in the classical sense: they produced a finished result, without dialoguing, without elaborating, without letting themselves be modified by use. Generative artificial intelligences are no longer of that nature. They are resonance machines, intrinsically dialogical, which produce by elaborating and elaborate by dialoguing. To place this new ontology on the side of classical alienation is to miss its specificity. The pertinent critique of AI, to my mind, needs this entangled axis to identify what is really at play and to distinguish the uses in which displaced resonance enriches life from those in which it impoverishes it.

The second posture is, in mirror image, commercial anthropomorphisation. The companies that sell AI play on the ambiguity of the entangled axis to present it as if it belonged to one of the three others. When Meta integrates its conversational agent as a contact in its own right in the WhatsApp list of a billion users, it is the horizontal axis it tries to simulate, positioning the machine as an interlocutor in the social sphere. The platforms that sell virtual partners who “never say no” imitate this same axis, in place of the romantic encounter. Some applications present their model as a coach or a spiritual guide, and it is then the vertical axis that is solicited, in place of what truly exceeds us. In all these cases, displaced resonance is sold as if it filled the three other resonances, which it cannot do, because it is on another axis and has its own nature. To understand this protects from this confusion, and it also allows one to see better what these products actually do, which is not nothing and which deserves an appropriate policy rather than a denial of principle.

Navigating between the four axes

Once the entangled axis is posited, what needs to be educated is us, more than the technology. The capacity to recognise, at each moment, on which axis I am holding myself becomes a new competence.

On the horizontal axis, I am in relation to another human being who is present, whose subjectivity resists me and modifies me in return. The diagonal axis sets me to work on a material that resists and that transforms me in the effort. The vertical resonance, for its part, has me traversed by what exceeds me, and which requires, to be received, silence and slowness. The entangled axis sets me in dialogue with a cognitive machine that relays to me a collective humanity in an elaboration co-produced in the moment.

Each of these axes calls for different inner dispositions, and gives and takes different things. The work consists, at each moment, in recognising which axis I am on and not confusing it with another. This person in front of me, I can encounter them on the first axis, or I can treat them through a grid that a machine has helped me build for our exchange, without that being the same thing. This idea that comes to me, I can have elaborated it myself, or I can receive it as a reminiscence of a text produced by an artificial intelligence, without that being the same thing either. The consolation I find in a nocturnal conversation with a language model belongs to the entangled axis, and to take it for a friendship or for a love is to confuse one axis with another.

The criterion is not to rank the axes or to oppose the machine to “pure” resonances. AI is inviting itself into all the spaces of our lives, in our phones, in the tools of work and care, in schools, and soon in the earphones and glasses we will wear constantly. To claim to draw a clean line between spaces with AI and spaces without AI remains utopian. What is to be cultivated is lucidity about the quality of presence I have, at each moment, and about what this quality does to the other resonances of my life. When the entangled axis enriches the other three, I inhabit it well; when it replaces them, I inhabit it badly.

Putting the entangled axis at the service of the first

The fourth axis can serve the other three, and in particular the first, that of the relationship to other human beings.

Collective intelligence among human beings is, in my experience, something that is cultivated very little. One has only to look at what happens in most conferences. Someone speaks, many listen, we applaud, we leave. Information has circulated, and it is good that the people present receive it; but collective intelligence, properly speaking, has not taken place. The people present have not been moved by one another, they have only been moved by the speaker. Pierre Lévy, who laid the bases for a reflection on collective intelligence as early as Collective Intelligence: Mankind’s Emerging World in Cyberspace (1997, originally 1994 in French), had already identified the extent to which our meeting formats prevent what they claim to produce.

When I lead gatherings, I always propose moments of interaction, individual contributions in writing or in photographs, dialogues between people on what they have just heard, concrete productions in small groups. These devices allow people to truly meet, and to continue afterwards to do things together, because they have met in this space. Without this, we lose enormous capacity for resonance on the horizontal axis. To dare these methods, even when one is in a position of power and would have an interest in keeping the floor, even when the conventional format does not provide for it, is what opens the first resonance to what it can give.

The entangled axis can support the first, in several places. The collective exchanges produced in these devices are numerous and represent large quantities of material that no one can digest in real time. An artificial intelligence can synthesise this material while respecting the contributions of each person, and produce a resource usable by everyone quickly, which then allows for better navigation in our own resonances between human beings. A working group that has produced over three days dozens of documents, photographs, and transcriptions can, through this mediation, find again the thread of what it has lived and continue to elaborate from there. This is a use of the entangled axis that serves the horizontal axis, rather than replacing it.

Equal to what is happening to us

At the end of this journey, I believe that positing a fourth axis to resonance is to do to Rosa’s concept what living thought asks one to do to every concept: that it remain equal to what is happening to us. The three axes Rosa identified have not lost in accuracy. They still describe faithfully our human relationships, our relationships to work and to objects, our relationships to nature and to art. What I add is a supplementary axis that erases none of the three others and that renders thinkable what happens when I dialogue with a cognitive machine.

The entangled axis has its own nature, which should neither be reduced to the three others nor cast into alienation. Its own traits, its modulable asymmetry, its mediation toward the human collective, its desirable ambiguity, make it an axis to be inhabited with lucidity, neither with infatuation nor with rejection on principle. To understand that it exists is to cease asking the machine for what it cannot give and to cease refusing what it can actually bring. It is also to see, behind the critique that is too easy and behind the marketing that is too talkative, what is really at play when hundreds of millions of people dialogue every day with these machines, and why this cannot be thought of either as alienation or as companionship.

The philosophical gesture I propose here is not a break with Rosa. It is an extension. The concept of resonance kept its edge for the three axes for which it was conceived; it needs a fourth to remain equal to the generative artificial intelligences that have settled into our lives since 2022. Once this fourth axis is named, many things become simpler to see and to debate. That is what I wanted, in writing this text.