What does a good DJ do when performing for an audience in a nightclub? What is it that they attend to when choosing the next record to be played in a specific moment of their set? The following considerations start from the assumption that a good DJ is particularly skilled at grasping and manipulating the atmosphere and vibe of a situation through musical means. “Ba no kuuki wo yomu,” as the Japanese say regarding this form of tacit sensitivity to social situations, the DJ aims to read the air of the situation to select the appropriate record or transition for the moment, to grasp and transform the audience’s current vibe. In Rainald Goetz’s novel Rave, we find a peculiarly concise description of the artistry of a DJ: In the situation described, the DJ is in danger of losing grip of the crowd’s vibe. He pre-listens to what might be the next record, to find out if “the new piece actually click[s]” (Goetz, 2020, page 79).

First, it is important to note that the DJ in this description is not merely searching for a record that simply fits the one currently playing; rather, he is looking for the right record “given the present situation of the party” and the way he wants to influence the party’s atmosphere musically. What he thereby aims to follow intuitively are “principles of consonance,” which do not merely relate to the composition of the line-up of records he might have anticipated while listening at home. Rather, he attends to the “multiple dimensions of the transitory rules governing the public’s reaction that are here and now momentarily in force” (Goetz, 2020, page 81 [my emphasis, TP]). At the same time, a good DJ not only reacts to and manipulates the crowd’s vibe in a given situation. They also fulfill what can be described as an “educational function:” A good DJ also aims at influencing and transforming the very conditions shaping the ways their audience listens to music, how they are able to switch from one musical vibe to another, or even to enhance their ability to immerse themselves in listening to rhythms and musical styles they were previously unfamiliar with.

In the following, these two ways in which a DJ influences the audience’s atmospheric experience are considered paradigmatic of how contemporary AI-supported applications increasingly manipulate the atmospheric dimension of our experience. However, unlike DJs, whose influence we consciously expose ourselves to, AI-powered applications manipulate us without us noticing or consenting to this influence. In this sense, they increasingly operate as “dark DJs.” On the one hand, they influence the atmosphere in which we find ourselves in a given situation by selectively presenting content. On the other hand, they transform and manipulate the very way in which we experience situations atmospherically. For reasons that will become clearer below, these two forms of manipulation are referred to as two forms of background exploitation—namely, atmospheric background hijacking and atmospheric background entrapment.

Atmospheres are ubiquitous phenomena. We hardly ever encounter a situation without an atmosphere (Bégout, 2020). This also holds for works of art, for novels, films, installations, and video games. As Niklas Luhmann (1981, page 176) argues, even abstract theories have atmospheres, “Weltstimmungsgehalte,” “world mood contents,” although they often seem unable to observe their own atmospheres. However, the ontological and epistemological status of atmospheres may appear somewhat puzzling. Having ingrained rather naturalist convictions, we tend to think that atmospheres are not really “out there” in the world, as they do not seem to exhibit the robustness and physicality of material objects and have a somewhat “psychic” feel to them. By contrast, we experience atmospheres not merely as subjective projections onto our surroundings. We experience them as permeating the situations we confront, as being really out there, also perceivable by others, not merely in our heads. This is even the case when we do not experience ourselves as involved in a particular atmosphere. We might enter a room and immediately grasp the atmosphere of sadness emanating from the three people sitting in the corner while feeling quite happy at the same time (Fuchs, 2013, page 22; Schmitz, 2019, page 134f.; Slaby, 2019).

This somewhat unclear onto-methodological status of atmospheres has led many phenomenologically leaning (this includes “4E”-approaches) theorists to argue that we should “attempt to behold and grasp atmospheres in their experiential presence before explicit reflection and conceptualization set in” (see also Böhme, 2017, page 17; Slaby, 2019, page 274). Following Edmund Husserl’s methodological guidance, the phenomenologically motivated advice thereby consists in performing a “phenomenological epoché” in which we “bracket” all ontological prejudices that go hand in hand with what he describes as the “natural attitude” (Husserl, 1983, page 58f.). Among other prejudices, we thereby bracket our naturalist ontological assumptions, often closely linked to a scientific image of the world, in order to free our view to see our “lifeworld,” our lived world of experience with all its experiential nuances and intricacies, all its relational, social, and atmospheric properties. As Husserl emphasizes, one of the essential epistemic characteristics of the entities belonging to this lifeworld is that they are experienced with the “experiential sense of ‘there for everyone’” (Husserl, 1973, page 5 [my translation, TP]; Moran, 2017, page 117). Thus, if we put aside our ontological prejudice regarding the existence of atmospheres stemming from our natural attitude, we experience them “as tangible, forceful, qualitative ‘presences’ in experiential space” (see also Bégout, 2020, page 199f.; Schmitz, 2019; Slaby, 2019, page 274), also perceivable by others. Accordingly, atmospheres have been described as objective, intersubjectively shared qualities of situations or spaces (Böhme, 2013; Fuchs, 2013; Schmitz, 2019; Slaby, 2019). Along this line of thought, “quasi-objectivity” has been proposed as one of the core characteristics a theory of atmospheres has to accommodate (Fernandez Velasco & Niikawa, 2025).

The presupposition of atmospheres as quasi-objective, however, becomes problematic if it goes hand in hand with a refusal to reflect on the fact that immediate experience is always mediated by conditions that can differ intersubjectively and that are themselves not part of what is given in experience. A lifeworld is not simply out there, ready to be grasped by whoever comes along; it is only to be directly experienced by those who have been socialized and enculturated accordingly. Famously, Pierre Bourdieu describes such conditions mediating immediate experience as the “habitus” of a subject (Bourdieu, 2006, page 54), which he defines as a “system of dispositions” in the “form of schemes of perception, thought and action.” According to him, these internalized schemes “ensure the active presence of past experiences” (Bourdieu, 2006, page 54) by shaping our immediate experience and understanding of situations, actions, and objects. Although Bourdieu himself highlights that ordinary behavior and interaction are “accompanied by the application of a spontaneous semiology” (Bourdieu, 2010, page 10) of experience provided through the habitus, the notion of the habitus is frequently interpreted as being primarily associated with internalized patterns of behavior and not so much with the way we intuitively experience and understand situations. To avoid such a misunderstanding, the conditions that mediate immediate experience at the individual level are in the following referred to as the “background.”1 The background hypothesis is thereby limited to two theses: First, that our immediate experience is mediated by conditions that are not themselves part of what is presented in experience. Second, the background is at least partially shaped by socialization and enculturation. Intersubjective differences in background can thus lead to diverging ways of grasping and understanding the same circumstances.

Now, if this is true, then we are precisely not able to recognize these mediating conditions from the perspective of the phenomenological attitude. Bourdieu therefore criticizes the phenomenological perspective for excluding “the question of the conditions of possibility of this experience,” namely, to recognize “the internalized structures” which provide “the illusion of immediate understanding, characteristic of the practical experience of the familiar universe” (Bourdieu, 2006, page 26).2 In phenomenological accounts, this methodological limitation often translates into a substantive presupposition: the assumption of a shared intersubjective or common (life)world, thereby discounting the possibility of diverging backgrounds. However, while it is true that we ordinarily experience situations, atmospheres, and entities with the “experiential sense of ‘there for everyone,’” such an experiential sense licenses no straightforward inference to ontological “quasi-objectivity.” As already stated, such an assumption is particularly problematic if it cannot be ruled out that the backgrounds mediating immediate experience might differ intersubjectively. Take the example of a cultural affordance: In many cultures, we find a tendency to perceive cups as affording to drink hot (but not cold) beverages from them in respective contexts.3 If we are socialized or enculturated in the respective manner, we intuitively experience cups as entities from which “one” is to drink hot beverages, but not cold ones. Frustrations of this expectation can accordingly lead to social sanctioning behavior. From a phenomenological perspective, the respective affordance appears as “quasi-objective,” as part of a common world. However, the establishment of cultural practices and customs hardly ever goes so far that we find no regular deviations within what can only be counterfactually assumed to be a homogeneous culture or lifeworld. The presupposition of common ground in the form of an intersubjective lifeworld is therefore always shaky (Luhmann, 1986). This holds even more true for atmospheres: On the one hand, atmospheres are much more ephemeral entities than cups and the cultural forms of use we intuitively associate with them. On the other hand, the way we grasp atmospheres even more clearly relies on our backgrounds, since atmospheres can hardly be argued to exist in any physically tangible sense. Atmospheres are not out there as such; they are out there for beings like us.

From this perspective, we can answer the question of the onto-methodological status of atmospheres more precisely, before we focus on atmospheres themselves: If their immediate experience is mediated by individual backgrounds, atmospheres are neither, as some phenomenologists would have us believe, simply quasi-objectively out there in our common lifeworlds, nor are they, as naturalists would have us believe, merely in the minds of individual subjects. Rather, atmospheres concern how we humans, mediated by our individual backgrounds, immediately experience situations.

Different subjects can experience an atmosphere in similar ways if they share sufficiently similar backgrounds with respect to salient characteristics of the circumstances they face. This phenomenon is what the Japanese refer to when they describe someone as “KY,” which stands for “kuuki yomenai,” someone who cannot read the air of situations, oftentimes referring to foreigners who “have a different Background” so that they “just bring a different mode of sensibility to bear” (Searle, 1998) on situations. We should therefore avoid two mistakes: 1. Naively conceiving atmospheres as quasi-objective characteristics of an inherently intersubjective lifeworld. 2. Pretending that atmospheres are only in our heads, not phenomena we directly experience.

To understand how and why atmospheres are central to human experience, we need to recognize the extent to which the “appresentation”—or, to use a different terminology, the “amodal perception” (cf. Nanay, 2010)—of “invisible” and “nonexistent” features constitute an integral part of ordinary experience. Wilfrid Sellars gives the example of the perception of “snow seen on a distant mountain” (Sellars, 2002, page 422):

If we acquired familiarity with certain types of situations, events, or things and their typical horizons of contiguity (the range of qualities, things, events, and objects that usually go along with them), we tend to also immediately experience their “invisible” (Musil, 1996a, page 573) features without having to consciously infer them. Past experiences sediment into the background in ways that shape our immediate experience by appresenting non-visible material and temporal parts and features of entities. Robert Musil exemplifies this phenomenon with the following examples:

What Musil describes as “invisible” are thus appresented features of an entity that can be brought into experiential view in the further course of experience as they belong to the experienced entity. Such features are usually not consciously inferred, believed in or postulated; they are—as phenomenologists prefer to say—immediately appresented (from Latin ad-praesentare) in experience.

However, as Musil adds, we not only appresent “invisible” aspects of things, but also features that are, at least from a physical or naturalist perspective, “nonexistent” [“das gar nicht Vorhandene”] (Musil, 1996a, page 573) in our current surroundings: One can think here of characteristics and features that, in our experience, frequently appear in connection with certain events, situations, or objects and are therefore experientially contiguous with them. Thus, we may typically associate beaches with palm trees, cocktails, and vacation, such that perceiving one of these elements immediately appresents the others. Or we may have internalized the typical song structures within a musical genre to such an extent that we intuitively anticipate the beginning of the chorus when listening to an unfamiliar song. In the same way, we also learn to associate things that potentially belong to or fit with typical situations and events, just as we develop a sense of what might come next in familiar situations. Appresentations can also be more idiosyncratic: For each of us, certain scents, for example, can evoke not only memories of specific people but also an entire cloud of elements, feelings, gestures, sounds, and colors, entire atmospheres linked with them and with the times when they were significant to us (Allen, 2014, page 148). In this way, everything we experience is involuntarily accompanied by anticipated horizons of what might come next as well as by clouds of appresented elements that are somehow linked with what we experience. Therefore, atmospheres rather belong to the realm of feeling than to the realm of explicit cognition, insofar as they involve intuitively sensed tendencies and associations rather than explicitly reflected and thought-out ones. To borrow a formulation Peter Strawson uses to describe the way imagination is constitutive for the perception of permanent objects, one can say that every experience “is, as it were, soaked with or animated by, or infused with—the metaphors are à choix—the thought of other past or possible” (Strawson, 1970, page 41) elements that are usually associated or contiguous with what is experienced. Beyond strictly perceived elements, atmospheric experience relies on the dimension of involuntarily appresented invisible and nonexistent elements.

It is this co-presentation of horizons of anticipations and clouds of associations of contiguous elements that accounts for the elusive and ephemeral character of the atmospheric. When we see a coffee cup, we do not just see a material object, nor do we solely grasp the affordance to drink hot beverages from it; we also immediately grasp the ephemeral atmospheric contexts to which the cup might belong, what might fittingly be done with it, and so on. Thus, we see the cup and its features in the light of the appresented horizons and clouds of associations. In this sense, atmospheres are directly experienced: Just as we immediately see sadness in another person’s facial expression or aggression in a raised fist, we immediately experience familiar situations as having specific atmospheres. We enter the room and immediately find ourselves influenced by the silliness our friends appear to be immersed in and entertaining. If we are familiar with situations such as parties, weddings, arguments, or funerals, we experience them directly in connection with particular atmospheric valences.

As already hinted at in the beginning, atmospheres have been philosophically conceptualized in a variety of ways. Most prominently, phenomenologists like Hermann Schmitz (2014) and Gernot Böhme have highlighted the fact that atmospheres, understood as “affective powers of feeling, spatial bearers of moods” (Böhme, 2017, page 16), play a crucial role in shaping our human experience of places and situations. Based on such an understanding, Thomas Fuchs defines atmospheres as “holistic affective qualities of spatial surroundings or interpersonal situations, that integrate their physiognomic, synesthetic and dynamic characteristics into a unified Gestalt” (Fuchs, 2013, page 22 [my translation, TP]). However, if the proposal that atmospheres are to be understood as being (even if just partly) constituted via appresentations of invisible and nonexistent elements is correct, it would be misleading to conceive them as Gestalt features, at least insofar as Gestalt features are usually conceptualized as higher-order constellational features grounded in (although not reducible to) lower-order perceptual features.4 Here, the understandable phenomenological impulse to describe atmospheres as something that is not in our heads but actually out there should not lead us to overlook the constitutive role that the dimension of appresented invisible and nonexistent elements plays in our atmospheric experience. Our respective backgrounds enable a “continuous decoding of the perceived” that is, however, “not consciously noticed” (Bourdieu, 2010, page 10). As Musil argues, atmospheric experience is “indebted to us” (Musil, 1996a, page 574); it is pre-reflectively read into the world to present it “in a light that emanates from ourselves” (Musil, 1996a, page 574). This light is never neutral but always carries inviting and soliciting, pushing or pulling qualities, valences, and relevancies (Lewin, 1917). This also explains why the atmospheric dimension of our experience is particularly susceptible to manipulation: the invisible and the nonexistent, which we involuntarily appresent in experience, depend on our backgrounds. And our backgrounds can be shaped and manipulated through habituation.

If the proposal put forward here is correct, atmospheres consist of more than merely the registration or integration of features of spatial surroundings or interpersonal situations into unified Gestalts. Rather, atmospheres are based on manifest and appresented elements, which are not simply registered neutrally, but are experienced in specific valences and relevancies. In this way, atmospheres belong to situations and objects as we directly experience them; however, they are nothing that we could directly point to in a given surrounding. As Musil describes it in Confusions of Young Törless in a scene where Törleß secretly watches his friend Beineberg from the side, the atmosphere of a situation, thing, or person is not revealed by a direct and focused gaze, “[b]ut precisely in that way—only half looking at him and half completing the picture in his imagination” (Musil, 2001, page 19). If one could subtract the invisible and the nonexistent from experience, the atmospheric would also disappear.

In his book Mythologies, Roland Barthes describes a whole range of objects, images, advertisements, and practices through the lens of the (culturally coded) atmospheres they evoke. What he analyzes as “myths” can, from the present perspective, be understood as culturally sedimented atmospheres. He gives the example of the atmosphere steak and chips had in his time within French society:

Another example he gives is the atmosphere of “Italianicity,” which was often evoked in French advertisements of his time by using specific names (like “Panzani”) or depicting certain colors and things (like spaghetti or tomatoes). Barthes argues that, in the French society of his time, Italianicity was an established atmosphere that could be evoked, alluded to, or played with to achieve specific atmospheric and affective effects. A French cliché of everything that’s Italian, so to speak. As he describes it: “Italianicity is not Italy, it is the condensed essence of everything that could be Italian, from spaghetti to painting.” (Barthes, 1977, page 48). What Barthes describes as Italianicity is therefore not something that tomatoes and spaghetti signify, tomatoes and spaghetti are not abstract symbols or signs that stand for Italianicity, rather, Italianicity is a specific evocative atmospheric assemblage of feelings, features, and elements—it names a culturally sedimented atmosphere that adheres to specific constellations of contiguous objects, rhythms, feelings, gestures, and so on. Whenever we are entrusted with design tasks, such as furnishing an interior, choosing the right elements for an occasion, preparing a party, or selecting appropriate clothing, we touch upon the dimension of the atmospheric. Whenever we follow “principles of consonance” (Goetz, 2020, page 80), which have more to do with intuition and gut feeling than with explicit thinking.

As previously suggested, the atmospheric dimension of our experience is an area where we are particularly susceptible to manipulation, as it depends on the backgrounds mediating immediate experience. The fact that human experience is mediated through backgrounds can be exploited by purposefully confronting people with content or situations that elicit specific behavior or are highly likely to affect them in a certain way. This form of manipulation is here referred to as background hijacking, insofar as it exploits preexisting background structures and the ways they shape experience. More specifically, atmospheric background hijacking involves the targeted confrontation of users with content, entities, or situations to exploit how they experience them atmospherically. Here, Barthes’ Mythologies present a whole arsenal of exemplary atmospheric entities and constellations, long exploited by advertisement and propaganda to strategically influence people’s behavior and experience. In today’s politically polarized situation, numerous everyday entities, behaviors, and situations—such as certain words, coronavirus masks, eating meat, carrying firearms, or displaying the rainbow flag—are charged with specific atmospheres and intuitively perceived in this way by corresponding audiences. The point here is not, however, that these entities or behaviors symbolize certain attitudes or worldviews in the abstract; rather, they are directly experienced by the respective groups as “loaded” or associated with specific affective and atmospheric valences. Regarding this phenomenon, Ludwig Wittgenstein gives the example of the experience of certain words and gestures: “Just think of the words exchanged by lovers! They’re ‘loaded’ with feeling. And surely you can’t just agree to substitute for them any other sounds you please.” (Wittgenstein, 1982, §712). We are confronted here with the effects of the intuitive somatic semiology provided by our backgrounds.

AI-supported content platforms can systematically exploit this fact by analyzing user data to microtarget content that selectively activates our pre-established atmospheric associations, thereby “pulling our strings” to influence us in desired ways. They can thus operate as dark DJs, playing specific content in specific situations to manipulate users’ behavior in an atmospheric way. On various content and social media platforms, algorithm-driven and often AI-supported decisions are continuously made about which content will be displayed to a consumer at a given moment, and these selections are frequently tailored toward particular forms of behavioral influence and affective engagement (Epstein & Robertson, 2015; Kremer et al., 2014). As whistleblower Sarah Wynn-Williams describes regarding the manipulation of affective states of users, Instagram and Facebook “monitor teenagers’ posts, photos, interactions, conversations with friends, visual communications, and internet activity on and off Facebook’s platforms and use this data to target young people when they are vulnerable,” especially “when they feel ‘worthless,’ ‘insecure,’ ‘stressed,’ ‘defeated,’ ‘anxious,’ ‘stupid,’ ‘useless,’ and ‘like a failure.’” (Wynn-Williams, 2025, page 333). One frequently cited example she gives is that “Facebook does work for a beauty product company tracking when thirteen-to-seventeen-year-old girls delete selfies, so it can serve a beauty ad to them at that moment” (Wynn-Williams, 2025, page 335). Although it is reasonable to assume that such strategies were already effective back in 2017 (for evidence for emotional contagion effects via social media, see Kramer et al. (2014)), with the help of large language models (LLMs)—trained and calibrated using individual and collective user data—in situ atmospheric exploitation nowadays already takes on entirely new dimensions (Afsar et al., 2022; Cai et al., 2023). Whereas classical recommender systems mainly predict which content is likely to be selected next (see also Picard, 1997, page 235), LLMs are particularly capable of modeling non-explicit semantic and associative structures because they are shaped by large-scale distributional patterns in training data, allowing them to encode patterns of association, contiguity, and contextual salience without these relations ever being explicitly represented or symbolically articulated. Modern, LLM-powered applications are thus much better equipped to track and manipulate what Goetz—regarding the DJ example—describes as the “multiple dimensions of the transitory rules governing the public’s reaction that are here and now momentarily in force” (Goetz, 2020, page 81 [my emphasis, TP]). Moreover, like real DJs, these applications are not limited to isolated interventions; instead, an advanced system could compose extended atmospheric content sequences, arranging content to orchestrate specific atmospheric or dispositional effects over longer periods.

This orchestration points to a more profound, arguably even more troubling form of background exploitation that is described here as background entrapment. Background entrapment exploits the fact that our backgrounds themselves can be transformed and recalibrated through experience, exposure, and habituation. It is a form of “long-horizon manipulative behavior” (Carroll et al., 2023) that can be brought about through algorithm- and AI-driven systems that aim toward recalibrating the background of users in a targeted and lasting way. Whereas background hijacking exploits pre-existing background structures in situ, background entrapment consists of their gradual recalibration over time, thereby reshaping the conditions through which events, situations, and objects are atmospherically experienced. In this way, applications are no longer only pulling our strings, but increasingly capable of modifying the very strings that can be pulled. An example of background entrapment regarding behavior is the way in which users are gradually accustomed to routine uses, hand gestures, etc., by their devices and applications, so that at some point they feel intuitively at home in their technical ecosystem and foreign in others. It is the kind of manipulation that Searle has in mind when he says that the “real disaster produced by television is not that children waste so many hours watching it, but that it alters their Background. It alters their whole mode of sensibility” (Searle, 1998, page 11). However, when one-to-many mass media still dominated the dissemination of information in society, such targeted recalibrations of the background could only be exerted in a relatively unfocused and homogeneous manner. In today’s digitalized world, where social media is increasingly moving toward personalizing and orchestrating the content shown to users based on individual and accumulated user data, it becomes increasingly possible to recalibrate the background of individual users in a targeted manner. Here, for the reasons already cited above, LLM-driven applications appear particularly capable of gradually reshaping the associative structures of users’ backgrounds through targeted, repeated exposure, thereby recalibrating the pre-reflective background conditions through which situations, objects, and actions are experienced. There are already empirical indications that several commercial platforms are in this way “switching to optimizing long term metrics” (Afsar et al., 2022; Cai et al., 2023; Carroll et al., 2023; see also Hansen et al., 2021), and background entrapment seems to be one promising path to reach this goal. Background entrapment that is specifically aimed at recalibrating the background regarding the atmospheric dimension of experience—i.e., the appresented horizons and clouds of anticipations and associations that are intuitively linked to behaviors, objects, and situations—is here referred to as atmospheric background entrapment. Think, for example, of the specific atmospheres associated with highly polarized subcultural milieus and the corresponding content that propagates these atmospheres and ingrains them into users’ backgrounds. Another example is the gradual drift through which users move from initially mainstream forms of doubt to more radical content. This logic of gradual recalibration applies directly to atmospheric background entrapment: for example through targeted, continuous exposure, a cheerful person may be gradually driven into a darker, more aggressive, or more insecure mindset (Jain et al., 2025; Kramer et al., 2014).

The general potential for manipulation and exploitation in the context of generative AI has been the focus of many contemporary debates (Carroll et al., 2023; Floridi, 2024; Ienca, 2023; Klenk, 2024; Weber-Guskar, 2024). However, current accounts tend to emphasize forms of exploitation, where AI-powered applications influence the current beliefs or emotions of their users in situ, which fall under the broad notion of background hijacking. In her influential book The Age of Surveillance Capitalism, Shoshana Zuboff highlights how the current form of digital and data-driven capitalism focuses on the “means of behavior modification” (Zuboff, 2019, page 292) to better control, predict, and exploit the behavior of users. Under the three forms of behavior modification she identifies, “tuning” (the subtle modification of behavior via cues and nudges given in situ) and “herding” (the modification of behavior by influencing the context of behavior through enabling or foreclosing action alternatives) fall under this notion of background hijacking. Against that, background entrapment is related to what Zuboff, following Burrhus Frederic Skinner, describes as the third form of behavior modification in surveillance capitalism, namely the “conditioning” of behavior (Zuboff, 2019, page 293).

However, the focus on background and atmospheric experience developed in this paper takes a more fundamental approach in that the manipulation of the background affects not only our behavior, but also, more fundamentally, the way we experience situations, events, and objects in the first place. Furthermore, the focus on the atmospheric dimensions of background hijacking and background entrapment complements an aspect central to our experience and, while clearly located on the “feeling” and “intuition” side of our experience, lacks the concreteness of emotions or beliefs.

As we have seen, AI-supported content platforms can, in a targeted manner, transform and manipulate our backgrounds to an unprecedented extent. Thereby, AI-supported microtargeting seems able to realize the slogan “If you are not paying for the product, you are the product” in a way we were not prepared for. It is no longer merely that we surrender our private data and thereby become a product for tech-companies so they can sell advertisements more efficiently; rather, these companies—and other actors operating in adjacent domains—may increasingly produce the way we pre-reflectively experience by intentionally shaping our backgrounds. One need not agree with every detail of the critique of the “culture industry” formulated by Max Horkheimer and Theodor W. Adorno in Dialectic of Enlightenment to endorse their claim that culture industry increasingly assumes a task Kant had, at least according to them, still attributed to the subject as an “active contribution,” namely “that they should, from the first, relate sensuous multiplicity to fundamental concepts” (Horkheimer & Adorno, 2002, page 98). According to Horkheimer and Adorno, the culture industry pre-schematizes sensuous multiplicity for consumers, thereby providing easily digestible content. Through atmospheric background entrapment, however, today’s culture industry appears capable of fulfilling this task not only with greater precision but also on a much deeper level: by shaping consumers’ backgrounds, it shapes the very conditions through which experience itself is mediated and schematized. Digital ecosystems already rely on having formed their users’ backgrounds in ways that make migration from one ecosystem to another difficult—a lock-in effect that is not only behavioral but also deeply atmospheric.

A final, anecdotal illustration can underline that the atmospheric dimension analyzed here is increasingly and explicitly recognized by actors operating at the intersection of data-driven prediction, AI, and behavioral influence as a central and particularly effective lever for the targeted manipulation of experience and behavior. Several figures close to the influential AI company Palantir—early Palantir employee Ryan Podolsky, current Palantir chief technology officer Shyam Sankar, and investor Christian Garrett—raised funds for a film production company, Founders Films. In the words of Shyam Sankar, the stated goal of this company is “to make content to make you proud to be an American. Not Pravda, entertainment” (Sankar 2025). As he puts it, what matters is “that there is a sense of heroism. It’s not just about war movies. It’s about what is the American spirit. And I like to say that, as a 5-year-old, I knew what it felt like to be an American before I knew civics, or political philosophy, or history” (Sankar 2025). The target of influence here is thus neither explicit belief nor the mere dissemination of narratives, but the targeted transformation of people’s pre-reflective, atmospheric experience.

If the previous reflections are correct, then the challenge posed by atmospheric background exploitation via AI cannot be met by the classical gestures of critique alone—by ever sharper reflection, demystification, or exposure of its mechanisms. Social media and content platforms now operate as unsolicited dark DJs, intervening not merely in situ but also through the slow, targeted recalibration of our backgrounds, thereby shaping the very world and atmosphere we immediately find ourselves in. What is at stake is thus no longer only the manipulation of current beliefs and emotions, but the targeted modification of the very air through which situations, events, and things appear to us. This demands a different kind of resistance—one less a matter of reflexive insight than of cultivating new habits of atmospheric attention: an ecological ethos of heightened awareness toward the socio-medial oikoi we inhabit and that shape not only who, but also how we are. With Musil, we may thus have to relearn how to perceive not only what is manifest, but also to refine our sensibilities toward the “invisible” and the “nonexistent” that silently shape the ways we find ourselves in the situations we encounter.