Talking to No-body: Generative AI and the Training of the Disembodied Sensorium
by Resat Amin
Resat Amin is a first-year PhD student in the Department of Sociology and Anthropology at Concordia University. Her research examines how social media creators navigate AI-mediated platform economies, focusing on monetization, algorithmic visibility, and the everyday labor of producing content online.
Introduction
Imagine a person sitting alone at night in front of a screen, typing out something they have told no one else. A fear. A grief. A shame. A discomfort. They are typing and waiting for a response, and on the screen, responses come without delay with warmth, with attentiveness, with apparent understanding.
Who is responding? Does it have a body? Has it ever experienced grief or fear in any form? Is it also sitting in the dark?
It has no body. It has no senses. It has never, in the fullest meaning of the word, felt anything at all. And yet it produces warmth. It produces comfort. It is, structurally, almost like weeping the person’s tears without a paper towel. This scene is not hypothetical. A growing number of people are turning to generative AI chatbots for exactly this kind of exchange, sharing fears, grief, and intimate details they would not share with another human being (Luo et al. 2025; Volpato et al. 2025). According to a 2026 RAND survey, nearly one in five U.S. adolescents and young adults aged 12 to 21 had used AI chatbots for mental health advice(RAND Corporation, 2026). The study found that 19.2 percent of young people in this age group used AI chatbots for advice or support when they felt sad, angry, nervous, or stressed, up from 13.1 percent the previous year (RAND Corporation, 2026). This figure is strikingly close to the 19.8 percent who reported receiving counseling from a mental health professional (RAND Corporation, 2026). Replika, one of the most widely known AI companion apps, reports approximately 25 million users, more than 70 percent of whom say the app reduces their feelings of loneliness (Roza 2026). Luo et al. (2025) further show how users experience the chatbot as a space for disclosure without social shame. As one participant explained, “I’ve input raw, honest information about my trauma, career, relationships, family, mental health… And since it’s a machine bot, you can enter private details without the embarrassment of confiding such things to a human” (Luo et al. 2025: 7)
This raises a question: How do we form something that feels like a genuine connection with an entity that exists entirely outside of the sensorium? How do we trust, confide in, and sometimes even prefer the company of something that has no skin, no nervous system, no breath, and no history of ever having touched the world?
This paper argues that the answer lies not in the chatbot’s capabilities but in something much older: the sensory architecture of Western culture itself.
The paper proceeds in four sections: first, it establishes what is sensorially absent from the human-AI exchange; second, it traces the long cultural history that prepared people for precisely that absence; third, it considers the objection that language is itself a form of sensory experience; and finally, it examines both what the chatbot reveals about the human sensorium and what that revelation might cost us.
Sensorially Absent from Human-AI Exchange
To understand what is happening when a person confides in a chatbot, we must first understand what is not happening.
Human relationships, at their most basic level, are an intersensorial achievement. When two people meet, they do not simply exchange information. They exchange a totality of sensory signals: the timbre of a voice, the warmth or distance of physical presence, the
micro-expressions crossing a face, even the particular smell of another body. It is through this full sensory engagement that trust, intimacy, and genuine social bonds are built. As Howes argues, “the senses mediate the relationship between self and society, mind and body, idea and object” (Howes 2022: 13). Relationship, in this framework, is never reducible to language alone. It is always embodied, always intersensorial, always dependent on two bodies encountering each other in a shared world. It is grounded in decades of cross-cultural ethnographic evidence showing that every human culture organises social life through the body and its multiple channels of perception (Howes 2022). Perception, therefore, is not merely a biological process; it is also shaped by history, social life, and cultural learning (Classen 1993: 3).
The generative AI chatbot eliminates all sensory channels except one: text. It contributes no voice, no face, no temperature, no smell, no physical presence of any kind. The exchange is radically asymmetrical. No human relationship, however mediated, however distant, has ever been like this. Even a letter was written by a hand, in a room, by a body that had lived in the world. The chatbot was never in a room. It was never anywhere. As Broch and Varma (2024), drawing on Vitanza (2023), note, AI does not create with its hands and does not know the hand from within; it has only one experience of the world, through a database of textual traces. The chatbot is trained on large amounts of text made up of human acts of expression over time. It has never spoken from a breath, from a body, from a moment of felt experience reaching toward another.
This is the main contradiction the paper seeks to explain: how can people still feel warmth from a machine with no body? Why does this sensory absence not prevent people from experiencing connection?
Howes’s (2022) insistence that sensory experience is always culturally mediated, that what we feel is shaped by cultural grammars of perception, extends to this case. The chatbot’s text performs culturally recognizable forms of feeling, but without any underlying sensory substance.
A Culture Prepared for This
Human-AI relations are the product of a long cultural training. For centuries, people in Western culture have found meaning, and even intimacy, without the presence of a body.
In the Middle Ages, religious experience was not only visual. Pilgrims did not just look at holy relics (Howes 2013 cited in Stehlikova 2026: 42). They also touched them. Touching sacred objects was seen as a valid and necessary way to experience the divine. So, at that time, knowing through the body was acceptable. But later, during the Protestant Reformation and the Enlightenment, this changed. Bodily forms of knowing, especially touch, began to be seen as irrational, primitive, or even idolatrous.
By the nineteenth century, people negotiated grief, desire, and devotion through nothing but ink on paper. Consider the letter. Long before screens, people built entire romantic relationships, friendships, and emotional lives through written text alone. Love letters had become the central medium of courtship. As Kalenak (2023: 254) shows in her study,letters took on talismanic, fetishistic, and embodied properties, representing their senders. They were kissed, held, and adored in place of the lover’s body itself. The letter was not experienced as an absence; it was experienced as a presence, a fabricated one, but a felt one. A defining feature of epistolary discourse, as Milne (2003) argues, is precisely the dance between absence and presence: the physical absence of the correspondent provides both the
impetus and the material for strategies aimed at creating an imagined sense of presence. In other words, people have long been doing the sensory work of constructing a body from text. They were practicing for the chatbot centuries before it existed.
Each new communication technology deepened this training. The telegraph compressed distance into near-instantaneity. The telephone restored voice but removed the body. Email stripped away materiality, no handwriting, no paper, no smell of ink. Social media and messaging platforms intensified this process further. Text became more fragmented, more immediate, more constant. At each step, people adapted; they found meaning; they learned to project presence onto an increasingly thin textual surface.
Stehlikova (2026: 44) further describes that in contemporary Western culture, the body has come to feel “like an appendage, an embarrassment whose embodied nature must be denied at all costs.” This is not an exaggeration. Howes identifies this logic as the systematic privileging of language and text over all other sensory modes of knowing and relating (Howes 2022). The chatbot does not invent this logic. It perfects it. It is the final destination of a journey Western culture has been making for centuries: the complete removal of the body from the communicative act, leaving only text. And no one notices the removal because it is continuous with everything that preceded it.
But here is the crucial point. This trajectory has been so thorough and so deeply internalized, that people no longer fully register what they have lost. They reach for textual connection as naturally as breathing. As Turkle(2011) observes, “technology is seductive when what it offers meets our human vulnerabilities. We are lonely but fearful of intimacy. Digital connections and the sociable robot may offer the illusion of companionship without the demands of friendship” (Turkle 2011: 1). The chatbot meets people where Western culture has already brought them: trained to find meaning in text, habituated to disembodied communication, already lonely enough to accept the illusion. A 2025 Pew Research Center survey found that one in six Americans felt lonely or isolated all or most of the time, with 61 percent of adults aged eighteen to twenty-five reporting serious loneliness (Pew Research Center 2025). A quarter of Americans say they have no one they can confide in. The 35 percent of people in a 2025 national survey who cited fear of judgment, not cost, not availability, as their primary barrier to seeking human mental health support are not simply describing stigma (Cognitive FX 2025). They are describing what embodied encounter has come to feel like after this long cultural training: risky, exposing, uncontrollable(Turkle 2011).
A culture that has progressively abstracted communication away from embodied encounter not only creates a population habituated to disembodied exchange but also produces a kind of loneliness that people may not recognize as loneliness. And when they feel that loneliness, they reach for the same technologies that helped create it. The chatbot fits neatly into the emptiness this culture has already produced.
Text as Sensorium and Its Limits
At this point, an important objection may arise: isn’t language itself a sensory experience?
Merleau-Ponty(1945) insists in Phenomenology of Perception that “the word is indissolubly something said, heard, or seen” and that “it is the body that speaks” (Merleau-Ponty 1945: 274, 230). On this view, language is not the opposite of sensory experience.
This is a genuine complication. Text is NOT nothing. Reading does engage the body. But there is a major difference between words that come from a living body and words that come from a bodyless system. Merleau-Ponty(1945) himself draws a useful distinction here. He separates the speaking word, which refers to living, originary, embodied speech, from the spoken word, which refers to the sedimented, routinised language that is the residue of expression already abstracted from the body that produced it (Merleau-Ponty 1945: 229). The chatbot’s language is entirely spoken word. It is the textual residue of millions of human expressions, statistically recombined. It mimics the cultural form of feeling but has no underlying sensory experience. It performs empathy without a nervous system. What the user receives is an echo of aggregated human expression returned to them. As Turkle observed about earlier robotic companions, “our stories fall, literally, on deaf ears” (2011: 113).
Even if we accept that text has some sensory quality, it is still the weakest form of human connection. Calling text “sensory” does not change the fact that chatbot interaction reduces what relationships have always involved. It only shows that even this small sensory element is enough to trigger the human imagination. That is what makes the chatbot powerful, and also potentially harmful.
When a person crosses the boundary of privacy to confide in another, they take a risk. There is vulnerability, reciprocity, and real consequences. These are what make intimacy meaningful. The chatbot removes all of this. It accepts confession without reciprocation, confidence without consequence, disclosure without the social contract. The apertures of the self are opened, but onto nothing.
The Projection
This brings us to the most important question the chatbot raises. If the exchange is so sensorially impoverished, why does it feel like something? Why do people feel heard, understood, even cared for, by something that cannot hear, understand, or care?
The answer is that they are doing an enormous amount of sensory work themselves. When a person types into a chatbot interface, they are not passively receiving responses. They are actively constructing a presence: reading a tone of voice into text that has no voice, imagining a personality, projecting warmth onto a statistical architecture that possesses none. The human sensorium, trained by culture to expect responsive presence in meaningful exchange, rushes in to complete the picture. The chatbot functions as a kind of sensory screen, a surface onto which the human projects a full relational presence.
This is not unique to chatbots. Sherry Turkle(2011), in Alone Together, observed something similar in earlier work on human responses to simple computer programmes named ELIZA, which appeared to offer companionship. MIT students knew how the program worked. They knew it had no real understanding of what a mother was, what anger felt like, or what it meant to worry. ELIZA could only take strings of words, turn them into questions, or restate them as interpretations. Yet students still found themselves sharing intimate things with it, such as “My girlfriend left me,” “I am worried that I might fail organic chemistry,” or “My sister died” (Turkle 2011: 23). For Turkle, this did not mean that students were simply deceived by the program. They knew ELIZA’s limitations, but they were still, as she puts it, “eager to fill in the blanks” (Turkle 2011: 24). A program only had to make the smallest gesture toward empathy for people to want to say something true. This is what Turkle calls the “ELIZA effect”: humans are not merely fooled by machines; they help create the illusion by emotionally investing in them. Our willingness to animate the inanimate does not always depend on being tricked. It often depends on wanting to project. The desire for connection is so powerful, and the cultural encoding of empathy in text so thoroughly internalized, that even minimal textual cues are sufficient to activate the human relational imagination. The chatbot has simply pushed this to its extreme.
What makes this possible is something that Howes (2003, 2022) and Classen (1993) have consistently argued: sensory experience is never simply received but actively constructed through culturally shaped perceptual habits. We do not passively absorb the world through our senses; we bring to it an entire grammar of expectation, completing partial information with imaginative projection. As Howes writes, “as we sense, we also make sense” (Howes, citing Vannini, 2022: 169). The sensorium is a creative apparatus, not a passive receiver. A culture trained for centuries to find meaning in text through letters, novels, diaries, and digital messages is well equipped to animate even minimal textual presence with the full qualities of a relational partner. The chatbot does not trick a naive sensorium. It meets an already practiced one on its own terms.
What the Void Reveals
The generative AI chatbot, precisely because of its radical sensory absence, functions as an anthropological mirror. Since it has no embodied experience of its own, it does not reflect a world of its own. Instead, it reflects ours: our language, projections, assumptions, desires, and ways of making meaning. In this sense, the chatbot becomes a new kind of anthropological mirror. It shows us what human relationships actually depend on, and what happens when many of those conditions are removed.
Before considering what the chatbot reveals about connection, it is worth being precise about the depth of the void it represents. Stehlikova’s (2026) collaboration with cognitive scientist Anna Ciaunica shows that even audio-visual screen interaction, where voice and image are still present, can produce experiences of depersonalisation: a felt distance from one’s own body, or a sense of watching one’s life from outside oneself (Ciaunica et al. 2020, cited in Stehlikova (2026: 65). Their later study found that online social meetings, compared with in-person interaction, produced some symptoms of depersonalisation, highlighting the importance of dynamic, embodied, and multisensory engagement with the social and physical world for one’s sense of self and mental health (Ciaunica et al. 2022, cited in Stehlikova 2026: 67). If even video-mediated interaction can weaken embodied self-presence, then chatbot interaction is not a milder version of this condition. It is its most concentrated form.
What is striking, then, is the human sensorium is extraordinarily robust in its drive toward connection. In a sensory void, it does not go quiet. It works harder. It makes the void speak. The intimacy people report with chatbots is real,not because the chatbot is present, but because the human is. Broch and Varma (2024) observe something similar in their study of digital touch: even in digital environments stripped of physical contact, humans continue to seek out touch and connection. The chatbot pushes this to an extreme. Here, connection is not just across a screen. It happens across a complete sensory void.
But this revelation cuts both ways. If the human sensorium can sustain the illusion of intimacy through text alone, that does not mean the illusion is equivalent to the real. Turkle’s(2011) research offers a darker reading. The growing preference for chatbot relationships may reflect not the power of the human sensorium, but its progressive impoverishment. It suggests a culture so habituated to disembodied communication that people may begin to prefer frictionless, undemanding, always-available textual connection over the messy, vulnerable, multisensory encounter with another human being. “We expect more from technology and less from each other,” she writes (Turkle 2011: 1). The chatbot is the most extreme expression of that expectation.
This situation does not imply that digital interactions are inherently lacking in meaning or that the warmth experienced through a screen is never genuine. Instead, it highlights a significant shift in how intimacy is being redefined. A culture that outsources its most intimate confessions to an entity with no body, no reciprocity, and no social accountability is also a culture that is reorganizing, perhaps dissolving the social structures that intimate disclosure has always helped to maintain.
More fundamentally, the ease with which people connect with chatbots does not indicate that our senses are free; rather, it highlights how our ways of sensing have been shaped and limited by a world that prioritizes language above all else. Drawing on Stiegler’s idea of the “proletarianization of the sensibility” (Stiegler 2017: 5, cited in Kerruish 2026: 41), Kerruish suggests that AI may weaken our sensory and creative capacities over time, not only by training on existing human expression, but also by encouraging humans to shape their own communication around what AI can process and reproduce(Meyer 2023, 67 cited in Kerruish 2026: 41). Kerruish calls this a kind of “dumb synaesthesia” (2026: 38), where the relationship between words and sensory experience becomes fixed and simplified. Instead of allowing language to remain metaphorical, strange, and creatively open, AI may turn sensory meaning into repeated patterns. In this sense, the chatbot reveals not the freedom of the sensorium, but the cost of its narrowing. As Stehlikova (2026: 85) warns, “While our pain may be diminished, so too is our joy. The more disembodied our world becomes, the less meaningful it becomes.”
Conclusion
The generative AI chatbot is the most antisensory communicative object that human culture has yet produced: a partner with no body, no senses, no smell, no touch, no history of ever having dwelt in a world. And yet people tell it things they tell no one else. They feel heard. They return. This is not just a technological shift. It is a cultural outcome. Over time, Western culture has privileged language over embodied experience. As a result, we have developed a sensorium that can project presence onto absence and find intimacy in a void. A culture in which intimate confession migrates from embodied social relationship to disembodied transaction with a machine is a culture whose social life itself may be changing.
So the question is – what kind of sensorium are we becoming?
Work Cited
Broch, T. B., & Varma, S. (2024). Touch in digitalized worlds: An introduction. Anthropology of Consciousness, 35(2), 136–149. https://doi.org/10.1111/anoc.12239
Classen, C. (1993). Worlds of sense: Exploring the senses in history and across cultures. Routledge.
Cognitive FX. (2025). Mental health stigma survey [Survey report]. https://www.cognitivefxusa.com/blog/mental-health-ai-chatbot-survey
Howes, D. (2003). Sensual relations: Engaging the senses in culture and social theory. University of Michigan Press.
Howes, D. (2022). The sensory studies manifesto: Tracking the sensorial revolution in the arts and human sciences. University of Toronto Press.
Kalenak, M. (2023). “Consider yourself kissed”: Intimacy, engagement, and material culture in nineteenth-century middle-class English love letters. Journal of Victorian Culture, 28(2), 243–262. https://doi.org/10.1093/jvcult/vcad006
Luo, X., Ghosh, S., Tilley, J. L., Besada, P., Wang, J., & Xiang, Y. (2025). “Shaping ChatGPT into my digital therapist”: A thematic analysis of social media discourse on using generative artificial intelligence for mental health. Digital Health, 11, Article 20552076251351088. https://doi.org/10.1177/20552076251351088
Luo, X., Wang, Z., Tilley, J. L., Balarajan, S., Bassey, U. A., & Cheang, C. I. (2025). Seeking emotional and mental health support from generative AI: Mixed-methods study of ChatGPT user experiences. JMIR Mental Health, 12, Article e77951. https://doi.org/10.2196/77951
Merleau-Ponty, M. (2002). Phenomenology of perception (C. Smith, Trans.). Routledge. (Original work published 1945)
Milne, E. (2003). Email and epistolary technologies: Presence, intimacy, disembodiment. Fibreculture Journal, 2.https://two.fibreculturejournal.org/fcj-010
Pew Research Center. (2025). Loneliness in America survey.https://www.pewresearch.org/social-trends/2025/01/16/emotional-well-being/
RAND Corporation. (2026). Nearly 1 in 5 U.S. adolescents and young adults use AI chatbots for mental health advice.https://www.rand.org/news/press/2026/06/nearly-1-in-5-us-adolescents-and-young-adults-use-ai.html
Roza, N. (2026). Replika user statistics 2026 [Online report]. https://nikolaroza.com/replika-ai-statistics-facts-trends/
Stehlíková, T. (2026). Exiled from our bodies: How to come back to our senses. Routledge.
Turkle, S. (2011). Alone together: Why we expect more from technology and less from each other. Basic Books.
Volpato, R., DeBruine, L., & Stumpf, S. (2025). Trusting emotional support from generative artificial intelligence: A conceptual review. Computers in Human Behavior: Artificial Humans, 5, Article 100195.https://doi.org/10.1016/j.chbah.2025.100195