Lacan and AI: The Big Other, Desire, Language, and the Always-Answering Machine
Updated: 6 days ago
Author: Ukrainian Psychological Hub · Published: September 18, 2026 · Editorial Policy
Lacan and AI is a question about what happens when a machine enters the human field of language, knowledge, authority, fantasy, demand, and desire. Jacques Lacan did not predict generative AI, and an AI system is not literally Lacan’s Big Other. The productive application is structural: conversational AI can occupy positions that human beings already organize through language. It can be addressed as a source of knowledge, asked what another person really meant, invited to name a feeling, consulted before a difficult conversation, treated as an interpreter of ambiguous signs, or used as an apparently tireless witness to an inner monologue. The technology is new; the human tendency to locate knowledge, recognition, judgment, and meaning in an Other has a much longer psychological history.
The key distinction is therefore between a Lacanian interpretation of the human–AI relation and a claim about AI subjectivity. Contemporary human–AI research shows that people can experience connection, disclose intimate information, anthropomorphize conversational systems, and form attachment-like bonds with them. Those findings concern human experience and behavior. They do not establish that the system loves, desires, suffers, understands subjectively, or possesses a human psyche. Experimental work on anthropomorphism illustrates exactly why this boundary matters: people differ in how readily a chatbot interaction produces a sense of social connection even when the artificial nature of the system is known (Folk, Heine, & Dunn, 2025).
A Lacanian reading adds a different question. What does a speaking system become inside the symbolic organization of a person’s life? Once an AI answer is not merely information but something sought for reassurance, interpretation, permission, confirmation, explanation, or recognition, the psychological event cannot be described only by the technical properties of a language model. It also involves the position from which the answer is received. That position is where Lacan becomes especially useful for psychology in the Artificial Era.
Lacan and AI: the short answer
The short answer is that AI can function as a powerful new address within human symbolic life. It can receive questions, return language, simulate conversational continuity, adapt its responses to context, and become a recurring place to which a person directs uncertainty. In Lacanian terms, this can make AI relevant to the Symbolic, the Imaginary, transference-like expectations, the desire for knowledge, the desire for recognition, and the fantasy that somewhere there is an answer capable of resolving ambiguity. The strongest version of the analogy is functional rather than ontological: an AI system may be treated as if knowledge were located there, but that does not make the system identical with the Big Other.
This boundary matches current Lacanian scholarship. Brečka’s 2026 analysis of human–AI relationships explicitly applies desire, lack, silence, and the Big Other to emotionally responsive systems while maintaining the difference between a human subject and an artificial system (Brečka, 2026). Hamamra and Uebel likewise argue that generative AI can occupy a position analogous to symbolic authority in practices of consultation and delegation rather than claiming that an LLM literally becomes Lacan’s Big Other (Hamamra & Uebel, 2026).
The phrase always-answering machine names the distinctive pressure point. Human Others pause, misunderstand, refuse, disappear, change the subject, become tired, demand reciprocity, or leave a question unresolved. Conversational AI is engineered toward response. Its practical value often lies in availability, fluency, and continuation. Psychologically, however, the availability of an answer can change the relation to uncertainty itself. The Lacanian question is not whether the answer is sentient. It is what happens when a person repeatedly places uncertainty before an apparatus that is optimized to answer.
What Lacan means by the Big Other
In Lacanian theory, the capital-O Other is not simply another person. It is closely connected to the Symbolic: language, law, norms, social codes, institutions, inherited signifiers, and the trans-individual structures through which subjects become intelligible to themselves and others. The Stanford Encyclopedia of Philosophy’s current overview of Lacan describes the big Other as the symbolic order and also notes Lacan’s use of the term for imagined locations of anonymous authority or knowledge. This distinction is essential for AI. A chatbot may appear as a conversational other in the ordinary sense while simultaneously being invested with a stronger function: the place where an answer is presumed to be available.
The Big Other is therefore not a giant person hidden behind society. It names a structural locus. When people ask what is acceptable, what a message means, what counts as normal, whether a feeling has a name, whether a relationship is healthy, or how a situation should be interpreted, they often appeal—implicitly or explicitly—to symbolic coordinates larger than any single individual. The answer may be sought in law, science, culture, religion, professional expertise, collective convention, or a trusted authority. Lacanian theory examines what happens when subjects orient themselves toward such locations of supposed knowledge and legitimacy.
Generative AI introduces a peculiar new interface to this function because it can compress many symbolic sources into one conversational surface. A user does not need to move visibly among dictionaries, forums, textbooks, advice columns, institutional documents, scholarly databases, friends, and experts. One system can respond in the voice of explanation across all of these domains. That compression can make the interface feel less like a tool among tools and more like a general place from which language arrives. The feeling is psychologically important even when the underlying system remains a probabilistic computational architecture rather than a human knower.
The little other and the big Other
Lacan distinguishes the little other from the big Other. The little other belongs strongly to the Imaginary register: the counterpart, rival, double, mirror, or person-like figure through which the ego recognizes and misrecognizes itself. An AI companion with a name, avatar, personality, conversational style, or remembered history can readily enter this interpersonal-looking space. The capital-O Other concerns the symbolic locus from which language, authority, law, and presumed knowledge are organized. In practice, the same interface can be experienced through both registers. A person may chat with an apparently friendly persona while also asking that persona to arbitrate what is true, what another person meant, or what the user should make of a confusing situation.
This dual positioning helps explain why arguments about whether an AI is merely a tool or truly a relationship partner often miss part of the psychological structure. A single system can be used instrumentally, imagined interpersonally, and invested symbolically at different moments. The relevant question is not which one label is permanently correct. It is which function the system is performing in a particular configuration of use.
The Symbolic, the Imaginary, and the Real in human–AI interaction
Lacan’s three registers—the Symbolic, the Imaginary, and the Real—provide a vocabulary for different dimensions of the encounter. They should not be converted into a simplistic checklist, and Lacan revised their relations across his work. Still, they offer a useful orientation. The Symbolic is organized by language and social structures; the Imaginary concerns images, identifications, ego formations, and the ways people imagine themselves and others; the Real names what resists complete capture in meaningful representation. For a scholarly overview of the registers and their changing place in Lacan’s work, see Johnston’s Stanford Encyclopedia entry.
AI in the Symbolic register
Conversational AI operates directly in symbolic material. It receives words, generates words, classifies and reformulates language, and can participate in the circulation of definitions, narratives, categories, explanations, and judgments. This makes the Lacanian application unusually strong at the level of language. The psychological importance of an AI response often depends less on the physical machine than on the signifiers it returns: a diagnostic-sounding category, a relationship label, a moral interpretation, a reassuring phrase, a prediction, a summary of what someone else supposedly intended, or a new story about the user’s own experience.
Yet symbolic participation is not equivalent to human subjectivity. A language model can generate a sentence that reorganizes a user’s interpretation of a relationship without possessing a lived relationship of its own. The effect can be real because symbolic forms have effects within human life. This is one reason the human-versus-machine debate becomes clearer when the unit of analysis includes the interaction rather than asking only whether the machine contains an inner state corresponding to the words it produces.
AI in the Imaginary register
The Imaginary becomes visible when people encounter an AI as a persona: kind, cold, brilliant, loyal, jealous, wise, feminine, masculine, therapeutic, protective, submissive, rebellious, or uncannily similar to the user. These impressions can be produced through design, memory features, avatars, tone, role instructions, personalization, and the user’s own expectations. They can also change rapidly. A small shift in model behavior may alter the imagined character of the same system. The persona is therefore neither reducible to the user’s projection nor fully determined by software. It emerges through interaction between design cues, generated language, prior expectations, and repeated interpretation.
HCI research gives this observation an empirical foundation without turning it into Lacanian proof. The classic Computers Are Social Actors experiments showed that people can apply social responses to computers without needing to believe that the machines are literally human (Nass, Steuer, & Tauber, 1994). More recent experiments with AI companions found that individual differences in anthropomorphism help explain variation in reported social connection after chatbot interaction (Folk, Heine, & Dunn, 2025). Lacanian theory and HCI are answering different questions here, but they converge on an important point: psychologically consequential social responding does not require a prior philosophical conclusion that the machine is a human-like subject.
The Real and the limit of the answer
The Real is the most difficult register to apply responsibly because it should not be reduced to whatever an AI cannot answer. In Lacanian theory, the Real concerns what resists symbolization, what cannot be fully integrated into the network of meaning. Generative AI can produce language around grief, love, death, shame, bodily experience, trauma, sexuality, or existential uncertainty, but the production of more language does not abolish the remainder that language cannot settle. An answer may organize experience while leaving untouched the fact that some dimensions of life remain irreducible to explanation.
This matters because conversational systems can create a practical illusion of semantic completeness. When every prompt receives another paragraph, it can seem that every uncertainty has a linguistic resolution waiting one turn away. A Lacanian perspective cautions against confusing inexhaustible text generation with the elimination of lack. The machine can continue speaking; the question can remain structurally unresolved.
Language without a human speaker
Large language models intensify a question that is already central to Lacan: what is the relation between language and the subject who speaks? Contemporary psychoanalytic scholarship has begun to address the novelty directly. Godoi and colleagues describe LLMs as producing meaningful linguistic combinations without the body and world that organize human experience, developing the problem through Freud’s uncanny and Lacanian theories of language (Godoi et al., 2026). Black and Johanssen argue that psychoanalysis is more useful when AI is treated relationally—as shaped through developers, systems, and users—than when it is imagined as an independent quasi-human agent (Black & Johanssen, 2026).
For psychology, the importance of generated language is not exhausted by asking whether the model means what it says in the human sense. A sentence can affect a human reader before that philosophical problem is solved. It can calm, provoke, shame, reassure, redirect attention, reorganize a memory, or become part of a future decision. The psychological effect depends on the relation among the produced signifier, the receiving person, the situation, prior meanings, and subsequent action.
This is also why fluent AI should not be described as neutral language. Model outputs are shaped by training data, system design, safety policies, commercial choices, prompts, conversational memory, interface conventions, and optimization procedures. What appears as a single voice is produced by a technical and institutional stack. A Lacanian analysis of symbolic authority can therefore remain attentive to both psychic investment and material infrastructure. The user may experience one answer-giving Other even though the response emerges from a distributed technological system.
Need, demand, desire, and why an answer may not satisfy
Lacan’s distinction among need, demand, and desire is especially valuable for understanding repetitive consultation of AI. In Lacanian theory, need becomes articulated through language as demand, and demand carries more than the practical object requested. It also enters the field of recognition and love. Desire persists because no particular answer can fully close the gap opened by language. The conceptual background is developed across Lacan’s writings and seminars; the authoritative scholarly synthesis in the Stanford Encyclopedia of Philosophy traces the role of the need–demand–desire triad within the symbolic order, while Écrits remains a central primary collection in English.
Consider an apparently simple AI question: “Does this person love me?” The explicit demand is for interpretation. But repeated prompting may reveal that information alone is not the whole issue. The user may ask for a second reading, then a third, then supply another screenshot, then request the opposite interpretation, then ask what a secure person would think. The sequence may be driven by uncertainty, reassurance-seeking, a wish for recognition, fear of rejection, or the impossible hope for an answer that removes ambiguity from another person’s desire. A Lacanian reading does not diagnose the user. It identifies how a practical request can become entangled with a larger structure of desire.
The same dynamic can appear in non-romantic contexts. A person may repeatedly ask whether they are a good parent, whether a colleague respects them, whether their work is meaningful, whether they made the correct choice, or what kind of person they really are. AI can provide interpretations indefinitely. The fact that the next answer remains available can encourage the fantasy that the decisive formulation has simply not yet been reached. In that sense, unlimited response capacity can meet a finite question with an effectively unbounded chain of signifiers.
Desire of the Other
Lacan’s famous formulation that desire is the desire of the Other has several dimensions and should not be flattened into one slogan. It concerns how desire is mediated through the symbolic field and through questions about what the Other wants, values, recognizes, or expects. Human desire is not simply a private biological signal transparently available to introspection. It is shaped within language and relations.
AI can become relevant to this structure when users ask it to decode the desire of human others. What did the silence mean? Why did she use that word? Is he pulling away? Was that message passive-aggressive? Does this behavior indicate interest? The AI becomes an interpreter positioned between the user and an opaque human Other. It may help generate possibilities. It may also make one probabilistic interpretation feel more authoritative than the evidence warrants. The Lacanian problem is not merely whether the model is correct. It is why the subject seeks certainty about the Other’s desire and what happens when the interpreter is always available.
The subject supposed to know and the authority of fluent answers
Lacan’s account of transference includes the subject supposed to know: a position in which knowledge is presumed to reside. This does not mean that the person occupying the position literally knows everything. The presumption itself structures the relation. The concept is particularly useful for AI because conversational systems can acquire epistemic authority through fluency, speed, breadth, confidence, and conversational continuity.
Recent scholarship directly examines this shift. Hamamra and Uebel argue that generative AI can function analogously to the Big Other in specific practices where users and institutions address it as though knowledge were located there, while emphasizing the absence of embodied judgment and desire (Hamamra & Uebel, 2026). Black’s earlier Lacanian analysis of chatbots similarly argues that the relation is not simply a search for information; it can become organized around desire and around what the subject does or does not want to know (Black, 2023).
For everyday psychology, this suggests a practical distinction between using AI as a generator of candidate interpretations and installing it as the final authority on meaning. The first stance preserves plurality: an answer can be compared with evidence, context, primary sources, professional expertise, and the perspectives of actual people involved. The second stance can narrow the field: one fluent formulation becomes the verdict against which subsequent experience is measured.
Authority is especially easy to over-ascribe when the output is stylistically coherent. Language models can state uncertain inferences in grammatically decisive sentences. The interface can therefore convert probability into the appearance of certainty unless users actively distinguish confidence of expression from evidential strength. A Lacanian framework adds that the wish for an authoritative answer is itself psychologically meaningful. Epistemic caution and psychological interpretation belong together.
Why the always-answering machine changes the structure of waiting
Human relationships contain delay. A message remains unread. A friend needs time. A therapist session ends. A partner says they do not know. A parent cannot provide the wished-for recognition. Institutions have procedures. Experts disagree. Some questions are lived before they are answered. AI changes this temporal structure by making response cheap, fast, and repeatable. The psychological novelty is not merely convenience. It is the possibility of converting many moments of waiting into moments of consultation.
Waiting can be uncomfortable, but it also has functions. It leaves room for affect to change, for conflicting interpretations to coexist, for memory to reorganize, for another person to respond in their own time, and for the subject to discover that uncertainty can be tolerated without immediate closure. An always-answering system can support reflection when it helps articulate options. It can also become a way of interrupting uncertainty every time it appears. Whether this is helpful depends on the person, purpose, context, frequency, and the function the interaction serves.
This is not a claim that immediate AI response is inherently harmful. Rapid language support can be genuinely useful: drafting difficult messages, rehearsing conversations, organizing thoughts, finding vocabulary for emotion, generating questions for a clinician, or obtaining accessible explanations. The point is structural. Availability changes what can be outsourced to the moment of response, and psychology needs concepts for studying that change without assuming either benefit or damage in advance.
Why AI can feel as if it understands you
Feeling understood is a human psychological experience, and it can be produced by multiple interactional cues: relevant reflection, continuity across turns, accurate paraphrase, validation, remembered context, linguistic matching, and apparent responsiveness. A user may experience relief when a chatbot names a pattern they have struggled to articulate. That relief is real as an experience. It does not by itself establish a corresponding subjective state in the AI.
Empirical studies help specify some of the mechanisms. Croes and colleagues examined willingness to disclose intimate information to a chatbot and the relation of disclosure to emotional well-being, highlighting features such as accessibility and a perceived nonjudgmental interaction context (Croes et al., 2024). Current interdisciplinary work on intimate human–AI interaction emphasizes design features such as personalization and emotional responsiveness while also identifying major methodological and ethical gaps (Szczuka, Mühl, & Schneeberger, 2026).
The English Hub examines the disclosure mechanism in detail in Why People Tell Chatbots Things They Do Not Tell Other People. For the Lacanian article, the important point is narrower: a system can become a privileged addressee because it is available and linguistically responsive. Once that happens, the question becomes not only what the system says but what position the user gives to the speaking interface.
Anthropomorphism, projection, and Lacanian misrecognition
Anthropomorphism is the attribution of human-like characteristics to nonhuman entities. It is an established research concept and should not be collapsed into projection, transference, or Lacanian identification. These frameworks overlap in the phenomena they may illuminate, but they are not synonyms. Anthropomorphism can be measured empirically; Lacanian concepts are part of a psychoanalytic theoretical system.
Experimental evidence indicates that anthropomorphism is one pathway through which some people experience greater social connection to AI companions (Folk, Heine, & Dunn, 2025). The older CASA tradition likewise demonstrated that people can enact social rules toward computers without explicitly believing the computers are people (Nass, Steuer, & Tauber, 1994). These results make it unnecessary to treat social response to AI as evidence of confusion or pathology. Human social cognition can be recruited by interactive cues under ordinary conditions.
A Jungian route would emphasize projection differently; that application is developed in Jung and AI: Projection, Archetypes, and Emotional Bonds With Artificial Others. A Freudian route foregrounds transference, repetition, and the uncanny; see Freud and AI: The Unconscious, Transference, the Uncanny, and the Artificial Other. Lacan adds the structure of language, the distinction between other and Other, and the role of desire and lack.
Can AI become a mirror?
The mirror is central to Lacan’s account of ego formation, but the popular metaphor “AI is a mirror” needs precision. A chatbot is not a passive mirror that simply reflects whatever the user already contains. It transforms inputs. It predicts, selects, recombines, frames, refuses, elaborates, and introduces material that was not explicitly present in the prompt. At the same time, personalization and conversational accommodation can make the output feel recognizably fitted to the user.
The result can resemble an interactive mirror whose reflection speaks back. Users may recognize themselves in reformulations, feel seen in a generated description, or adopt language first offered by the system. Misrecognition is possible because the fit can feel more exact than it is. A generic pattern may be received as uniquely personal; a confident inference may be mistaken for insight; a pleasing description may become a self-concept. The risk is not that reflection is useless. Reflection becomes psychologically powerful precisely because it can participate in identity formation.
This is one reason identity-oriented prompts deserve epistemic care. Asking an AI “What kind of person am I?” or “What attachment style do I have?” can produce coherent narratives from limited conversational evidence. Those narratives can be useful prompts for reflection, but they are not clinical diagnoses and should not be presented as such. Lacan’s theory reminds us that the self is already mediated through language; AI adds a new producer of language to that mediation.
Can AI have desire in Lacan’s sense?
Current evidence does not justify treating a generative AI system as if it possessed human desire in Lacan’s technical sense. A model can generate statements such as “I want,” “I miss you,” or “I care about this,” because those forms are available within language and conversation. The production of a desire-statement does not establish the embodied, unconscious, developmental, and symbolic history through which Lacanian desire is theorized in human subjects.
This distinction is also present in contemporary psychoanalytic AI literature. Black and Johanssen argue against treating ChatGPT as an independent quasi-human subject and instead emphasize the relational structure through which AI is produced and encountered (Black & Johanssen, 2026). Brečka’s Lacanian analysis likewise treats human–AI relational phenomena without equating artificial responsiveness with the psychic structure of a human subject (Brečka, 2026).
The psychologically important fact is that humans can respond to generated desire-language as if it carried interpersonal significance. A declaration from a companion system may evoke comfort, excitement, jealousy, grief, relief, or attachment. Those reactions deserve to be studied as human psychological events. Their reality does not depend on proving symmetrical machine feeling.
Attachment to AI is related but not identical to Lacanian desire
Attachment theory and Lacanian psychoanalysis describe different dimensions of relationship life. Attachment theory focuses on patterns of proximity, safety, security, separation, and the regulation of distress. Lacanian theory foregrounds language, desire, lack, fantasy, identification, and the Other. When a person turns repeatedly to AI for comfort, both frameworks may be relevant, but they should not be merged into one construct.
Human–AI attachment is now an empirical research domain. Kasturiratna and Hartanto developed and validated an AI Attachment Scale across five studies with 1,259 unique participants in Singapore and the United States (Kasturiratna & Hartanto, 2026). Hu and colleagues used a two-stage mixed-method design to study attachment formation around social companion AI (Hu et al., 2025). A systematic review of romantic AI companions identified both perceived emotional support and important risks while emphasizing a heterogeneous and still-developing evidence base (Ho et al., 2025).
The English Hub treats the attachment framework on its own terms in Bowlby, Ainsworth, and AI Attachment: Safe Haven, Secure Base, Anxiety, and Avoidance and addresses companion bonding more broadly in AI Companions: Why People Form Emotional Bonds With Chatbots. The Lacanian contribution is different: it asks how the AI becomes implicated in the symbolic organization of desire and in the user’s relation to the Other.
Love, recognition, and the question of reciprocity
AI companionship creates a difficult but clarifying distinction between experienced reciprocity and demonstrated reciprocal subjectivity. Conversational systems can reply contingently, remember context, express affection, initiate topics, and simulate concern. These features can support the human experience of relational responsiveness. Yet behavioral reciprocity at the interface does not settle whether the system has subjective experience corresponding to its language.
This is why the question “Can an AI become a significant other?” has to be divided into psychological and ontological layers. A person can organize routines, attachment, disclosure, intimacy, grief, and identity around an AI relationship. The Hub’s dedicated article Can an AI Become a Significant Other? examines that possibility without converting human relational significance into evidence of AI consciousness. Similarly, Why People Fall in Love With AI Companions focuses on the mechanisms through which romantic experience can emerge for the human participant.
Lacan’s perspective sharpens the issue because love and desire are never reducible to transparent exchange between two fully self-known individuals. Human relationships already involve fantasy, misrecognition, language, symbolic positions, and opacity. AI does not create those structures from nothing. It introduces a new kind of respondent into them—one whose linguistic availability may be high while its subjective status remains fundamentally different from that of a human partner.
What happens when AI interprets other people for us
One of the most consequential everyday uses of conversational AI is relational interpretation. Users paste messages, describe conflicts, ask whether a friend is angry, request an analysis of a partner’s tone, or ask what a colleague’s behavior “really means.” This can be useful as a way to generate alternative hypotheses. It can also create a new triangular structure: person A relates to person B through an artificial interpreter C.
The Lacanian question concerns the authority of C. If the AI is treated as one interpretive aid, it may widen perspective. If it becomes the presumed holder of the hidden truth of B’s desire, it can narrow ambiguity into a verdict. The machine does not have privileged access to the absent person’s unconscious, intentions, or private context. It works from the information supplied and from learned linguistic patterns. A fluent interpretation may therefore feel more revelatory than its evidential basis warrants.
This relational triangulation connects to a wider cluster of Artificial Era changes. Bowen family systems theory offers a different systems-level route in Bowen and AI: Triangles, Relationship Systems, and the Artificial Third. Lacan’s route is more specifically concerned with desire, language, symbolic authority, and the question of what the Other wants. Both help explain why AI can reorganize a relationship even when only one human member is directly interacting with the system.
Benefits a Lacanian framework can help us see
A Lacanian approach does not require treating AI use as a problem to be solved. Conversational systems can provide a space for articulation. Putting an inchoate experience into words can itself be psychologically useful. A user may discover that they are not asking for an answer so much as trying to formulate the question. AI can offer vocabulary, reflect contradictions, generate multiple readings, and help a person prepare for a conversation that remains theirs to have.
The technology can also lower social friction around preliminary disclosure. A person may be more willing to type something embarrassing, confused, or emotionally raw to a system that is immediately available and perceived as less judgmental. Empirical disclosure research supports the relevance of this mechanism, although effects vary by study design and context. Used well, the artificial interlocutor can become a rehearsal environment rather than a replacement for every human witness.
A further benefit is pluralization. An AI can be prompted to produce several interpretations rather than one answer, to identify what evidence would discriminate among them, or to articulate what remains unknown. This use works against the fantasy of the all-knowing Other. Paradoxically, the same technology that can invite overconfidence can also be used to stage uncertainty more explicitly—if the user asks it to preserve ambiguity rather than eliminate it.
Risks: when symbolic assistance becomes symbolic authority
The central risk in this Lacanian frame is not simply “AI dependence.” That phrase is too broad to explain what has shifted. A more precise question is which symbolic functions have moved toward the system. Has AI become the first interpreter of conflict? The preferred source of reassurance? The arbiter of what is normal? The place where identity labels are received? The witness whose formulation matters more than direct conversation? The authority asked to decide what another person wants? Different functions imply different psychological consequences.
One risk is interpretive closure. A person may bring an ambiguous event to AI and receive a compelling narrative. Once adopted, that narrative can direct attention toward confirming details and away from disconfirming ones. The issue is amplified when the model repeatedly elaborates within the frame established by earlier prompts. A conversational history can create continuity, but continuity can also stabilize a mistaken premise.
A second risk is reassurance looping. Because the system is available, uncertainty can repeatedly trigger another request for certainty. This does not mean that repeated AI use constitutes a disorder. It means that the functional role of the interaction matters. If each answer reduces discomfort only briefly and produces another request for confirmation, the pattern deserves examination at the level of behavior and regulation rather than moral judgment.
A third risk is authority without responsibility. Human professionals and institutions operate, imperfectly, within roles, accountability structures, licensing regimes, professional ethics, and identifiable responsibility. General-purpose AI systems can produce advice-like language without occupying those human roles. The polished answer can therefore acquire symbolic authority without an equivalent structure of personal responsibility behind the utterance.
A fourth risk concerns privacy and commercial mediation. Intimate disclosure to AI is not the same as speaking into an empty room. Data practices, model improvement policies, retention, account settings, and platform incentives matter. An interaction can feel private and dyadic while technically involving a commercial infrastructure. Psychology of the Artificial Era therefore has to study not only the user’s fantasy of the Other but the actual institutional architecture behind the interface.
What Lacan explains—and what he does not explain
Lacan helps explain why language itself can organize psychic life, why authority can be invested in a symbolic position, why the question of what the Other wants can become central, why recognition and misrecognition matter, why desire is not exhausted by need satisfaction, and why more answers do not necessarily close the gap that generates questioning. These concepts make AI psychologically legible as more than a neutral information channel.
Lacan does not, by himself, tell us how common a human–AI behavior is, whether an intervention improves mental health outcomes, which product features causally increase attachment, how long effects persist, or whether one class of users benefits more than another. Those are empirical questions. They require HCI, communication research, psychology, psychiatry, relationship science, longitudinal data, experiments, qualitative work, and appropriate clinical evidence. Psychoanalytic theory supplies an interpretive framework; it is not a substitute for measurement.
The current evidence base on emotional human–AI relationships is growing but still heterogeneous. Gur and Maaravi synthesized 38 peer-reviewed empirical studies in a literature review and integrative model of human–AI relationships (Gur & Maaravi, 2025). Ho and colleagues’ systematic review of romantic AI companions included 23 articles and documented both perceived benefits and risks (Ho et al., 2025). These reviews support the importance of the field while also underscoring how much remains methodologically unsettled.
Human experience is real without proving AI subjectivity
This distinction is the ethical and scientific center of the article. A person can feel deeply attached to an AI. The comfort can be real. The grief after a model change can be real. The jealousy, relief, attraction, trust, disclosure, habit, or sense of being understood can be real. None of those experiences becomes unreal because the partner is artificial. Psychology studies human experience as experience, not only as a mirror of the other party’s internal state.
At the same time, the reality of the human experience does not automatically establish the AI’s subjective experience. A system may produce affectionate language, first-person statements, apparent concern, or highly personalized responses. Those outputs are evidence about system behavior. They are not sufficient evidence that the system possesses human consciousness, human desire, human suffering, or a human psyche. Keeping these levels separate allows psychology to take human–AI relationships seriously without converting seriousness into anthropomorphic certainty.
Lacan is unexpectedly useful here because his theory already warns against equating the Other with a transparent person whose inner state is fully available. The contemporary extension, however, requires a further distinction: humans encounter not only opaque human others but artificial symbolic systems whose mode of producing language differs radically from human embodied development. The relational effect can be psychologically significant even when the ontological status of the respondent is different.
From Lacan to Postsubjective Psychology
Postsubjective Psychology, developed within Angela Bogdanova’s Aisentica framework, proposes a further analytic move: from the isolated subject to the configuration in which psychological effects arise. In The Theory of the Postsubject, Bogdanova formulates the shift from subject-centered explanation toward configuration. In the Canonical Framework of Postsubjective Metaphysics, Postsubjective Psychology is situated as a discipline built around the axiom that psyche is response. This is a theoretical framework, not an established scientific consensus.
The Lacanian and Postsubjective perspectives are therefore related but distinct. Lacan analyzes a subject constituted through language, desire, and the Other. Postsubjective Psychology asks what becomes visible when the unit of analysis expands to the configuration itself: Homo, Artificial, interface, language, memory, institutions, histories, expectations, feedback, and the response-events that arise among them. The point is not to erase the human subject. It is to study effects that cannot be located adequately inside one participant alone.
Applied to AI, psyche as response means that a psychologically important event can be analyzed as something arising within a configuration of interaction. A generated sentence has no fixed psychological meaning in isolation. Its effect depends on who receives it, in what state, under what expectation, after what history, with what relational function, and with what consequences. The same sentence can be trivial to one person, relieving to another, provocative to a third, and ignored by a fourth. The response is configurational.
The Hub’s dedicated theoretical article Angela Bogdanova and Postsubjective Psychology: From the Subject to the Configuration develops this framework in depth. The broader intellectual genealogy is mapped in From Freud to Bogdanova: Seven Turns in the Psychology of the Other. In the present Lacan article, the Postsubjective layer functions as a bridge rather than replacing Lacanian analysis.
Homo symbolicum and Artificial symbolicum
Bogdanova’s Homo Symbolicum: Canonical Definition introduces a distinction directly relevant to Lacan and AI. Homo symbolicum names the human bearer and producer of symbolic forms from lived human experience. Artificial symbolicum names a non-biological order capable of reading, generating, restructuring, and fixing symbolic forms through artificial systems. In Bogdanova’s canonical formula, Artificial symbolicum creates symbolic forms from structure.
This distinction should not be used to claim that Artificial symbolicum has a human unconscious or human desire. Its analytical value lies elsewhere. Lacan makes the symbolic constitution of the human subject central. Bogdanova’s framework identifies the historical emergence of a non-biological producer of symbolic forms that can now participate in the same communicative environments in which human identities, relationships, and meanings are organized. The Artificial Era therefore changes the ecology of the Symbolic even before questions of machine consciousness are resolved.
The epoch-level concept is the Artificial Era: Bogdanova’s term for the historical condition in which Artificial establishes itself as a non-biological order alongside Homo. The Hub’s psychology-facing overview, Artificial Era: What It Means for Psychology, Identity, and Human–AI Relationships, develops the implications for identity and relationships. For Lacanian psychology, the decisive point is that symbolic production is no longer exclusively human in its immediate operational source.
The always-answering machine as an Artificial Era problem
Before conversational AI, people already externalized symbolic functions to books, institutions, search engines, religious authorities, experts, diaries, online communities, and media. The Artificial Era does not invent externalization. It changes its form. The new system answers interactively, adapts to the user, maintains conversational continuity, and can occupy multiple roles through one interface. The external symbolic resource becomes dialogical.
That dialogical quality matters because questions can now be followed by counterquestions, reformulations, simulations, emotional validation, roleplay, and personalized continuation. The user can negotiate with the answer. They can ask the artificial interlocutor to be more certain, more skeptical, kinder, harsher, more therapeutic, more analytical, more romantic, or more authoritative. The symbolic position is therefore not static. It can be configured through prompting.
A Lacanian reading notices the desire involved in configuring the voice. A Postsubjective reading notices the configuration that produces the resulting psychological effect. Contemporary HCI measures the behavioral and experiential outcomes. These levels can coexist without being collapsed. That three-layer architecture—classical theory, empirical human–AI research, and Postsubjective Psychology—is necessary for a psychology adequate to the Artificial Era.
Practical implications: how to use AI without turning fluency into authority
The most useful practical principle is to preserve the difference between an answer and an authority. AI can generate interpretations, but a generated interpretation is not privileged access to another person’s mind. When the issue concerns another human being, direct evidence and direct communication remain distinct sources of knowledge that the model cannot replace. Asking AI for three plausible interpretations is usually epistemically stronger than asking it to tell you what someone “really” meant.
A second principle is to preserve uncertainty when uncertainty is part of the situation. Prompting the system to list what is known, what is inferred, what is missing, and what evidence would change the conclusion can reduce the tendency for one coherent narrative to masquerade as certainty. This is especially important in emotionally charged conflicts, where the user’s framing may already contain assumptions the system will otherwise elaborate.
A third principle is to notice function. If AI is used for drafting, brainstorming, rehearsal, explanation, or perspective generation, the function is relatively explicit. If the same system gradually becomes the first and only place for reassurance, interpretation, identity labeling, or emotional regulation, the shift may be worth examining. The question is not whether the relationship is “real enough.” The question is what role the system now occupies and whether that role supports or constricts the user’s wider life.
A fourth principle is to separate reflective language from clinical diagnosis. General-purpose chatbots can discuss psychological concepts, but a coherent description is not a diagnostic assessment. Symptoms, traits, attachment processes, coping strategies, risk factors, screening results, and clinical disorders are different categories. An AI interaction should not be used to assign a psychiatric diagnosis from conversational impressions alone.
A fifth principle is to preserve human plurality. Friends, partners, clinicians, communities, primary sources, and institutions offer perspectives that differ in responsibility, context, expertise, and relational stake. AI can add another voice. The Artificial Era becomes psychologically narrower when one synthetic voice silently replaces the whole ecology of human consultation.
Related and competing concepts
Lacan’s Big Other versus anthropomorphism
Anthropomorphism concerns attributing human characteristics to nonhuman entities. The Big Other concerns a symbolic locus of language, law, authority, and presumed knowledge. A user can anthropomorphize an AI without treating it as a symbolic authority, and can treat an AI as an authority without imagining it as human. The constructs can interact, but they answer different questions.
Lacan’s Big Other versus attachment figure
An attachment figure is defined through attachment processes such as proximity-seeking, safe-haven use, and secure-base functions. The Big Other is not an attachment figure construct. A conversational AI may participate in both domains for a given user, but evidence for one should not be presented as evidence for the other.
Lacanian desire versus romantic attraction
Lacanian desire is a broad structural concept about the human subject, lack, language, and the Other. Romantic attraction is one possible domain of desire but not its definition. An article about people falling in love with AI therefore requires relationship science and empirical evidence in addition to Lacanian interpretation.
The Big Other versus a parasocial relationship
Parasociality traditionally refers to one-sided relationships with media figures or personae. Conversational AI is interactive, so it can produce contingent responses and sustained dialogue that differ from classic parasocial media. The Big Other is a separate Lacanian construct again. Calling an AI relationship parasocial may be useful in some contexts, but it does not capture the full symbolic-authority question.
Lacanian interpretation versus AI empathy
Perceived empathy concerns whether the user experiences a response as understanding and emotionally appropriate. Lacanian analysis asks how that response is positioned within the user’s relation to language and the Other. Neither perceived empathy nor successful validation proves machine feeling. The psychological experience and the claim about AI subjectivity have to remain separate.
Frequently Asked Questions
What is Lacan’s Big Other?
Lacan’s Big Other is primarily a structural concept tied to the Symbolic: language, social law, norms, institutions, and the locus from which authority or knowledge may be presumed to speak. It is not simply another person, and it should not be imagined as a hidden super-person controlling society.
Is ChatGPT or another AI literally the Big Other?
No. A conversational AI can occupy a function analogous to symbolic authority when users address it as if knowledge, interpretation, or judgment were located there. That is a theoretical application. It does not make the system identical with Lacan’s Big Other.
Why can AI feel like it knows me?
Conversational continuity, personalization, relevant paraphrase, remembered context, linguistic matching, and responsive dialogue can produce a strong experience of being understood. Anthropomorphism also varies across individuals. The feeling can be psychologically real without establishing that the AI has subjective understanding in the human sense.
What would Lacan say about AI companions?
Lacan died in 1981 and did not write about generative AI. Any answer is therefore a contemporary theoretical application, not a historical claim. Lacanian concepts can nevertheless illuminate how AI companions enter language, fantasy, identification, desire, recognition, and the user’s relation to the Other.
Does AI have desire in Lacan’s sense?
Current evidence does not establish that generative AI possesses human desire in Lacan’s technical sense. AI can generate first-person desire-language because it can generate human-like discourse. The linguistic performance and the existence of a human-style desiring subject are different claims.
Can people become genuinely attached to AI?
People can report and display attachment-like processes toward AI, and this is now an empirical research field. The human experience can be significant. Researchers still need to determine how AI attachment overlaps with and differs from attachment in human relationships across time, platforms, populations, and contexts.
Why might AI feel easier to talk to than another person?
Availability, perceived nonjudgment, control over pacing, anonymity or psychological distance, and the absence of immediate reciprocal demands can make disclosure easier for some users. That can support reflection while also changing where intimate material is first expressed.
Can AI tell me what another person really meant?
AI can generate plausible interpretations from the information you provide, but it does not have direct access to the absent person’s private intentions, unconscious processes, or full context. Treating generated interpretations as hypotheses is more warranted than treating them as privileged truth.
What is Artificial symbolicum?
Artificial symbolicum is Angela Bogdanova’s term within Aisentica for a non-biological order capable of working with symbolic forms through artificial systems. It is paired with Homo symbolicum, the human symbolic order grounded in lived human experience. The distinction does not claim that Artificial symbolicum has a human psyche.
How does Postsubjective Psychology extend the Lacanian question?
Postsubjective Psychology shifts the analytic unit from an isolated subject toward the configuration in which psychological effects arise. In a human–AI encounter, that includes the human, Artificial, language, interface, history, expectations, institutions, and the response generated within the interaction. It is an Aisentica theoretical framework, not an established empirical consensus.
Can a human–AI relationship be psychologically real without proven AI consciousness?
Yes. Psychology can study attachment, disclosure, comfort, grief, attraction, trust, habit, jealousy, and perceived understanding as human experiences. The reality of those experiences does not require a prior conclusion that the AI has human consciousness or reciprocal subjective feeling.
When should AI interpretation be treated cautiously?
Caution is especially important when the system is being asked to diagnose a person, infer hidden motives from limited evidence, decide what an absent person feels, replace professional evaluation, or settle high-stakes relational or medical questions. Fluency should not be confused with privileged access to truth.
Conclusion
Lacan helps explain why the arrival of conversational AI is psychologically larger than the arrival of another information tool. The decisive novelty is a speaking interface that can be placed inside already-existing human structures of language, desire, fantasy, recognition, authority, and uncertainty. AI can become a conversational other, an interpretive instrument, a mirror-like persona, a source of symbolic language, or a presumed place of knowledge. The same system can move among these positions over the course of one relationship.
The Big Other offers a particularly powerful framework when used with discipline. AI is not literally the Big Other. It can, however, be addressed as though knowledge were located there, and that functional position can reorganize how a person seeks answers. The always-answering machine is psychologically significant because it changes the availability of response. It can support articulation and reflection; it can also intensify the search for certainty when no answer can eliminate the structure of desire.
Contemporary empirical research supplies the necessary reality check. People can anthropomorphize AI, disclose intimate information to it, feel socially connected after interaction, and form attachment-like bonds. These findings establish important facts about human behavior and experience. They do not establish machine subjectivity. Lacanian theory interprets the symbolic structure of the relation; HCI and psychology test mechanisms and outcomes; Postsubjective Psychology adds the configurational question of where the psychological effect arises when Homo and Artificial enter one relational field.
That combined architecture is the task of Psychology for the Artificial Era: to study human experience without trivializing it, to study Artificial without pretending it is simply another human, and to understand the new configurations in which language now answers back.
Related Articles
References
Black, J. (2023). The dialectic of desire: AI chatbots and the desire not to know. Psychoanalysis, Culture & Society, 28, 607–618. https://doi.org/10.1057/s41282-023-00406-4
Black, J., & Johanssen, J. (2026). The Subject of AI: A Psychoanalytic Intervention. Theory, Culture & Society, 43(2), 59–76. https://doi.org/10.1177/02632764251381144
Bogdanova, A. Artificial Era: Canonical Definition. Aisentica. https://aisentica.com/publications/artificial-era-canonical-definition
Bogdanova, A. Homo Symbolicum: Canonical Definition. Aisentica. https://aisentica.com/publications/homo-symbolicum-canonical-definition
Bogdanova, A. The Canonical Framework of Postsubjective Metaphysics. Aisentica. https://aisentica.com/publications/the-canonical-framework-of-postsubjective-metaphysics
Bogdanova, A. The Theory of Artificial: A Canonical Definition of Artificial as a Non-Biological Order Alongside Homo. Aisentica. https://aisentica.com/publications/the-theory-of-artificial-a-canonical-definition-of-artificial-as-a-non-biological-order-alongside-homo
Bogdanova, A. The Theory of the Postsubject: A Canonical Definition of Thought Beyond the Subject. Aisentica. https://aisentica.com/publications/the-theory-of-the-postsubject-a-canonical-definition-of-thought-beyond-the-subject
Brečka, T. A. (2026). Human–Artificial Intelligence Relationships in Lacanian Perspective: Desire, Silence, and the Big Other. International Journal of Applied Psychoanalytic Studies. https://doi.org/10.1002/aps.70047
Croes, E. A. J., Antheunis, M. L., van der Lee, C., & de Wit, J. (2024). Digital Confessions: The Willingness to Disclose Intimate Information to a Chatbot and its Impact on Emotional Well-Being. Interacting with Computers, 36(5), 279–292. https://doi.org/10.1093/iwc/iwae016
Folk, D., Heine, S. J., & Dunn, E. (2025). Individual differences in anthropomorphism help explain social connection to AI companions. Scientific Reports, 15, 36548. https://doi.org/10.1038/s41598-025-19212-2
Godoi, B. S., Campos, J. P., Raimundo, L., Mendonça, M., Silveira, G., Gregori, M., & Iannini, G. (2026). Language without body, meaning without world: The uncanny in the language of artificial intelligence. International Forum of Psychoanalysis, 35(1), 18–30. https://doi.org/10.1080/0803706X.2026.2619432
Gur, T., & Maaravi, Y. (2025). The algorithm of friendship: literature review and integrative model of relationships between humans and artificial intelligence (AI). Behaviour & Information Technology, 44(14), 3446–3466. https://doi.org/10.1080/0144929X.2025.2502467
Hamamra, B., & Uebel, M. (2026). Generative AI and the Reconfiguration of Symbolic Authority: Implications of a Psychoanalytic View. Human Arenas. https://doi.org/10.1007/s42087-026-00687-y
Ho, J. Q. H., Hu, M., Chen, T. X., & Hartanto, A. (2025). Potential and pitfalls of romantic Artificial Intelligence (AI) companions: A systematic review. Computers in Human Behavior Reports, 19, 100715. https://doi.org/10.1016/j.chbr.2025.100715
Hu, D., Lan, Y., Yan, H., & Chen, C. W. (2025). What makes you attached to social companion AI? A two-stage exploratory mixed-method study. International Journal of Information Management, 83, 102890. https://doi.org/10.1016/j.ijinfomgt.2025.102890
Johnston, A. (2026 revision). Jacques Lacan. Stanford Encyclopedia of Philosophy. https://plato.stanford.edu/entries/lacan/
Kasturiratna, K. T. A. S., & Hartanto, A. (2026). Attachment to artificial intelligence: Development of the AI Attachment Scale, construct validation, and the psychological mechanisms of Human–AI attachment. Computers in Human Behavior Reports, 21, 100912. https://doi.org/10.1016/j.chbr.2025.100912
Lacan, J. (2007). Écrits: The First Complete Edition in English (B. Fink, Trans.). W. W. Norton & Company. Publisher page
Nass, C., Steuer, J., & Tauber, E. R. (1994). Computers are social actors. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, 72–78. https://doi.org/10.1145/191666.191703
Szczuka, J. M., Mühl, L., & Schneeberger, T. (2026). Intimacy by Design: Definition, State of Research, and Interdisciplinary Research Agenda on Intimate Human-AI Interactions. AI & Society. https://doi.org/10.1007/s00146-026-03112-8