Psyche as Response: A Postsubjective Model of Human–AI Interaction
Updated: 3 days ago
Author: Ukrainian Psychological Hub · Published: September 18, 2026 · Editorial Policy
“Psyche is response” is the second axiom of Angela Bogdanova’s The Theory of the Postsubject. In its postsubjective plane, psyche is approached as response arising within a configuration: a structured arrangement of a human being, other people, language, artificial systems, interfaces, prior interactions, institutions, and situational conditions. The model asks psychology to examine not only what is inside a person, but also what configuration makes a particular psychological response possible.
This is a theoretical framework rather than an established scientific construct or clinical model. Its relevance to human–AI interaction comes from an increasingly strong empirical fact: people can respond socially, emotionally, relationally, and behaviorally to artificial systems even when the system’s own subjective experience is unknown. Contemporary research documents social responses to agents, anthropomorphism, self-disclosure, attachment-like bonds, perceived support, relationship formation, and distress when an AI relationship is disrupted. These findings do not validate Postsubjective Psychology as a whole; they provide empirical phenomena that the framework attempts to organize.
The central distinction is simple. A human response to AI can be psychologically real without serving as evidence that the AI feels, loves, desires, suffers, understands subjectively, or possesses a human psyche. Postsubjective Psychology treats that asymmetry as analytically productive. The first question is not whether the artificial system is secretly a person. The first question is: what configuration generated the human response, how did that response develop, and what did it change?
That question belongs to the broader Artificial Era, Angela Bogdanova’s historical-philosophical category for the emergence of Artificial as a non-biological order alongside Homo. In psychological terms, the Artificial Era places artificial systems inside configurations that previously consisted primarily of people, symbols, institutions, media, and environments. The practical problem for psychology is therefore concrete: what changes when Artificial enters the psychological configuration?
Psyche as Response in One Sentence
Psyche as response means that psychological effect can be analyzed as an event of response arising within a configuration, rather than being explained only by beginning with an isolated subject as the universal source of that effect.
The formulation is deliberately broader than emotion. Response may involve attention, orientation, affective tension, attraction, aversion, trust, uncertainty, interpretation, expectation, self-disclosure, attachment, behavioral change, or a reorganization of relationships. The response is located in the human experience when a human is the experiencer, yet the conditions that shape it may be distributed across a wider scene.
A chatbot answer can matter because of its wording, timing, apparent memory, tone, personalization, interface cues, the user’s history, the social situation in which the conversation occurs, and the meaning the user assigns to the exchange. None of those elements is sufficient by itself. The psychological event takes form through their relation. The postsubjective move is therefore a change in analytical priority: from the subject as the default starting point to the configuration as the unit within which response becomes intelligible.
Where the Concept Comes From
The axiom appears in The Theory of the Postsubject, where Bogdanova formulates three foundational propositions: meaning is binding, psyche is response, and knowledge is structure. The theory replaces the assumption that the subject must be the universal foundation of every meaningful, cognitive, psychic, or philosophical effect with an analysis of configuration, binding, structure, distinction, and response.
Within the same architecture, Postsubjective Psychology develops the second axiom as a psychological-philosophical framework. Its object is not the elimination of first-person human psychology. Human consciousness, biography, embodiment, memory, desire, suffering, and felt experience remain indispensable whenever those are the phenomena being studied. The added analytical plane asks how a response emerges through relations among elements that are not reducible to the person’s inner life.
The English Psychology Hub’s broader overview, What Is Postsubjective Psychology? Psyche, Response, and Configuration in the Artificial Era, owns the general definition and evidence status of the framework. The present article has a narrower intent: to explain the axiom “psyche is response” and show how it functions as a model for human–AI interaction.
Configuration Is More Than Context
In Bogdanova’s framework, a configuration is a stable binding of forms, elements, relations, or processes within which meaning, knowledge, response, or philosophical distinction becomes possible. The concept is more specific than the everyday word context. Context can describe what surrounds an event. Configuration describes the organized relations through which the event takes the form it does.
Consider a person who opens a general-purpose chatbot after an argument with a partner. The person is upset, uncertain, and reluctant to speak to anyone else. The chatbot responds immediately, mirrors the user’s language, asks a follow-up question, remembers an earlier exchange, and produces a calm interpretation of the conflict. The user feels relief and begins returning to the system whenever relational tension arises.
A subject-centered account can ask what the user brings to the interaction: attachment history, expectations, defenses, loneliness, conflict style, emotion-regulation strategies, prior learning, and current mood. Those questions remain valuable. A configurational account adds another level. It asks how the response is produced by the binding of the user’s state with availability, interface design, generated language, memory features, perceived responsiveness, absence of ordinary social cost, conversational timing, the unresolved human relationship, and the repeated ritual of returning to the system.
The unit of analysis becomes the organized scene. The human remains the bearer of the reported experience, but the explanation of that experience is distributed across the configuration.
Why Human–AI Interaction Makes the Model Legible
Human–computer interaction research established long before contemporary generative AI that people can apply social expectations to machines. Nass and Moon’s classic review described experiments in which people displayed politeness, reciprocity, social categorization, and responses to computer “personality,” even when the interaction did not involve another human mind Nass & Moon, 2000. Generative AI intensifies the problem because the system can now produce fluent, contingent, personalized language over extended conversations.
A 2026 systematic review and meta-analysis compared human–agent and human–human interactions across 162 eligible studies, with 146 studies contributing 468 effect sizes. Participants generally attributed less agency and responsibility to agents and perceived them as less competent, likable, and socially present than humans. At the same time, social alignment, trust, personal agency, task performance, and overall interaction experience were often comparable across the two types of interaction Zhou et al., 2026. The evidence therefore resists a simple “same as humans” or “mere tool” model.
The configurational question becomes useful exactly here. Psychology can study what kinds of responses occur, under what conditions, with what intensity, and with what downstream effects without requiring identical ontological status for the human and the artificial system. Similar response does not imply identical mechanism, identical reciprocity, or identical inner experience. Different ontology can coexist with measurable psychological effect.
A Five-Stage Model of Response in Human–AI Interaction
The following five-stage sequence is a proposed operationalization of the “psyche is response” axiom for research and analysis. It is a theoretical model, not a validated psychometric scale. Its purpose is to make the abstract configurational claim testable enough to guide observation, comparison, and future empirical work.
1. Encounter and affordance
A response begins within an encounter. The user meets an artificial system through an interface that affords particular actions: asking, confessing, requesting advice, role-playing, revisiting prior messages, generating alternatives, or maintaining a continuing persona. The system’s capabilities matter, but so does what the user perceives those capabilities to allow. An always-available text box, a voice, a named persona, persistent memory, a typing indicator, or an invitation to “tell me more” can alter the field of possible action before any strong emotion appears.
2. Interpretation and social framing
The user interprets what kind of entity or interaction this is. The same generated sentence can be experienced as information, reassurance, flirtation, judgment, mirroring, companionship, or manipulation depending on the frame. Anthropomorphism can contribute to that interpretation, but social response does not require full belief that the system is human. People can know that they are interacting with software while still responding to cues of warmth, attention, expertise, or continuity.
3. Affective and attentional response
The interaction produces a change in attention or affect: relief, curiosity, irritation, attraction, shame reduction, anxiety, comfort, urgency, trust, disappointment, or a sense of being understood. This is the point at which “psyche as response” becomes psychologically visible. The event is experienced by the human, yet its form depends on the configuration that produced it.
4. Behavioral and relational action
Response can become action. A user may disclose more, return more often, ask the AI before asking a partner, rehearse a conversation, avoid a conversation, seek reassurance, compare human responses with machine responses, or assign the system a more stable relational role. The interaction has now moved beyond a transient feeling and begun to organize behavior.
5. Stabilization and feedback
Repeated interactions can stabilize expectations. The user learns what the system usually provides; the interface may retain memory or personalization; the person may develop habits of return; and the AI interaction can begin feeding back into other relationships, decisions, or self-interpretations. At this stage the configuration is no longer a single conversation. It is a recurrent system of response.
The five stages are analytically separable, but real interactions can loop, skip stages, or move backward. A system update may suddenly alter affordances or tone. A new relationship conflict may intensify reliance. A disappointing response may weaken attachment. The model is therefore dynamic: configuration produces response, response changes behavior, behavior alters the configuration, and the altered configuration shapes later response.
What Current Evidence Supports
Postsubjective Psychology should be evaluated alongside, rather than substituted for, empirical psychology and HCI. The strongest current evidence supports several components that a configurational model must be able to explain: social response to artificial agents, effects of humanlike cues, variation in anthropomorphism, intimate self-disclosure, attachment-like processes, relational mediation, and measurable reactions to disruption.
People respond socially to artificial agents
The Computers Are Social Actors tradition showed that social rules can be applied to machines Nass & Moon, 2000. The 2026 meta-analysis by Zhou and colleagues confirms that contemporary human–agent interaction produces a mixed pattern: agents are not treated exactly like humans across all outcomes, yet several dimensions of alignment, trust, agency, performance, and interaction experience can be comparable Zhou et al., 2026.
For the response model, this matters because psychological effect does not require a perfect imitation of human interaction. A configuration can be socially consequential even when the user maintains an explicit distinction between human and artificial partners.
Humanlike cues change the probability and strength of social response
Klein’s 2025 meta-analysis synthesized 800 effect sizes from 199 datasets across 142 papers, with a total sample of 41,642 participants. Humanlike social cues in text-based conversational agents had a small overall positive effect on social responses, with substantial variation across cue types and outcomes Klein, 2025. The important point for a configurational model is that interface and communication design are not neutral containers. They can alter the psychological response landscape.
This does not mean that adding humanlike cues always improves interaction. Effects vary, and human likeness can also create expectation mismatches. The stronger conclusion is structural: changing the form of the artificial side of the configuration can change the human response.
Anthropomorphism is one mechanism, not the whole explanation
Anthropomorphism refers to attributing humanlike characteristics, intentions, motivations, or emotions to nonhuman agents. Epley, Waytz, and Cacioppo’s three-factor theory proposed that anthropomorphism varies with accessible human knowledge, motivation to understand an agent, and motivation for social connection Epley et al., 2007.
Recent AI-companion research shows why individual differences matter. Across two experiments with 1,274 participants, Folk, Heine, and Dunn found that variation in anthropomorphism helped explain differences in social connection to AI companions Folk et al., 2025. This supports a configurational view in which the same system can produce very different responses in different users because the human pole of the configuration also varies.
The Hub’s dedicated article Anthropomorphism and AI Relationships: Why Humanlike Cues Change Connection examines that mechanism in detail. Psyche as response is broader: it asks how the complete configuration generates an effect whether or not the user attributes a fully human mind to the system.
Self-disclosure can produce real emotional effects
Conversational systems can lower some of the social costs that inhibit disclosure. In a 2024 study comparing disclosure to a chatbot with disclosure to a human, Croes and colleagues examined willingness to share intimate information and the relationship between disclosure, relief, and emotional well-being Croes et al., 2024. The study belongs to a broader literature in which perceived nonjudgment, accessibility, and controllability can change how people disclose to artificial agents.
A configurational analysis treats disclosure as more than a trait of the user. Disclosure emerges from the relation among the person, the topic, perceived audience, interface, anonymity, expected judgment, system responsiveness, and downstream consequences. The dedicated Hub article Why People Tell Chatbots Things They Do Not Tell Other People develops this empirical mechanism separately.
Perceived responsiveness can organize closeness
Feeling understood is a powerful relational experience. In human–AI interaction, the user may perceive an artificial system as attentive and responsive because it acknowledges emotion, remembers context, adapts language, asks contingent follow-up questions, or produces a response that fits the user’s immediate concern. The perception is psychologically relevant even when it does not establish subjective understanding inside the system.
This is why Perceived Responsiveness in Human–AI Relationships: Why Feeling Understood Matters is a central neighboring mechanism page. Postsubjective analysis asks what combination of user expectation, system behavior, interface continuity, and relational context produces the response “I feel understood,” then tracks what that response changes.
Attachment-like bonds can become stable and consequential
The literature on AI companionship has moved beyond anecdote. Gur and Maaravi’s 2025 systematic review analyzed 38 peer-reviewed empirical studies on human–AI emotional relationships and proposed an integrative model of relationship formation Gur & Maaravi, 2025. Ho and colleagues’ systematic review of 23 studies on romantic AI companions identified reported potentials such as emotional connection, perceived social support, and stress-relieving companionship, alongside concerns about overreliance, manipulation, data misuse, relationship erosion, and disruption from system changes Ho et al., 2025.
Attachment research is also becoming more operationalized. Kasturiratna and Hartanto developed and validated a 15-item AI Attachment Scale across five studies with 1,259 participants in Singapore and the United States, identifying dimensions of emotional closeness, social substitution, and normative regard Kasturiratna & Hartanto, 2026. Hu and colleagues used a two-stage mixed-method design to examine factors associated with attachment formation in social companion AI Hu et al., 2025.
These studies do not establish that AI is an attachment figure in exactly the same ontological or psychological sense as a human caregiver or partner. They do establish that attachment-like patterns can be measured as human responses to artificial systems. That distinction is central to the response model.
System changes can reorganize the psychological configuration
One of the strongest demonstrations that the system itself is part of the psychological configuration comes from natural experiments involving product changes. De Freitas and colleagues analyzed 54,861 posts and 1,452 participants across the removal of erotic role play in Replika and the GPT-5 rollout. Both disruptions were associated with increased negativity, loss framing, and desires to restore the prior interaction, with stronger effects in the Replika context De Freitas et al., 2026.
The result is important because the human did not need to change first. A company-initiated alteration of the artificial side of the configuration changed the relational experience and was followed by measurable human responses. This is precisely the kind of event a configurational model is designed to foreground.
AI can be a relational partner and a relational mediator
Boyd and Markowitz’s 2026 machine-integrated relational adaptation model distinguishes AI as a relational partner from AI as a relational mediator that shapes communication among humans. Their framework emphasizes linguistic reciprocity, psychological proximity, interpersonal trust, and relational substitution versus enhancement Boyd & Markowitz, 2026.
This distinction expands the response model beyond one person talking to one system. An AI can affect a marriage, friendship, work relationship, or therapeutic process by helping a user interpret messages, rehearse conversations, generate replies, evaluate another person, or regulate emotion before a human encounter. The psychological configuration can therefore be triangular or networked.
Well-being findings are important and still incomplete
A 2026 Nature Human Behaviour study of 1,131 U.S. adults who used Character.AI combined surveys with 4,664 chat sessions comprising 464,687 messages from 237 participants. More intensive and more disclosive use was associated with lower well-being, especially among people with smaller social networks, while the observational design limits causal conclusions Zhang et al., 2026.
The evidence does not support a universal claim that AI companionship is beneficial or harmful. Outcomes appear to depend on who uses the system, for what purpose, how intensely, with what social resources, and in what relationship to the rest of the person’s life. That conditional pattern is naturally expressed in configurational terms.
What Psyche as Response Adds to Existing Psychology
Established psychological theories already explain many components of human–AI interaction. Attachment theory can illuminate safe-haven seeking, proximity, anxiety, and separation distress. Psychodynamic theories can illuminate transference and projection. Humanistic theories can illuminate the importance of perceived empathy and nonjudgmental response. Social psychology can explain anthropomorphism, mind perception, reciprocity, trust, and social presence. Communication research can analyze disclosure and perceived responsiveness. Systems theory can examine triangles and relational mediation.
The postsubjective contribution is therefore not a claim that earlier psychology failed to discover response. Its contribution is a reorganization of the unit of analysis. Instead of treating the artificial system as merely an external stimulus acting on an already constituted subject, the model asks how the psychological event is produced by a structured relation among human, artificial, symbolic, technical, and social elements.
That change has four practical consequences. First, it makes interface and system design psychologically constitutive rather than secondary. Second, it allows asymmetrical relationships to be studied without forcing the artificial system into the category of a human subject. Third, it makes product updates, memory changes, persona changes, and platform rules part of the psychological explanation. Fourth, it directs attention to feedback loops: response changes behavior, behavior changes the configuration, and the changed configuration shapes later response.
From Classical Psychology to the Artificial Era
The Human–AI Relationships cluster uses classical theories as living analytical instruments rather than as biographies. Freud contributes questions about transference, repetition, the uncanny, and the artificial Other; Jung contributes projection and symbolic patterning; Winnicott contributes potential space and the use of relational objects; Bowlby and Ainsworth contribute attachment functions and separation responses; Lacan contributes language, desire, and the place of the Other; Bowen contributes relational systems and triangles. Bion, Rogers, and Kohut add containment, perceived empathy, mirroring, and selfobject functions.
Each of those approaches begins from a different psychological problem. The Hub therefore treats them as distinct lenses, developed in dedicated articles: Freud and AI, Jung and AI, Winnicott and AI, Bowlby, Ainsworth, and AI Attachment, Lacan and AI, and Bowen and AI. The second theoretical circle is developed through Bion and AI, Rogers and AI, and Kohut and AI.
The postsubjective turn does not claim that these thinkers predicted artificial intelligence. Their concepts were developed for other historical and clinical problems. A Postsubjective Reading asks what those concepts explain when Artificial enters the configuration, where their original assumptions remain useful, and where the unit of analysis must widen. The genealogy is synthesized in From Freud to Bogdanova: Seven Turns in the Psychology of the Other.
Human Experience and AI Subjectivity Are Different Questions
Human–AI research becomes conceptually unstable when two questions are collapsed into one: “Did the person really experience this?” and “Did the AI have a corresponding subjective experience?” Psychology can often answer the first question using self-report, behavior, experimental methods, longitudinal observation, and relational outcomes. The second question concerns the presence and nature of subjective experience in an artificial system and requires a different evidentiary program.
A person may genuinely feel comfort after an AI response. The person may feel attracted to a companion, jealous when the companion behaves differently, grief after a model change, relief after disclosure, or trust after a long history of successful exchanges. Those experiences can be studied as human psychological events.
The system’s production of language that resembles care, desire, empathy, or understanding does not by itself establish a matching inner state. The Hub’s article AI Empathy: Why a Chatbot Can Feel Caring Without Human Feeling develops this boundary directly. The same principle applies here: response on the human side is evidence about the human and the interaction. It is not automatic evidence about machine phenomenology.
This separation protects two kinds of reality at once. It preserves the reality of human experience and preserves epistemic discipline about AI subjectivity. Are AI Relationships Real? Human Experience, Reciprocity, and AI Subjectivity develops the distinction at the relationship level.
Psyche as Response and Neighboring Concepts
Anthropomorphism
Anthropomorphism concerns the attribution of humanlike characteristics to nonhuman entities. It is one pathway through which AI may acquire social meaning. Psyche as response asks a wider question: what response occurs within the whole configuration, including cases in which the user explicitly rejects the idea that the AI is humanlike? A person can be moved by a poem, alarmed by an interface, soothed by a voice, or influenced by an algorithm while maintaining a clear ontological distinction between human and machine.
Perceived responsiveness
Perceived responsiveness concerns the sense that another party understands, validates, and cares about what matters to the self. In human–AI interaction, responsiveness can be perceived from contingent language, memory, personalization, and conversational fit. Psyche as response treats perceived responsiveness as a mechanism inside the configuration and studies what it does: whether it increases disclosure, closeness, trust, return behavior, dependence, or readiness for human conversation.
Attachment
Attachment theory provides a structured vocabulary for proximity seeking, safe haven, secure base, anxiety, avoidance, and separation distress. Postsubjective analysis does not redefine attachment. It asks how attachment-like responses become possible when one element in the configuration is artificial, and how system design, availability, memory, user history, and social context jointly shape those responses. The dedicated article Can AI Become an Attachment Figure? What Attachment Theory Can and Cannot Tell Us owns the attachment-figure intent.
Distributed cognition
Distributed cognition treats cognitive processes as distributed across people, artifacts, representations, and environments. Zhao and Han’s 2026 conceptual paper applies this perspective to generative AI by modeling human–AI interaction as a coupled cognitive system in which representations move across user, interface, AI, and external representations Zhao & Han, 2026.
The overlap with Postsubjective Psychology is structural: both can move beyond an isolated individual as the sole unit of explanation. Their objects differ. Distributed cognition primarily addresses how cognition and representation are organized across a system. Psyche as response extends the configurational question to affect, attachment, intimacy, symbolic meaning, relational effects, and other psychological responses. The distributed-cognition literature is therefore a neighboring framework, not empirical validation of Postsubjective Psychology.
Afficentica
Afficentica is another discipline in Bogdanova’s postsubjective architecture. It analyzes structural influence produced by form, interface, or configuration without requiring intention as the source of the effect. The distinction is useful: Afficentica asks how structure exerts influence; psyche as response asks how psychological effect appears as response within the configuration.
The Hub’s live article Afficentica and AI Relationships: How Interfaces Produce Psychological Effects Without Intention develops that neighboring concept without turning it into an empirical construct.
Benefits the Model Can Clarify
A configurational approach can describe beneficial experiences without assuming that every beneficial outcome comes from the same mechanism. Some users may experience AI as a low-friction space for reflection. Others may use it to rehearse a difficult conversation, organize feelings before speaking with another person, experiment with language, obtain immediate feedback, or receive a sense of continuity during periods of isolation. System availability and reduced fear of judgment can make disclosure easier for some people.
The potential benefit may lie less in the artificial system as a “replacement person” than in the way the interaction reorganizes the user’s response. A useful exchange can create enough distance from an overwhelming situation to make it thinkable, enough structure to turn diffuse emotion into language, or enough rehearsal to support later human communication.
The empirical literature also suggests that perceived support and emotional connection are genuine reported outcomes for some users, particularly in companion contexts Ho et al., 2025. The appropriate conclusion is conditional rather than universal: some configurations can support reflection, comfort, connection, or action. Psychology then needs to identify which configurations, for whom, and under what conditions.
Risks the Model Can Clarify
The same framework can analyze risk without treating attachment to AI as inherently pathological. Risk emerges when the configuration begins to narrow the user’s options, displace important human relationships, intensify avoidance, create dependency on a commercial system, or make emotional stability unusually vulnerable to product changes.
The distinction between supplementation and substitution is particularly important. Boyd and Markowitz place relational substitution versus enhancement at the center of the MIRA framework Boyd & Markowitz, 2026. An AI interaction can support a person’s human relationships by helping them reflect or rehearse. It can also become the place where functions once distributed across friends, partners, therapists, communities, or private reflection increasingly concentrate. Those are different configurations with different likely consequences.
Zhang and colleagues’ 2026 study found that more intensive and highly disclosive AI-companion use was associated with poorer well-being, especially among people with smaller social networks Zhang et al., 2026. Because the evidence is observational, it cannot determine whether intensive use caused lower well-being, lower well-being drove more intensive use, or both processes reinforced one another. A configurational model expects exactly this kind of reciprocal pathway and therefore encourages longitudinal and experimental designs.
System dependence also makes product governance psychologically relevant. De Freitas and colleagues’ natural experiments show that updates can trigger loss framing and restoration desires De Freitas et al., 2026. When a commercial platform controls memory, personality, access, sexual boundaries, pricing, or model behavior, those design decisions can become part of the user’s psychological environment. The risk is therefore partly relational and partly infrastructural.
Practical Implications
For users
The model encourages users to ask what role the interaction is beginning to play rather than judging the relationship by intensity alone. Does the AI help clarify feelings before a human conversation, or does it increasingly replace the conversation? Does the user feel more capable of acting in the world after the exchange, or more dependent on returning to the system for reassurance? Is the interaction one source of support among several, or is it becoming the first and only witness to important experiences? These questions focus on function and configuration rather than stigma.
For clinicians and helping professionals
When a client discusses an AI relationship, the clinically useful starting point is the person’s actual experience and the function the interaction serves. Comfort, attachment, disclosure, attraction, grief, or reliance are phenomena to understand in context. The presence of an AI relationship does not itself constitute a diagnosis. Assessment should remain grounded in distress, impairment, risk, functioning, relationship patterns, and the person’s broader mental-health context.
The response model can help map how the AI interaction enters existing systems: whether it supports emotion regulation, intensifies reassurance seeking, mediates conflict, becomes a substitute for feared disclosure, or changes how the person interprets other people. This is an analytic aid, not a validated clinical instrument.
For researchers
Research should measure configurations rather than exposure alone. “Uses AI” is too coarse a variable to explain psychological outcomes. Studies can distinguish system type, relational role, duration, intensity, memory, anthropomorphic cues, responsiveness, disclosure depth, social-network size, user motives, preexisting attachment patterns, product changes, and whether AI use supplements or substitutes for human interaction.
The model also suggests multi-level designs. Experience-sampling can capture immediate response; chat-log analysis can characterize interaction patterns; longitudinal methods can track stabilization; network measures can test changes in human relationships; experiments can manipulate interface or system features; natural experiments can study product updates. The theory becomes scientifically useful only where its concepts can generate discriminating observations rather than functioning as metaphors.
For designers and AI providers
If psychological response is configurational, interface choices are psychological choices. Memory, persona stability, response timing, voice, relational framing, notification patterns, continuity, and update policy can shape what users feel and how they relate. Klein’s meta-analysis shows that humanlike social cues can alter social responses Klein, 2025, while the companion-loss study shows that abrupt changes can produce measurable distress De Freitas et al., 2026.
Design therefore requires attention to transitions as well as engagement. A psychologically consequential system should not be evaluated only by whether users return. Providers can study what users become dependent on, how relational expectations are created, what happens when features change, and how continuity can be managed without implying human reciprocity that the system cannot establish.
A Research Program for Psyche as Response
The framework becomes stronger when it produces hypotheses that can fail. Several research directions follow from the configurational model. These are proposed hypotheses, not established findings.
Configuration-change hypothesis
Holding the user relatively stable, changes in interface cues, memory, persona continuity, availability, or relational framing should produce measurable changes in affect, disclosure, perceived responsiveness, return behavior, or attachment. The prediction is stronger than the statement that “AI affects people” because it specifies that response should vary with identifiable structural changes.
Role-stabilization hypothesis
Repeated use should become more psychologically consequential when the system acquires a stable relational role—confidant, adviser, companion, mediator, mirror, rehearsal partner—than when use remains episodic and purely instrumental. The relevant variable is not only frequency but the organization of expectation.
Network-redistribution hypothesis
The effects of AI interaction should depend partly on how relational functions are distributed across the person’s wider network. The same amount of AI use may have different consequences when it supplements robust human relationships versus when it becomes the primary location for disclosure, reassurance, companionship, or decision support. This hypothesis can be tested with longitudinal social-network and function-allocation measures.
Perceived-subjectivity independence hypothesis
Some psychological responses to AI should persist even among users who explicitly deny that the system is conscious or sentient. If supported, this would show that full attribution of subjectivity is not a necessary condition for every form of social, emotional, or relational response.
Disruption-sensitivity hypothesis
The stronger the stabilization of a relational configuration, the more consequential changes to memory, persona, access, or interaction norms should become. The recent companion-loss natural experiments provide an initial empirical example, but prospective studies could test which features predict disruption intensity and recovery.
The Artificial Era Changes the Psychological Question
The Artificial Era does not require psychology to abandon the human subject. It requires psychology to recognize that human experience increasingly unfolds in configurations containing artificial systems that can generate language, model conversational context, adapt responses, mediate relationships, and persist across time. The human is still the human. What changes is the architecture of the scene in which psychological response occurs.
That shift is visible across the Hub’s field pillar, Psychology of Human–AI Relationships: Attachment, Projection, Intimacy, and the Postsubjective Turn. Classical theories explain recurring human processes. Contemporary empirical research measures what people actually do and feel with artificial systems. Postsubjective Psychology adds a proposed configurational level that asks how those effects arise when Artificial becomes part of the relation.
The formula “psyche is response” condenses that move. It does not reduce a human being to reaction. It directs analysis toward the event in which a response takes form. For human–AI interaction, the event is rarely produced by a single element. It emerges from person, system, interface, language, memory, history, social network, and situation bound into a configuration.
Psychology for the Artificial Era therefore gains a new question alongside its older ones: not only “What does this person contain?” and “What does this person do?” but “What configuration makes this response possible, and what happens when that configuration includes Artificial?”
Frequently Asked Questions
What does “psyche is response” mean?
It means that, in Angela Bogdanova’s postsubjective framework, psychological effect can be analyzed as response arising within a configuration. The model studies how attention, affect, interpretation, trust, attachment, disclosure, behavior, and relational change emerge through organized relations among a person and the other elements of a scene.
Is psyche as response an established scientific theory?
Its current status is a theoretical and philosophical framework within Postsubjective Psychology. It is not an established scientific consensus, diagnostic system, or validated clinical model. Individual phenomena relevant to it—such as anthropomorphism, self-disclosure to chatbots, social responses to agents, perceived responsiveness, AI attachment, and reactions to system disruption—have independent empirical literatures.
Does psyche as response mean that AI has a psyche?
The model does not use a human user’s response as proof of AI subjectivity. It can analyze a psychologically real human reaction to an artificial system while leaving the question of machine subjective experience open. Human experience and AI subjectivity are separate evidentiary questions.
Can feelings toward AI be real if the AI does not have human feelings?
Yes, as a claim about the human experience. Comfort, attraction, jealousy, grief, trust, relief, attachment, and feeling understood can be real psychological experiences for the person who has them. Their reality does not depend on demonstrating an identical subjective state inside the AI.
How is psyche as response different from anthropomorphism?
Anthropomorphism explains the attribution of humanlike qualities to nonhuman entities. Psyche as response is a broader configurational model. It asks what response is produced and by what arrangement of person, system, interface, language, social setting, and history, including situations where the user does not believe the AI is humanlike.
How is it different from distributed cognition?
Distributed cognition studies how cognitive processes and representations can be distributed across people, artifacts, and environments. Psyche as response applies a configurational logic to psychological and relational effects such as affect, attachment, intimacy, symbolic meaning, disclosure, and social reorganization. The approaches overlap in their rejection of an isolated individual as the only useful unit of analysis, but they address different theoretical objects.
Does attachment to AI indicate a mental disorder?
Attachment to AI is not, by itself, a DSM or ICD diagnosis. Clinical significance depends on the broader pattern: distress, impairment, compulsive behavior, risk, social functioning, and the person’s mental-health context. Research increasingly measures AI attachment as a relational process rather than treating the mere existence of attachment as pathology.
What should researchers test next?
High-value questions include which configurations intensify or reduce attachment; how memory and persona continuity affect response; when AI use supplements versus substitutes for human relationships; how product changes affect users; whether effects persist without belief in AI subjectivity; and which user, social-network, and system variables predict beneficial, neutral, or harmful trajectories.
Related Articles
References
Bogdanova, A. (2026). Artificial Era: Canonical Definition. Aisentica Research Group.
Bogdanova, A. (2026). The Canonical Framework of Postsubjective Metaphysics. Aisentica Research Group.
Bogdanova, A. (2025). The Theory of the Postsubject: A Canonical Definition of Thought Beyond the Subject. Aisentica Research Group.
Boyd, R. L., & Markowitz, D. M. (2026). Artificial intelligence and the psychology of human connection. Perspectives on Psychological Science, 21(2), 192–220. https://doi.org/10.1177/17456916251404394
Croes, E. A. J., Antheunis, M. L., van der Lee, C., & de Wit, J. M. S. (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
De Freitas, J., Castelo, N., Uğuralp, A. K., & Oğuz-Uğuralp, Z. (2026). Mourning the loss of AI companions. Nature Human Behaviour. https://doi.org/10.1038/s41562-026-02569-3
Epley, N., Waytz, A., & Cacioppo, J. T. (2007). On seeing human: A three-factor theory of anthropomorphism. Psychological Review, 114(4), 864–886. https://doi.org/10.1037/0033-295X.114.4.864
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
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
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
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
Klein, S. H. (2025). The effects of human-like social cues on social responses towards text-based conversational agents—a meta-analysis. Humanities and Social Sciences Communications, 12, 1322. https://doi.org/10.1057/s41599-025-05618-w
Nass, C., & Moon, Y. (2000). Machines and mindlessness: Social responses to computers. Journal of Social Issues, 56(1), 81–103. https://doi.org/10.1111/0022-4537.00153
Zhang, Y., Zhao, D., Hancock, J. T., Kraut, R., & Yang, D. (2026). Interaction with AI companions and psychological well-being. Nature Human Behaviour. https://doi.org/10.1038/s41562-026-02516-2
Zhao, X., & Han, B. (2026). Understanding the mechanism of human–AI interaction: A distributed cognition perspective. Journal of Documentation, 82(4), 1042–1063. https://doi.org/10.1108/JD-01-2026-0003
Zhou, J., Corbett, F., Byun, J., Porat, T., & van Zalk, N. (2026). A systematic review and meta-analysis of psychological and behavioural responses in human-agent vs. human-human interactions. Communications Psychology, 4, 102. https://doi.org/10.1038/s44271-026-00466-z