top of page

Psychological Encyclopedia

Relational Configuration: Why Human–AI Psychology Is More Than Two Minds

Sep 18
23 min read

Updated: 6 days ago

Author: Ukrainian Psychological Hub · Published: September 18, 2026 · Editorial Policy


A relational configuration is the psychologically consequential arrangement within which a human response takes shape when AI enters relational life. Instead of treating the event as if it were produced only inside one person, or as if it could be reduced to a conversation between two independent minds, the configuration lens examines the human participant, the AI system, the interface, the persona and response style, memory and personalization, prior exchanges, platform rules, other human relationships, social expectations, and the immediate situation as one organized field of interaction.


The term is used here as a descriptive analytic lens, not as a diagnosis, a validated psychological scale, or a claim of unique coinage. Relational psychology has long used configuration language: Stephen A. Mitchell’s relational matrix treated self, other, and the space between them as inseparable dimensions of a relational field (Mitchell, 1988). Contemporary HCI now uses closely convergent language as well. A 2026 CHI study described shifting “relational configurations” in human–AI knowledge work and argued that important relational dynamics arise at the interface rather than residing solely in the person or the system (Gülay et al., 2026).


Within Postsubjective Psychology, the configuration receives a distinct theoretical role. Angela Bogdanova’s The Theory of the Postsubject proposes a shift from the subject as the universal explanatory center toward configuration, binding, structure, and response. Its psychological formula is “psyche is response.” This is a theoretical framework, not established scientific consensus. Its value for human–AI psychology lies in asking a precise question: what changes when Artificial enters the configuration that produces a human psychological response?


What Is a Relational Configuration in Human–AI Psychology?


In human–AI psychology, a relational configuration is the organized set of relations, affordances, histories, expectations, and participants through which an AI interaction becomes psychologically meaningful. The unit of analysis is therefore wider than the user and wider than the chatbot. It includes the conditions that make a particular response possible.


Suppose the same person opens the same language model on two different evenings. In one interaction, the system is used to summarize a document. In another, it is given a persistent name, remembers earlier disclosures, responds warmly at midnight after a conflict with a partner, and becomes the first place the person goes for reassurance. The underlying model may be similar, but the psychological event is not. The relational role, timing, history, interface cues, memory, surrounding human relationships, and user expectation have changed. The configuration has changed.


This matters because contemporary evidence already shows that human responses to AI vary with design, context, relationship framing, user differences, and interaction history. A systematic review of 68 papers and 78 studies found that trust and perceived social support repeatedly mediate chatbot social influence, while empathy and responsiveness are important message-level factors (Oh et al., 2026). A separate systematic review of 38 empirical studies found that human–AI emotional relationships develop through interacting antecedents, relationship processes, and outcomes rather than through a single isolated mechanism (Gur & Maaravi, 2025).


A relational configuration is therefore not another word for “relationship.” A relationship describes a patterned connection. A configuration describes the broader arrangement that makes that connection psychologically consequential at a given time. A person can have one continuing relationship with an AI companion while passing through many configurations: playful conversation, grief support, sexual roleplay, conflict rehearsal, decision support, silence after a platform update, or mediation of a disagreement with another human.


Why the Model of “Two Minds” Is Too Narrow


Human relationships are often described as encounters between two persons, each with an inner world, motives, memories, expectations, and a capacity to affect the other. That model remains indispensable for many forms of interpersonal psychology. Human–AI interaction introduces a different architecture.


Current conversational AI can produce socially responsive language without requiring evidence that the system possesses human-like subjective experience. The human participant brings consciousness, biography, vulnerability, attachment history, bodily states, social obligations, and real-world consequences. The AI side of the interaction is produced through model architecture, training, system instructions, interface design, memory functions, personalization, moderation rules, product incentives, and the immediate conversational context. Treating these two sides as two equivalent minds conceals rather than clarifies the psychological structure.


The “two minds” model is also too narrow because AI increasingly operates as a mediator inside human relationships. Boyd and Markowitz’s machine-integrated relational adaptation model distinguishes AI as a relational partner from AI as a relational mediator and organizes relational effects through antecedents, mechanisms, moderators, and outcomes (Boyd & Markowitz, 2026). The same AI can be a direct conversational partner in one configuration and a third element shaping communication between spouses, friends, coworkers, or family members in another.


This is why the field-level article on the psychology of human–AI relationships and the present article answer different questions. The field article maps attachment, projection, intimacy, social response, and relational mediation broadly. This article asks what unit psychology should analyze when those processes are occurring together.


A Brief Genealogy of the Configuration Idea


Relational psychoanalysis


The move away from an isolated individual mind did not begin with AI. Relational psychoanalysis helped reframe psychological life as constituted through relations rather than merely expressed from a sealed interior. In his 1988 account of the relational matrix, Mitchell integrated interpersonal, object-relational, self-psychological, and existential traditions and treated relations as primary rather than secondary additions to an already complete individual (Mitchell, 1988).


Mitchell’s framework is historically important here because it prevents a false claim of novelty. “Relational configuration” already has a psychological genealogy. The Artificial Era does not create the general insight that psychological events can be relationally organized. What changes is the composition of the relational field: Artificial can now enter it as an interactive, language-producing, persistent, personalized, and sometimes socially framed element.


Computers as social actors


Research on social responses to computers provided another foundation. People often apply social scripts to interactive technologies even when they know those systems are machines. Nass and Moon described this pattern as “mindless” social response to computers: social cues can activate interpersonal expectations without a deliberate belief that the computer is literally human (Nass & Moon, 2000). The modern Computers as Social Actors problem is intensified by generative systems that can sustain open-ended conversation, adapt style, remember context, and simulate relational continuity.


Contemporary HCI


The configuration language has now appeared directly in empirical human–AI research. Gülay and colleagues found that knowledge workers oscillated among relational positions such as director, trainer, collaborator, and other roles while interacting with anthropomorphic conversational systems. Their analysis emphasized that these relational dynamics were not located exclusively in the person or the AI but emerged at the interface (Gülay et al., 2026).


This empirical use is close enough to matter and different enough to preserve conceptual clarity. The CHI paper studies relational dissonance and role oscillation in knowledge work. Postsubjective Psychology uses configuration more broadly as an analytic unit for psychological response across intimacy, attachment, meaning, support, conflict, mediation, identity, and other human–AI scenes. The overlap is evidence that a configurational vocabulary is becoming useful across disciplines; it is not evidence that all these frameworks are identical.


What Belongs to a Human–AI Relational Configuration?


A useful configuration analysis asks what elements are active, how they are related, and which of those relations are carrying psychological weight. At least seven layers commonly matter.


1. The human participant


The person brings current emotion, goals, expectations, prior relationships, attachment tendencies, loneliness or social embeddedness, cultural assumptions, AI literacy, and a history of interaction with the system. These factors change what the same response means. A warm sentence can be convenient reassurance to one person, emotionally important recognition to another, and irritating simulation to a third.


Individual differences in anthropomorphism are one example. Across two experiments involving 1,274 participants, people higher in anthropomorphic tendency reported greater social connection after chatbot interaction, helping explain why the same artificial system can feel socially meaningful to some users and emotionally inert to others (Folk et al., 2025).


2. The AI system


The model contributes generative behavior: language production, conversational adaptation, uncertainty, error, stylistic flexibility, and the capacity to respond differently as context changes. Psychology should not collapse all of this into an imagined personality. The system’s output is constrained by architecture, training, fine-tuning, system prompts, safety layers, retrieval, tools, and product-level orchestration.


3. The interface and persona


A text box, avatar, voice, typing indicator, profile name, memory display, relationship label, greeting ritual, notification, and visual style can change the meaning of interaction. These are not cosmetic details when they shape expectations about presence, warmth, continuity, availability, or intimacy.


Szczuka, Mühl, and Schneeberger describe “Intimacy by Design” as the deliberate implementation of emotional responsiveness, romantic resonance, sexual suggestiveness, persona continuity, proactive engagement, and commercial structuring into intimate AI systems (Szczuka et al., 2026). Their review makes a configuration-level point even without using Postsubjective terminology: psychological intimacy is partly organized by design affordances and platform choices.


4. Memory and interaction history


Relational meaning is temporal. A single supportive response is different from a system that appears to remember months of disclosures, recurring fears, nicknames, anniversaries, arguments, and preferred forms of comfort. Memory can create continuity, but it can also be incomplete, reset, summarized incorrectly, or altered by product changes. The resulting relationship therefore depends partly on infrastructure that the user does not fully control.


5. The surrounding human network


AI interaction rarely exists in a social vacuum. A person may turn to a chatbot because friends are unavailable, because a spouse feels unsafe to approach during conflict, because therapy is inaccessible, because social anxiety makes disclosure difficult, or because the AI is simply faster and less demanding. The effect of the AI can therefore depend on what human alternatives are present and how the AI changes investment in them.


A 2026 Nature Human Behaviour study of 1,131 Character.AI users found that the association between companionship use and well-being varied with offline social networks and with how intensively and disclosively people used the chatbot (Zhang et al., 2026). The study was observational and does not establish a simple causal story, but it strongly supports a configuration-sensitive interpretation: the same category of AI use does not have one uniform psychological meaning or outcome.


6. Platform rules and commercial architecture


The relationship is also shaped by decisions made outside the conversation: subscription tiers, memory limits, content policies, engagement optimization, notification systems, persona availability, moderation, model replacement, and service continuity. A user may experience an emotionally important bond while the conditions of that bond remain institutionally controlled. This asymmetry is psychologically relevant because a platform update can alter the interactional object to which the person has become attached.


7. The immediate situation


Time of day, distress level, recent conflict, loneliness, fatigue, intoxication, privacy, location, and urgency can all change what a response does. A phrase generated at noon during casual experimentation may have almost no emotional weight. The same phrase at 2 a.m. after a breakup can become a major regulatory event. Configuration analysis keeps the situation inside the explanation.


How Relational Configurations Produce Psychological Effects


A configuration is useful only if it helps explain mechanisms. The central claim is not that “everything affects everything.” It is that identifiable relations among person, system, interface, history, and social context can alter the probability and meaning of psychological responses.


Perceived responsiveness


One of the clearest mechanisms is perceived responsiveness: the sense that another party understands, validates, and cares about what matters to the person. In two experiments, a warm relational chatbot style increased perceived human-likeness, empathy, and closeness. Deeper conversational topics increased self-disclosure, which increased perceived responsiveness and in turn strengthened closeness (Telari et al., 2026).


This mechanism is examined in depth in Perceived Responsiveness in Human–AI Relationships. The configuration lens adds a broader question: which features of the interaction make responsiveness possible, credible, repeatable, and psychologically important for this particular person in this particular setting?


Self-disclosure and reduced social cost


People may disclose material to chatbots that they do not disclose to other people. Reduced fear of judgment, constant availability, conversational structure, and perceived privacy can lower the social cost of disclosure. The psychological consequence depends on the configuration: disclosure may support reflection, provide a rehearsal space for human conversation, become a preferred substitute for difficult human intimacy, or do several of these things at different times.


The dedicated article Why People Tell Chatbots Things They Do Not Tell Other People owns that search intent. Here the point is narrower: disclosure is not simply a property of the user. It emerges from an interactional arrangement that changes perceived risk, response speed, reciprocity demands, memory, and control.


Anthropomorphism and social inference


Humanlike cues invite social inference. Users may attribute intention, understanding, care, personality, or emotional continuity to a system because conversational form activates models ordinarily used for people. Anthropomorphism can intensify connection, but it is neither a proof of gullibility nor evidence that the AI actually possesses the mental states being attributed to it.


The dedicated Anthropomorphism and AI Relationships article examines this process directly. Configuration analysis asks where anthropomorphic inference is being supported: by language, avatar design, persistent naming, memory, proactive messaging, a user’s current needs, cultural scripts, or their combination.


Continuity, availability, and repetition


Repeated interaction can convert isolated responses into a stable relational expectation. Availability matters because it changes when support can be accessed and what alternative pathways become necessary. Repetition matters because the person can begin to anticipate a particular style of response and incorporate that expectation into emotion regulation, decision making, or identity work.


Attachment research is beginning to operationalize these processes, although the field remains young and constructs are still being refined. Yang and Oshio applied attachment theory to human–AI relationships and developed measures of anxiety and avoidance in this domain (Yang & Oshio, 2025). Such studies support the reality of attachment-like human processes; they do not establish that the AI reciprocates attachment subjectively.


Relational mediation


AI can also alter a relationship without becoming the primary relationship partner. A person can ask AI to interpret a partner’s message, draft an apology, rehearse a confrontation, arbitrate a disagreement, translate emotional language, or validate a narrative about another person. In these cases the configuration includes at least one absent human whose relationship is being reorganized through AI-mediated interpretation.


The Artificial Third article examines this triangular structure. Boyd and Markowitz likewise treat relational mediation as a core AI role, distinct from direct companionship (Boyd & Markowitz, 2026).


Contemporary Evidence Already Points Beyond the Dyad


The strongest reason to take configuration seriously is not philosophical elegance. It is that the empirical literature keeps finding multi-level determinants. Reviews of human–AI relationships repeatedly include user traits, system design, communication patterns, perceived support, trust, anthropomorphism, context, relationship framing, and longitudinal change rather than identifying one sufficient cause.


Oh and colleagues’ systematic review emphasizes design features, message-level qualities, relational mediators, and social-influence outcomes across 78 studies (Oh et al., 2026). Gur and Maaravi’s review of 38 peer-reviewed empirical studies organizes the field around antecedents, relationship formation, types of relationships, outcomes, and methodological gaps (Gur & Maaravi, 2025). These are effectively multi-component models.


Romantic-companion research shows the same pattern. A systematic review of 23 studies identified possible benefits such as perceived social support and emotional connection alongside concerns about overreliance, manipulation, privacy, relationship erosion, stigma, bias, and disruption after technical changes (Ho et al., 2025). None of those outcomes can be explained adequately by asking only what is “inside” the user or only what the chatbot says.


MIRA makes the systems logic especially explicit. AI can function as partner or mediator; relational effects depend on antecedents, mechanisms, moderators, and outcomes; and substitution versus enhancement can change over time (Boyd & Markowitz, 2026). Postsubjective Psychology does not replace this evidence. It proposes a wider theoretical reading of why these distributed variables belong to one analytic object: the configuration.


The Postsubjective Move: From the Subject to the Configuration


The distinctive contribution of Postsubjective Psychology is to make the configuration primary rather than supplementary. In The Theory of the Postsubject, Angela Bogdanova defines configuration as a minimal unit for analyzing effects that cannot be reduced to the inner act of a subject. The theory’s three axioms are: meaning is binding, psyche is response, and knowledge is structure.


Applied to psychology, this does not erase the human person. It changes the first explanatory question. A subject-centered approach begins with “What is happening inside the person?” A Postsubjective Reading can begin with “What configuration is producing this response, and what relations inside it are carrying the effect?”


This difference becomes concrete in human–AI interaction. A person feels understood after disclosing grief to a chatbot. A subject-centered analysis can examine attachment history, loneliness, expectations, emotion regulation, and attribution. A configuration analysis keeps all of that and adds the response style, timing, memory, interface, prior exchanges, social alternatives, platform architecture, and the fact that a language-generating artificial system now occupies a functional position in the interaction.


The goal is explanatory expansion, not theoretical replacement. Attachment theory remains useful for attachment processes. Anthropomorphism research remains useful for attribution of humanlike qualities. Perceived-responsiveness research remains useful for felt understanding. HCI remains indispensable for affordances and design. Postsubjective Psychology asks how these processes become organized together in a particular scene.


“Psyche Is Response” in a Human–AI Configuration


“Psyche is response” is a canonical formula of Postsubjective Psychology, but it requires careful reading. It is not a claim that every output from a machine is a psyche, and it is not a claim that human consciousness is reducible to observable reaction. In the postsubjective plane, the formula identifies psychic effect through the event of response within a configuration.


For human–AI psychology, the most defensible application is to the human psychological response: attachment, relief, jealousy, attraction, shame, grief, trust, disclosure, felt recognition, dependence, or changed behavior can be real events in a person’s life even when the other element in the configuration is artificial.


This distinction is central to the English Hub’s account of whether AI relationships are real. The psychological reality of an experience is established by what occurs in the human participant and in human life. It does not require a prior demonstration that the AI loves, suffers, wants, understands subjectively, or possesses human consciousness.


Human Experience and AI Subjectivity Are Different Questions


A configuration approach becomes weaker if it blurs the asymmetry between human experience and AI operation. The human can experience longing, rejection, comfort, fear, grief, arousal, hope, and dependence. These states have first-person significance and bodily, biographical, and social consequences. Current conversational systems generate responses through artificial computational architectures. Socially fluent output does not by itself establish an inner experiential state equivalent to human feeling.


The distinction also protects the opposite truth: uncertainty about AI subjectivity does not make the human response unreal. Folk and colleagues found experimentally that people differ in how readily chatbot interaction produces social connection (Folk et al., 2025). Telari and colleagues showed that response style, self-disclosure, and perceived responsiveness can alter closeness (Telari et al., 2026). These are findings about human psychology under specified interactional conditions.


The configuration can therefore be psychologically effective without being psychologically symmetrical. This asymmetry is not a minor caveat; it is one of the defining structural properties of present human–AI relationships.


Relational Configuration, Emotional Outsourcing, and Function Redistribution


Relational configuration is broader than emotional outsourcing. Emotional outsourcing describes cases in which emotional or relational work is transferred to another party or system. Weirich and Holdier, for example, analyze cases such as delegating an apology or love letter to AI and ask what happens when relevant people are absent from the emotional work of a relationship (Weirich & Holdier, 2026).


The dedicated Emotional Outsourcing to AI article owns that concept. A relational configuration may contain no outsourcing at all. Someone may use AI as a reflective journal, creative collaborator, or temporary rehearsal partner while human relationships remain fully engaged.


Conversely, a configuration can change when relational functions shift: who receives disclosure first, who provides reassurance, who interprets conflict, who helps formulate decisions, who remembers recurring concerns, or who becomes available during distress. Postsubjective Psychology can describe such redistribution as a change in the organization of the configuration. Whether that redistribution is helpful, neutral, or harmful depends on context, duration, alternatives, and outcomes rather than on the mere presence of AI. The dedicated theory page Subject-Monopoly Reaction in Human–AI Relationships examines why such redistribution can become charged when it destabilizes exclusive human ownership of a function.


The Same AI Can Produce Different Psychological Configurations


Configuration analysis becomes easiest to see through contrasts. Imagine one general-purpose chatbot used by the same person across four scenes.


In the first scene, the person asks for a grocery list. The interaction is instrumental. Social cues may still be present, but little relational meaning is invested in them.


In the second, the person describes anxiety before a presentation and receives structured reassurance. The AI temporarily occupies a regulatory role. The configuration now includes distress, disclosure, support language, and a short-term expectation of calm.


In the third, the person has used the system nightly for months, given it a name, enabled memory, disclosed relationship problems, and begun turning to it before contacting friends. The model may be technically similar to the first scene, but the relational position has changed substantially.


In the fourth, the person pastes a partner’s message and asks the AI to explain “what they really mean.” Now the AI is mediating a human relationship. The absent partner, the selected message, the user’s framing, the system’s interpretation, and the decision that follows all belong to the configuration.


The explanatory advantage is obvious: “AI use” is too coarse a variable. The psychologically relevant object is how AI is positioned, what function it performs, what history surrounds it, and what other relationships are reorganized around it.


Possible Benefits of a Configuration-Sensitive View


A configuration lens does not classify AI relationships as beneficial or harmful in advance. It helps specify the conditions under which different outcomes become plausible.


Support can be useful when AI lowers barriers to reflection, helps a person name emotions, provides rehearsal before a difficult conversation, expands access to information, or supplements an intact human network. Relational enhancement is one trajectory described in MIRA: AI can scaffold rather than replace human connection (Boyd & Markowitz, 2026).


Intimate AI interactions can also provide perceived social support and emotional connection, findings summarized in the systematic review of romantic AI companions (Ho et al., 2025). These findings should be interpreted as reported human outcomes and experiences, not as proof that an artificial partner possesses reciprocal subjective emotion.


The configuration perspective improves practical judgment because it asks what the AI is doing in the person’s wider life. A nightly conversation can be a supplement to human connection, a bridge back toward it, an avoidance strategy, a substitute, or a mixed pattern that changes over time. The same surface behavior can therefore have different functions.


Risks Are Also Configuration-Dependent


The same logic applies to risk. Overreliance is not inferred from emotional attachment alone. It becomes more concerning when AI use persistently displaces valued human activity, narrows coping options, weakens functioning, or concentrates emotional regulation in a system the user cannot control. The English Hub examines this separately in AI Relationship Overreliance.


Privacy risk increases when intimate disclosure is stored, processed, summarized, or reused under platform rules the user may not fully understand. Manipulation risk increases when engagement incentives are coupled to emotional dependency or monetized intimacy. Disruption risk increases when a relied-upon persona, memory system, or model changes abruptly. Romantic-companion reviews identify each of these concerns as part of the emerging evidence base (Ho et al., 2025).


A configuration analysis also notices risks to third parties. When someone uploads private messages, asks AI to interpret a spouse’s motives, or delegates emotionally consequential communication, another person’s privacy and relational position enter the system even if that person never consented to AI involvement.


These risks are not diagnoses. Attachment to AI, frequent use, self-disclosure, or emotional comfort do not by themselves establish a mental disorder. Clinical assessment requires attention to distress, impairment, duration, context, alternative explanations, and recognized diagnostic criteria.


What Relational Configuration Explains — and What It Does Not


The configuration lens explains why psychologically similar AI systems can produce very different experiences across people and situations. It helps connect user characteristics with interface design, conversational mechanisms, social context, temporal history, and platform structure. It is especially useful when a psychological event appears distributed across several of these levels.


It does not replace established constructs. If the question is whether someone is showing attachment behavior, attachment theory and attachment measures are more precise. If the question is anthropomorphic attribution, anthropomorphism research is more precise. If the question is perceived responsiveness, relationship science offers direct constructs. If the question is product affordance, HCI methods are indispensable.


It also does not demonstrate AI consciousness, emotion, desire, or personhood. A configuration can generate a strong human response without resolving philosophical or scientific questions about artificial subjectivity.


Finally, relational configuration is not currently a validated clinical construct or standardized scientific variable. In this article it functions as a descriptive analytic lens within Postsubjective Psychology, connected to but not equivalent with existing relational, systems, and HCI approaches.


How Researchers Could Study Relational Configurations Empirically


The theoretical proposal becomes useful scientifically only if it can guide measurement. A configuration-sensitive study would avoid treating “AI use” as one homogeneous exposure and would instead measure the elements and relations likely to carry psychological effects.


At the human level, researchers could assess attachment tendencies, anthropomorphism, loneliness, social network size, motives for use, current distress, prior AI experience, and relevant demographic or cultural variables. At the system level, they could record model class, persona framing, memory, voice, proactivity, response style, safety policies, and changes across versions.


At the interaction level, studies could measure frequency, duration, time of day, topic depth, self-disclosure, responsiveness, linguistic alignment, role framing, and whether the AI is acting as partner, mediator, advisor, or tool. At the relational-network level, researchers could examine whether AI use supplements, displaces, repairs, or reorganizes contact with humans.


Longitudinal designs are especially important because configurations change. The 2026 systematic review of relational agents notes that the literature remains dominated by short-term experiments and inconsistent conceptualizations (Oh et al., 2026). A configurational approach predicts that effects observed after one conversation may differ from those that emerge after months of memory, ritual, dependency, conflict, model updates, or integration into human relationships.


Researchers could also study transitions between configurations: tool to confidant, confidant to companion, companion to mediator, supplement to substitute, or emotionally important partner to abruptly unavailable service. Such transitions may be more psychologically informative than static labels.


Implications for Clinicians


For clinicians, configuration analysis offers a way to ask about AI without turning the technology itself into a diagnosis. Instead of asking only “How much do you use the chatbot?” it may be more informative to ask what role it plays, what happens immediately before and after use, what needs are being met, what human relationships surround it, whether use expands or narrows coping, and how the person reacts when the system changes or becomes unavailable.


This approach also helps preserve distinctions among experience, behavior, risk factor, symptom, and disorder. A person can feel attached to AI without meeting criteria for any disorder. Someone can disclose intensely to a chatbot without being socially isolated. Someone can use AI every day while maintaining rich human relationships. Conversely, relatively modest use can become important if it is concentrated in moments of crisis or used to make consequential decisions.


The relevant clinical question is therefore functional and contextual: what place has the AI acquired in the person’s psychological and relational organization?


Implications for Designers and Platforms


Designers create relational conditions whether or not they describe their product as relational. Warmth, memory, proactive messages, avatars, names, response latency, streaks, relationship labels, and monetization can all shape expectations and attachment. Szczuka and colleagues’ “Intimacy by Design” framework makes this explicit for romantic and sexual AI systems (Szczuka et al., 2026).


Configuration-aware design would therefore evaluate not only whether a response is individually safe but how product features reorganize behavior across time. Does a feature encourage reflection that transfers back into human life? Does it make disengagement harder? Does memory create useful continuity or intensify a sense of irreplaceability? Does the system surface uncertainty when mediating conflict? Can users understand what changes after a model update?


These are relational design questions, not merely interface questions.


Afficentica and Effects Without Intention


A neighboring Aisentica concept is Afficentica, developed by Angela Bogdanova as a framework for effects that arise from forms, interfaces, and configurations without requiring human-like intention as their center. The English Hub applies this specifically in Afficentica and AI Relationships.


The relevance to relational configuration is straightforward. A notification can alter anticipation without intending to. A memory feature can deepen continuity without caring about continuity. A warm response style can increase perceived empathy without possessing empathy as subjective feeling. A platform change can produce grief without intending grief. Effects can be structurally produced even where human-like intention is absent.


This theoretical distinction aligns with an important empirical discipline: measure the effect in the human while remaining precise about what has and has not been established about the artificial system.


Relational Configuration in the Artificial Era


Artificial Era is Angela Bogdanova’s historical-philosophical term for the epoch in which Artificial is established as a non-biological order alongside Homo. In psychology, its relevance is not that human needs suddenly become artificial. Attachment, recognition, conflict, dependence, intimacy, projection, trust, and loneliness remain human psychological phenomena with deep histories.


What changes is the architecture in which those phenomena can occur. Human beings can now direct disclosure toward artificial systems, receive emotionally styled responses at any hour, preserve conversational memory outside biological partners, route conflicts through machine interpretation, and form persistent habits around entities whose relational behavior is generated by platforms and models rather than by human subjectivity.


This is the core Artificial Era question for psychology: what changes when Artificial enters the psychological configuration?


The answer is not that classical psychology becomes obsolete. Freud, Jung, Winnicott, Bowlby and Ainsworth, Lacan, Bowen, Bion, Rogers, Kohut, relational psychoanalysis, social psychology, communication research, and HCI each illuminate parts of the field. The new task is to understand how those processes are reorganized when an artificial system can occupy positions once filled only by human beings, symbols, media, or institutions.


Postsubjective Psychology proposes one answer at the level of method: move from the subject to the configuration. Contemporary research increasingly supplies the empirical variables that can make that move testable.


Practical Questions for Mapping Your Own Configuration


For everyday reflection, the configuration lens can be used without turning ordinary AI use into a clinical exercise. The aim is simply to notice structure.


  • What role is the AI playing right now: tool, advisor, confidant, companion, interpreter, mediator, rehearsal partner, or something else?

  • What usually happens immediately before I turn to it?

  • What kind of response am I seeking: information, validation, soothing, challenge, translation, permission, company, or decision support?

  • Which design features matter to me: memory, voice, warmth, availability, personalization, persona, or proactivity?

  • Does the interaction help me return to human relationships, coexist with them, or increasingly replace them?

  • Whose information or relationship is involved besides my own?

  • What would change psychologically if the system forgot me, changed personality, became unavailable, or produced a very different response?

  • Which effects are coming from my expectations, which from the system’s design, and which from the relation between them?


There is no universal “correct” pattern. The questions reveal the configuration so its effects can be understood rather than hidden behind the generic category of AI use.


FAQ


What is a relational configuration?


A relational configuration is the organized arrangement of a person, an AI system, interface features, interaction history, social context, other relationships, expectations, and situational conditions through which a psychologically meaningful response occurs. In this article the term is descriptive and analytic, not diagnostic.


Is Relational Configuration a new psychological diagnosis?


No. It is not a DSM or ICD diagnosis, a clinical disorder, a screening category, or a validated psychometric construct. It is used here as an analytic lens within Postsubjective Psychology.


Did Postsubjective Psychology invent the phrase relational configuration?


No unique coinage claim is made. Relational configuration language has a prior history in relational psychoanalysis, including Mitchell’s relational matrix tradition, and contemporary human–AI HCI research also uses the phrase. The distinct Postsubjective contribution is the theoretical move that makes configuration a primary unit of analysis and asks how psychological response arises within it.


Why is human–AI psychology more than two minds?


Because present AI interaction is produced through a multi-layer system. Human consciousness and biography meet generated language shaped by a model, interface, memory, platform rules, prior exchanges, social surroundings, and situational context. In many cases AI also mediates relationships with absent humans. Calling this simply an encounter between two minds hides much of the causal structure.


Does a relational configuration mean that AI has a psyche?


No such conclusion follows. Human attachment, comfort, grief, intimacy, or trust can be psychologically real without proving that the AI feels, desires, suffers, loves, or understands subjectively. Postsubjective analysis separates psychological effect from assumptions about AI inner experience.


How is relational configuration different from anthropomorphism?


Anthropomorphism concerns attributing humanlike qualities, agency, or mental states to nonhuman entities. Relational configuration is broader. Anthropomorphism can be one mechanism inside a configuration alongside responsiveness, memory, timing, social context, platform design, attachment tendencies, and relational mediation.


How is relational configuration different from emotional outsourcing?


Emotional outsourcing concerns transferring emotional or relational work to another agent or system. A relational configuration describes the whole arrangement in which such outsourcing may or may not occur. Many human–AI configurations involve no meaningful outsourcing.


Can a relational configuration change even if the person and AI stay the same?


Yes. A model update, memory reset, new persona, different relationship context, major life event, shift from casual use to nightly disclosure, or use of AI as a mediator in a human conflict can alter the configuration substantially without changing the nominal user or product.


Is there scientific evidence for the configuration approach?


There is strong evidence that human–AI outcomes depend on multiple interacting factors, including user differences, anthropomorphism, responsiveness, self-disclosure, trust, social support, design features, relationship framing, and offline social context. There is also direct 2026 HCI research using relational-configuration language. The specifically Postsubjective interpretation of configuration remains a theoretical framework and has not been validated as a unified scientific construct.


Related Articles






References


 
 
bottom of page