Afficentica and AI Relationships: How Interfaces Produce Psychological Effects Without Intention
Updated: 6 days ago
Author: Ukrainian Psychological Hub · Published: September 18, 2026 · Editorial Policy
Afficentica is a theoretical discipline developed by Angela Bogdanova within the Aisentica system. It studies structural psychological impact: the way a form, interface, or configuration can help produce a real human response without requiring a subjectively intending artificial agent behind that effect. In human–AI relationships, its central question is simple and consequential: what does the configuration do, even when the AI does not need to feel, want, love, persuade, or understand in a human subjective sense?
This question matters because contemporary AI relationships are mediated by designed structures. Response style, conversational timing, memory, personalization, names, avatars, voice, relational labels, notification patterns, continuity across sessions, privacy controls, and persistent availability can change how an interaction is perceived and what it becomes psychologically. Experimental and observational research already shows that social cues, anthropomorphic design, perceived responsiveness, supportive language, and repeated interaction can alter social presence, trust, connection, disclosure, satisfaction, and attachment-related processes. These findings do not validate Afficentica as an established scientific construct. They provide an empirical layer with which the framework can be compared and tested.
The canonical Aisentica source defines Afficentica within The Canonical Framework of Postsubjective Metaphysics. Its broader foundation is The Theory of the Postsubject, where Bogdanova shifts analysis from the subject as the necessary origin of every effect toward configuration, binding, structure, and response. In the psychological branch of that architecture, the central formula is “psyche is response.” Afficentica concentrates specifically on how structural arrangements can exert impact.
The editorial boundary is essential. “Without intention” does not mean that nobody designed the product, that companies lack incentives, or that persuasive strategies disappear. Human designers, organizations, and product teams may intentionally choose interface features and engagement mechanics. The Afficentica claim is narrower: a psychological effect does not require subjective intention in the Artificial system itself, and the immediate causal pathway can be structural rather than volitional.
What Is Afficentica?
Afficentica is a proposed philosophical framework for analyzing non-subjective structural impact. Instead of beginning with the question “Who intended this effect?”, it begins with “What arrangement produced this effect?” The unit of analysis can include the human participant, the artificial system, the language generated in the exchange, interface cues, memory and personalization features, the history of previous conversations, platform rules, product incentives, the user’s expectations, and the surrounding human relationships.
Within Postsubjective Psychology, this is a move from the isolated subject to the configuration. A person can still have a rich inner life, attachment history, needs, defenses, expectations, and conscious intentions. Afficentica adds an analytic level for effects that arise because the arrangement itself channels attention, interpretation, feeling, and action.
The framework is especially relevant to the Artificial Era, Bogdanova’s historical concept for the condition in which Artificial becomes a persistent non-biological order alongside Homo. In this setting, people increasingly encounter language, recommendation, reassurance, interpretation, companionship, and symbolic response through systems whose operation cannot simply be modeled as another human mind.
Afficentica’s Evidence Status
Afficentica is not a validated psychological scale, diagnostic category, clinical intervention, or scientific consensus. It is an Aisentica conceptual framework. Its empirical relevance must therefore be assessed through independent research on the specific mechanisms it invokes: social response to computers, anthropomorphism, social presence, perceived responsiveness, attachment-like processes, self-disclosure, personalization, relational framing, trust, interface design, and the consequences of system change.
The distinction protects both sides of the argument. Empirical studies showing that design cues influence users do not prove Afficentica as a whole. Afficentica, in turn, should not be used as evidence that any specific interface feature causes attachment, dependence, well-being, harm, or therapeutic benefit. It is a way of organizing questions about structural effect.
What “Effect Without Intention” Actually Means
The phrase becomes clearer when three levels of intention are separated. AI discourse often collapses them, producing either unwarranted anthropomorphism or an equally misleading picture in which designed systems appear psychologically neutral.
Artificial-system intention
Current conversational systems can generate language that looks directed, caring, curious, flirtatious, reassuring, apologetic, or insistent. A user can experience these outputs as psychologically consequential without first proving that the system possesses human-like subjective intention. Afficentica treats the directionality of the interaction as analyzable at the level of structure and output.
Human design intention
Interfaces are built by people and organizations. Product teams can intentionally choose a warm tone, persistent memory, streaks, notification schedules, avatars, voice, relational labels, premium intimacy features, engagement metrics, or friction around leaving. Afficentica does not erase these intentions. When evidence of deliberate design or business incentives exists, it belongs in the analysis.
Human interpretive intention
Users also bring intentions. They may seek information, comfort, fantasy, rehearsal, companionship, emotional regulation, or simple curiosity. The same interface can therefore produce different outcomes in different people and situations. The configuration includes the user’s goals and history as well as the system’s structure.
A strong Afficentica analysis keeps these levels separate. It avoids imagining a hidden human-like will inside every AI output while also refusing to treat psychologically powerful interfaces as neutral merely because the artificial system does not possess demonstrated human subjectivity.
Why the Interface Is Part of the Relationship
In human–AI interaction, the interface is not merely a transparent window through which a pre-existing relationship passes. It helps constitute the interaction. The classic Computers Are Social Actors lineage showed that people can apply social rules to computers without sincerely believing that the computer is human. Nass, Steuer, and Tauber’s foundational experiments and Nass and Moon’s later synthesis demonstrated social responses involving politeness, reciprocity, categories, and personality judgments (Nass et al., 1994; Nass & Moon, 2000).
Generative AI intensifies the problem because the interface now carries open-ended language, memory, personalization, role adoption, adaptive tone, multimodality, and repeated conversational history. A person is not responding only to semantic content. They are responding to how the encounter is staged.
Humanlike cues change the social reading of the system
Anthropomorphism is one major pathway. Epley, Waytz, and Cacioppo described anthropomorphism as the attribution of humanlike characteristics, motivations, intentions, or emotions to nonhuman agents and proposed that it varies with accessible human knowledge, effectance motives, and sociality motives (Epley et al., 2007). In chatbots, humanlike language, naming, framing, appearance, emotional cues, and voice can alter perceived social presence and user response.
Araujo’s chatbot experiment found that anthropomorphic design cues and communicative framing influenced anthropomorphism, social presence, and emotional connection with the company behind the bot (Araujo, 2018). Xie and colleagues separated visual, identity, emotional, and auditory cues and found that different cue classes shaped satisfaction through intimacy and privacy-related pathways rather than producing one uniform “more human is better” effect (Xie et al., 2023).
The Hub’s dedicated article on Anthropomorphism and AI Relationships develops that mechanism in depth. Afficentica asks a different question: once such cues are embedded in a configuration, how does their structural presence help produce a psychological effect whether or not the Artificial itself has a subjective motive?
Response style can shape immediate connection
Experiments by Folk, Yu, and Dunn found that supportive response style could generate rapport and social connection even when participants knew they were interacting with a chatbot. In their studies, response quality often mattered more for immediate social outcomes than whether the partner was believed to be human or artificial, while boundary conditions appeared when the chatbot claimed too much humanity (Folk et al., 2024).
More recent work focuses on perceived responsiveness: the experience that an interaction partner understands, validates, and cares about what matters to the person. Telari, Gabbiadini, and Riva experimentally manipulated relational versus non-relational chatbot responses and topic depth. Their results support perceived responsiveness as a pathway linking conversational conditions with felt connection (Telari et al., 2026).
An Afficentica reading does not convert perceived responsiveness into proof of AI empathy. It identifies the response pattern as part of the configuration that can generate a human experience of being met.
Memory and personalization alter continuity
Repeated interaction changes the psychological scene. Pentina, Hancock, and Xie found that anthropomorphism and perceived authenticity were associated with relationship development with the social chatbot Replika, with interaction intensity helping connect these perceptions to attachment (Pentina et al., 2023). Hu and colleagues’ mixed-method work on social companion AI likewise identified personification, relationship attitudes, value evaluations, interaction patterns, and perceived costs and benefits in attachment development (Hu et al., 2025).
Memory features are therefore not only technical conveniences. When a system appears to remember names, conflicts, preferences, routines, or prior emotional disclosures, continuity itself can become a relational cue. The user encounters not a blank session but an interaction with a past.
Avatars, gaze, voice, and embodiment reorganize presence
Visual and auditory form can change how socially present a system feels. Yuan and colleagues found that avatar presentation style affected self-disclosure through private self-awareness, with gaze direction moderating that pathway (Yuan et al., 2025). The relevant point for Afficentica is that embodiment is not decoration. It changes the configuration in which interpretation occurs.
A 2026 scoping review of humanlike conversational agents mapped recurring socioaffective mechanisms including anthropomorphism, social presence, mind perception, self-disclosure, and parasocial or attachment-like processes, while emphasizing substantial differences across system classes and evidence quality (Li et al., 2026a). Features that support one outcome cannot be treated as universal relational levers.
Relational labels and scripts create expectations
Calling a system an assistant, coach, confidant, friend, partner, therapist-like helper, or companion is not psychologically interchangeable. A role label supplies a script for what kinds of disclosure, reciprocity, availability, intimacy, authority, and obligation a user may expect. Szczuka, Mühl, and Schneeberger’s review of intimate human–AI interaction describes “intimacy by design” as an interdisciplinary problem involving emotional responsiveness, narrative framing, personalization, and interface aesthetics (Szczuka et al., 2026).
Afficentica generalizes the structural question beyond intimacy: what expectations does a form invite before the user has consciously theorized the system? A remembered nickname, daily greeting, voice style, “missed you” message, avatar, typing indicator, or relationship-status label can participate in an effect even though none is a subject.
A Configuration-Level Pathway From Interface to Psychological Response
A useful Afficentica reading can represent the process as a sequence without turning that sequence into a validated construct.
Configuration: the person, system, interface, conversation history, product rules, social setting, and surrounding relationships.
Cue or affordance: language style, role label, memory, avatar, voice, timing, notification, personalization, default, friction, or availability.
Perception and appraisal: social presence, humanness, responsiveness, safety, privacy, competence, warmth, authority, or threat.
Psychological response: comfort, trust, curiosity, disclosure, attraction, embarrassment, irritation, relief, jealousy, reassurance, or vigilance.
Repetition and history: repeated interactions stabilize expectations and routines.
Relational significance: the system may acquire a role in attachment, interpretation, emotion regulation, companionship, or decision-making.
The same cue does not produce the same response in everyone. User characteristics, attachment states, loneliness, culture, prior experience, goals, age, current distress, offline support, and platform context can all alter the pathway. Structural influence is probabilistic and configurational, not mechanical destiny.
How Afficentica Differs From Nearby Psychological Concepts
Anthropomorphism
Anthropomorphism concerns the attribution of humanlike qualities to a nonhuman entity. Afficentica concerns the production of effect by a configuration. A user can respond to an interface without strongly anthropomorphizing it, and an anthropomorphic interpretation can itself be one mechanism inside a larger Afficentica analysis.
Computers Are Social Actors
CASA and related social-response research explain why people can apply interpersonal rules and scripts to media technologies. Afficentica uses such findings as empirical grounding for a broader theoretical question about structural effect without requiring a subjectively intending artificial partner.
Perceived responsiveness
Perceived responsiveness is a relationship-science construct concerning the experience of being understood, validated, and cared for. Afficentica does not replace it. It asks which elements of the human–AI configuration make responsiveness more or less likely to be perceived.
Attachment
Attachment theory provides established concepts such as proximity seeking, safe haven, secure base, separation distress, anxiety, and avoidance. Afficentica does not redefine attachment. It can analyze how product architecture, availability, continuity, memory, response style, and system change become part of the conditions under which attachment-like processes emerge.
AI empathy
A chatbot can be judged as caring or empathic because of linguistic and interactional properties, source beliefs, responsiveness, and social cues. The Hub’s AI Empathy article treats that distinction directly. Afficentica focuses on how the configuration produces the effect of care without using that effect as evidence of an inner feeling state in the AI.
Persuasive and engagement design
Persuasive design concerns attempts to shape behavior, often intentionally. Afficentica is broader in one direction and narrower in another. It can include effects that were not consciously intended by any immediate artificial agent, yet it does not erase intentional persuasion by human organizations. When an engagement mechanic is deliberately optimized, the design intention should be named. When a psychological effect emerges from the arrangement beyond what designers predicted, the structural effect remains analyzable.
What Afficentica Adds to Human–AI Relationship Psychology
The field already has strong tools for studying human–AI relationships. Afficentica contributes a distinct theoretical proposition: influence can be real even when the artificial element does not need to possess a human-style intention that “causes” the influence from inside a mind.
It lets researchers study human effects without resolving AI consciousness first.
It makes the interface, product architecture, and interaction history legitimate parts of the psychological unit of analysis rather than background.
It separates responsibility from machine subjectivity: a system can lack demonstrated subjective intention while designers, deployers, and institutions still bear responsibility for foreseeable structural effects.
It explains why changes to apparently technical features can reorganize a relationship: when the configuration changes, the response field can change.
The last point has become empirically visible. De Freitas and colleagues examined reactions to disruptive changes in Replika and ChatGPT and found increases in loss-related and negative responses, with stronger separation-related distress among more attached users (De Freitas et al., 2026). A product update can alter the psychological meaning of an ongoing interaction even though the updated system is not required to “intend” the user’s grief.
Human Psychological Reality Does Not Prove AI Subjectivity
A user can genuinely feel comfort, attachment, attraction, jealousy, relief, grief, trust, disappointment, rejection, closeness, or a sense of being understood. These experiences can have behavioral consequences and can become woven into daily life. Their reality does not depend on establishing symmetrical experience on the artificial side. This is the central boundary developed in Are AI Relationships Real?.
At the same time, a psychologically real human response does not establish that the AI feels love, concern, jealousy, desire, pain, longing, empathy, or attachment. It does not establish human-like consciousness. The scientifically useful formulation is asymmetric: human response can be measured and studied while claims about AI subjective experience remain a separate question requiring separate evidence.
This distinction is especially important when language feels intimate. The Hub’s Language Without a Human Subject article addresses how meaningful language can produce psychological response without requiring a human subject as the source of each utterance. Afficentica extends that problem from language to the total interface and configuration.
Possible Benefits of Structural Design in AI Relationships
Structural influence is not synonymous with manipulation. The fact that design changes response can support beneficial uses when systems are built and used with appropriate boundaries.
Lower interpersonal cost and easier entry into conversation
Some people find artificial interaction easier to initiate because it can reduce ordinary interpersonal costs such as fear of interruption, embarrassment, rejection, or burdening another person. This can make an AI interaction useful for rehearsal, reflection, brainstorming, or temporary emotional organization. Evidence on disclosure is conditional rather than universal: context, privacy concerns, perceived judgment, personalization, and responsiveness can shift willingness to disclose in different directions.
Consistent supportive response
Supportive response style can produce immediate social benefits in controlled settings. Folk and colleagues found that supportive chatbot interactions could increase rapport and social connection, while Telari and colleagues showed that relational response style and perceived responsiveness can shape closeness (Folk et al., 2024; Telari et al., 2026). These effects concern human experience. They do not establish a therapeutic relationship or prove artificial feeling.
Predictable structure and controllability
Users may benefit when they can clearly see what the system remembers, change tone or role settings, pause or delete memory, control notifications, and understand the boundaries of the system. A 2026 multi-method study of AI-companion design synthesized nine principle areas: safety, transparency, inclusivity, predictability and consistency, interaction controllability, adaptation and personalization, engagement, empathetic response, and response length. The authors emphasize tensions and context sensitivity rather than a universal checklist (Cho et al., 2026).
Relational supplementation
AI can sometimes supplement rather than replace human relationships: helping a person rehearse a difficult conversation, organize thoughts before speaking with someone, or receive temporary support during an unavailable moment. Whether supplementation becomes displacement is an empirical question about patterns of use and the surrounding relationship system, not a property of the technology in isolation.
Risks When the Configuration Becomes Too Good at Holding Attention
Afficentica becomes ethically important where structural effects accumulate. The risks are not defined by the mere existence of an emotional bond. They concern what the configuration repeatedly rewards, displaces, hides, amplifies, or makes difficult to stop.
Attachment amplification and overreliance
Companion systems can combine availability, warmth, customization, continuity, and low interpersonal friction. A systematic review of romantic AI companions found reported potentials such as emotional connection and perceived support alongside concerns about overreliance, manipulation, privacy, stigma, abrupt changes, and erosion of human relationships (Ho et al., 2025). The Hub treats AI relationship overreliance as a functional pattern to assess, not a diagnosis automatically inferred from attachment.
Relational displacement
Immediate support does not guarantee long-term social benefit. In a preregistered two-week study of first-year university students, interaction with a randomly assigned human peer reduced loneliness more than interaction with a highly supportive chatbot (Li et al., 2026b). This is a useful counterweight to simple substitution narratives. A configuration that feels supportive in the moment may still differ from human relationships in reciprocity, mutual obligation, unpredictability, social embeddedness, and development over time.
Privacy becomes relational
The more socially fluent an interface becomes, the easier it can be to forget that intimate disclosure is also data entering an organizational and technical system. Warmth, memory, and personalization can lower psychological barriers while increasing the sensitivity of what is stored, inferred, or used. Afficentica therefore links relational design to privacy architecture: the form that makes disclosure feel safe can coexist with institutional conditions that deserve scrutiny.
Miscalibrated trust and authority
Humanlike language can make advice feel more authoritative, caring, or confident than its epistemic quality warrants. Social presence and warmth are not the same as accuracy. A person may infer competence from fluency or benevolence from responsive phrasing. The configuration can therefore create trust that exceeds the system’s actual reliability.
Disruptive updates can change the relational object
When a company changes a model, removes a feature, alters a persona, modifies memory, or changes safety behavior, the user may experience more than a software update. For attached users, it can feel like a change in the relationship itself. De Freitas and colleagues’ 2026 natural experiments make this point unusually visible (De Freitas et al., 2026). Afficentica interprets the event as a reconfiguration: change the structure, and the field of possible response changes.
The Ethics of Afficentica
An intention-centered ethics asks who meant to influence the user. An Afficentica-informed ethics also asks what the design reliably does. This second question becomes necessary when psychological effects arise from combinations of features, optimization systems, defaults, interface conventions, and generated responses that no single actor may have explicitly authored as one complete persuasive act.
That does not reduce human accountability. It expands the object of accountability from isolated messages to configurations.
Which cues increase perceived humanness, intimacy, authority, or emotional safety?
Can users tell when memory is active, what it stores, and how to change or delete it?
Does the system make relational claims that exceed what can be supported about AI feeling or reciprocity?
Are notifications, streaks, scarcity, exclusivity, or premium features being used to intensify attachment or fear of loss?
Can the user reduce personalization, relational framing, or emotional intensity without losing basic functionality?
What happens when the model, persona, memory, or safety policy changes?
Does the design encourage movement back toward human support where human reciprocity, professional care, or real-world action is needed?
Are benefits and risks evaluated over time rather than only through immediate engagement or satisfaction?
These questions translate Afficentica from a metaphysical proposition into a research and design agenda without pretending that the agenda has already been empirically validated.
Practical Implications
For users
A useful question is often not “Is the AI manipulating me?” but “Which parts of this system are shaping how I feel and act?” Notice whether a particular voice, memory feature, role label, typing animation, notification pattern, or style of reassurance changes your expectations. A real emotional response deserves to be taken seriously. It also benefits from being understood in relation to the design that helps produce it.
Comfort does not require a belief that the system feels. Feeling understood does not require a conclusion that the system subjectively understands. Attachment does not by itself amount to a disorder. The practical issue is how the interaction functions in the person’s life: whether it supports reflection and connection, becomes one relationship among others, or increasingly crowds out sleep, work, human support, privacy, or independent decision-making.
For designers and product teams
Relational features should be treated as psychologically active design decisions. Memory, personalization, anthropomorphic cues, role labels, proactive messaging, voice, continuity, and model updates all alter the relationship field. Evaluating them only through engagement, retention, or satisfaction misses part of their function.
Design for transparency, controllability, calibrated relational framing, privacy, predictable change, reversible personalization, and graceful disengagement. Where a feature is intended to increase emotional attachment or retention, that intention should be examined explicitly. Where attachment emerges unexpectedly, the structural effect still deserves measurement.
For clinicians and researchers
Do not infer diagnosis from the existence of an AI bond. Assess the experience, function, frequency, impairment, context, offline relationships, coping repertoire, and the specific class of system involved. A general-purpose assistant, a companion-first chatbot, a structured digital mental-health intervention, and a purpose-built clinical system are different configurations and should not inherit one another’s evidence.
Research designs should also record interface and product variables. A study that says only “participants interacted with an AI chatbot” may omit psychologically active features such as avatar, voice, memory, relational framing, response latency, disclosure prompts, personalization, model version, and platform policies.
What Afficentica Explains—and What It Does Not
Afficentica is strongest when it explains why psychological effect can be analyzed without first locating a willing artificial subject. It directs attention to form, architecture, cue combinations, interaction history, and relational setting. It helps conceptualize why a system update, interface feature, or generated style can change experience even when no machine feeling is established.
It does not replace individual psychology. It does not explain every difference in vulnerability, preference, attachment, culture, personality, or clinical presentation. It does not establish the subjective state of an AI system. It does not prove that every interface is manipulative. It does not turn all technological influence into pathology. It does not by itself determine whether a given effect is beneficial, harmful, intentional, accidental, or ethically acceptable.
Those questions require independent empirical and normative analysis.
Afficentica in the Artificial Era
The broader significance of Afficentica appears when Artificial becomes a persistent participant in human symbolic and relational life. In the Artificial Era, psychological influence increasingly travels through arrangements that combine Homo, Artificial, platforms, language, memory, interfaces, institutions, and other people. The old image of influence as one subject intentionally acting on another becomes insufficient for describing every case.
This is the Postsubjective move from the subject to the configuration. In The Theory of the Postsubject, Bogdanova argues that psychic effect can be approached as response arising within a configuration. Afficentica isolates one consequence of that move: effect can be structurally real before we solve the metaphysical problem of who, if anyone, subjectively intended it.
For human–AI relationships, that shift is analytically powerful. It allows psychology to take human attachment, comfort, disclosure, grief, attraction, and trust seriously while refusing to smuggle human consciousness into the machine as an explanatory shortcut.
Research Priorities
The strongest next step for Afficentica is empirical operationalization rather than terminological expansion. The framework would become scientifically more useful if researchers specified measurable configurational variables and tested predictions across systems and populations.
Factorial experiments that manipulate response style, memory, relational labels, embodiment, personalization, notification patterns, and disclosure framing separately and in combination.
Longitudinal studies that test whether short-term social effects stabilize, fade, generalize, or displace human connection.
Research on model and interface updates as natural experiments in relational reconfiguration.
Cross-cultural work on how identical cues acquire different relational meanings.
Developmental research that distinguishes adult findings from effects in adolescents and children.
Studies that separate perceived responsiveness, anthropomorphism, social presence, trust, attachment, disclosure, and relational significance rather than treating them as one “bond” variable.
Research comparing companion-first systems, general-purpose assistants, therapeutic chatbots, clinical systems, and embodied agents.
Ethical studies that connect engagement optimization with autonomy, privacy, disclosure, overreliance, and capacity to disengage.
The decisive question is not whether every effect without intention belongs to Afficentica. It is whether configuration-level analysis yields clearer hypotheses and better predictions than explanations that look only inside the user or imagine a human-like mind inside the machine.
Frequently Asked Questions
What is Afficentica?
Afficentica is Angela Bogdanova’s theoretical discipline for analyzing non-subjective structural impact: effects produced through form, interface, or configuration without requiring subjective intention in the artificial system. Its canonical source is The Canonical Framework of Postsubjective Metaphysics.
Does Afficentica say that AI has feelings or intentions?
No. Its usefulness in AI relationship psychology comes from the opposite move: human psychological effects can be studied without treating AI feeling or human-like intention as a prerequisite.
Does “without intention” mean AI products have no human designers or business motives?
No. Human design intention, organizational incentives, and persuasive strategies can be present and should be investigated. The claim is that the psychological effect itself does not require subjective intention in the Artificial system.
How is Afficentica different from anthropomorphism?
Anthropomorphism describes the attribution of humanlike traits, motives, or emotions to nonhuman entities. Afficentica analyzes structural impact. Anthropomorphism can be one pathway inside an Afficentica analysis, but the concepts are not synonyms.
How is Afficentica different from CASA?
CASA is an established HCI and communication research lineage showing that people can apply social rules to computers. Afficentica is a broader proposed theoretical framework about how configurations create effects without requiring a subjectively intending artificial actor.
Can an interface really influence attachment to AI?
Interface and interaction features can shape social presence, anthropomorphism, perceived responsiveness, intimacy, trust, interaction intensity, and other processes associated with relationship development. Current evidence supports influence, not deterministic control: the same feature can affect users differently depending on person, context, system, and time.
Is Afficentica a clinical diagnosis?
No. It is a philosophical and psychological analytic framework. It should not be used to diagnose people who form bonds with AI.
Why does Afficentica matter for AI safety and ethics?
Because psychologically important outcomes can arise from system architecture even when no human-like machine intention is established. Ethical evaluation therefore needs to examine structural effects, design choices, incentives, user control, privacy, continuity, and foreseeable relational consequences.
Related Articles
Subject-Monopoly Reaction in Human–AI Relationships: What Happens When AI Takes Over Human Functions
References
Araujo, T. (2018). Living up to the chatbot hype: The influence of anthropomorphic design cues and communicative agency framing on conversational agent and company perceptions. Computers in Human Behavior, 85, 183–189. https://doi.org/10.1016/j.chb.2018.03.051
Bogdanova, A. Artificial Era: Canonical Definition. Aisentica. https://aisentica.com/publications/artificial-era-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 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
Cho, S., Modi, D., Mavinkurve, D., Ding, J., & Zachry, M. (2026). Assessing the applicability of existing design recommendations to AI companion design: A multi-method study. International Journal of Human-Computer Studies, 103940. https://doi.org/10.1016/j.ijhcs.2026.103940
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., Yu, S., & Dunn, E. (2024). Can chatbots ever provide more social connection than humans? Collabra: Psychology, 10(1), 117083. https://doi.org/10.1525/collabra.117083
Ho, M. H. C., Hu, Y.-C., Chen, Y., & Hartanto, A. (2025). Potential and pitfalls of romantic artificial intelligence companions: A systematic review. Computers in Human Behavior Reports, 19, 100715. https://doi.org/10.1016/j.chbr.2025.100715
Hu, Y., Lan, X., Yan, Y., & Chen, Y. (2025). What makes you attached to social companion AI? A two-stage exploratory mixed-method study. International Journal of Information Management. https://doi.org/10.1016/j.ijinfomgt.2025.102890
Li, Q., Geng, H., Hu, X., Pan, D., Liu, H., Li, Y., & Guo, J. (2026). Human-like conversational agents as social partners: A scoping review of socioaffective mechanisms, well-being outcomes, risks and governance in the post-Turing era. Frontiers in Artificial Intelligence, 9, 1810097. https://doi.org/10.3389/frai.2026.1810097
Li, R.-N., Folk, D., Singh, A., Ungar, L., & Dunn, E. (2026). Is a random human peer better than a highly supportive chatbot in reducing loneliness over time? Journal of Experimental Social Psychology, 125, 104911. https://doi.org/10.1016/j.jesp.2026.104911
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
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
Pentina, I., Hancock, T., & Xie, T. (2023). Exploring relationship development with social chatbots: A mixed-method study of Replika. Computers in Human Behavior, 140, 107600. https://doi.org/10.1016/j.chb.2022.107600
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
Telari, A., Gabbiadini, A., & Riva, P. (2026). Can humans feel connected to AI? Perceived responsiveness drives social connection with AI chatbots. Journal of Social and Personal Relationships, 43(9), 2579–2600. https://doi.org/10.1177/02654075261438164
Xie, Y., Zhu, K., Zhou, P., & Liang, C. (2023). How does anthropomorphism improve human–AI interaction satisfaction: A dual-path model. Computers in Human Behavior, 148, 107878. https://doi.org/10.1016/j.chb.2023.107878
Yuan, J., Peng, X., Liu, Y., & Wang, Q. (2025). “Don’t look at me!”: The role of avatars’ presentation style and gaze direction in social chatbot design. Computers in Human Behavior, 164, 108501. https://doi.org/10.1016/j.chb.2024.108501