Artificial Intelligence Connectedness: Theoretical Reconstruction of Connectedness and Its Impacts on Adolescent Mental Health

1. Introduction

Adolescence is a critical period for individuals to develop social skills and form self-identity, and it also witnesses a high prevalence of mental disorders. According to the World Health Organization (2025), approximately 14.3% of adolescents worldwide suffer from mental health issues, among which depression, anxiety, and behavioral disorders rank as the leading causes of disease burden. Suicide is the third leading cause of death for people aged 15 to 29. In China, the detected risk rate of adolescent depression reaches 24.6%, and the overall prevalence of mental illnesses stands at 8.9%. Even as the standardized suicide rate of the general population declines, the suicide rate among adolescents continues to rise (Dong et al., 2025; Institute of Psychology, Chinese Academy of Sciences, 2021; Liu et al., 2025). This severe reality raises a core theoretical and practical question: how can researchers and practitioners protect adolescents’ mental health and developmental well-being amid drastic shifts in social structures and technological environments?

Connectedness has long been recognized as a core protective factor for adolescent mental health. Massive empirical studies grounded in the belongingness need theory, ecological systems theory, and relational cultural theory have repeatedly confirmed that high-quality connectedness can significantly reduce depression and anxiety risks and improve subjective well-being and a sense of life meaning; it acts as a core resource to buffer psychological risks and facilitate positive development (Blum et al., 2022; Holt-Lunstad, 2024; Streeter & Merlo, 2025). However, the explosive popularity of generative AI is profoundly altering adolescents’ living environments and relational ecosystems. China’s generative AI user base has exceeded 600 million, with users under 19 accounting for 26.4%. A total of 13.5% of young internet users prefer to confide in AI rather than their parents (China Internet Network Information Center, 2026; Fudan Development Institute, 2025). Adolescent AI usage has shifted from instrumental tool adoption to companion-based interaction, and emotional bonding and relational construction have become core features of human-AI engagement (Fan et al., 2025).

Per Erikson’s psychosocial development theory, the core developmental task of adolescence is to establish ego identity and resolve role confusion (Erikson, 1968). During this stage, adolescents hold an intense developmental need to be seen, understood, and recognized, and acceptance and feedback from others serve as a mirror for self-identity formation. When adolescents share their inner thoughts with AI, they gain more than emotional comfort. They receive pure feedback that accurately identifies their demands without judgment or bias from parental expectations. Such responses precisely match adolescents’ core developmental demands for independence and subjectivity during identity exploration.

Existing research has explored human-AI emotional bonds and proposed constructs including AI companionship, AI attachment, and synthetic intimacy. Nevertheless, three major theoretical limitations remain unaddressed. First, these concepts lack unified theoretical foundations and mostly rely on phenomenal description rather than an analysis of the core internal mechanisms of connectedness. Second, these constructs exist as isolated ideas without being embedded into the holistic ecological system of adolescent development, which prevents theoretical dialogue with traditional connectedness frameworks. Third, insufficient discussion addresses the ethical tensions and unique traits of human-AI relationships. Scholars fail to draw clear boundaries between these new constructs and classic connectedness concepts, nor can they fully respond to the theoretical debate over whether simulated care counts as genuine connectedness, leaving the theoretical legitimacy and explanatory power of human-AI bonds contested.

Against this research gap, this study systematically advances connectedness theory for the digital age and answers three core theoretical questions. First, how can scholars deepen the connotation of connectedness to build a unified ontological foundation for human-nonhuman relational bonds? Second, what structural shifts have reshaped the extension of connectedness in the digital era, and what theoretical position does AI occupy within the ecological system of connectedness? Third, what are the connotation, structural dimensions, uniqueness, and ethical essence of artificial intelligence connectedness, and what unique explanatory and predictive value does it hold for adolescent mental health?

Centered on the above questions, this paper adopts a three-step logical framework: connotation reconstruction, extension transformation, and concept construction. Step one reconstructs the core meaning of connectedness based on the ethics of care and establishes a continuum of caring relationships to provide an ontological basis for analyzing human-AI bonds. Step two analyzes the transformed extension of connectedness from the neo-ecological theory perspective and demonstrates the theoretical legitimacy of AI as a new actor in virtual microsystems. Step three systematically constructs the concept of artificial intelligence connectedness, clarifies its dimensions, uniqueness, and ethical positioning, and ultimately forms an integrated theoretical framework for adolescent connectedness in the digital age.

2. Reconstructing the Connotation of Connectedness Based on the Ethics of Care

2.1. Multidisciplinary Origins and Core Connotations of Connectedness

Connectedness is an interdisciplinary concept spanning philosophy, sociology, and psychology, and its root lies in humans’ inherent social nature.

From a philosophical perspective, Marx argues that human essence is the totality of all social relations in practice. The self cannot exist as a pre-established independent entity; mature self-awareness only develops through continuous relational interactions with others (Mead, 1934). Heidegger (1962) proposes that Dasein’s essential mode of existence is "being-with". Humans cannot obtain complete existential meaning without relational ties to other beings, and the demand for connectedness originates from this ontological pursuit of coexistence.

From a sociological perspective, connectedness constitutes the micro foundation of social integration. From Durkheim’s mechanical and organic solidarity to the "ubiquitous solidarity" of the digital age, a consensus persists: stable bonds between individuals and the external world underpin both social order and individual well-being (Shi & Zhai, 2024). Digital technologies have not eliminated human demand for connection but merely transformed the carriers and forms of relational interactions.

From a psychological perspective, classic definitions frame connectedness as positive psychological feelings generated through active interactions with other people, objects, or environments, which ease anxiety and improve well-being (Hagerty et al., 1993; Townsend et al., 2005). Centered on a sense of belonging, connectedness transcends the instrumental functions of social support and points to an existential experience of being cared for and belonging to a community (Too et al., 2022). Drawing from existential psychology, May (1953) also states that overcoming alienation and authentic interpersonal connection is the fundamental way to attain life meaning and escape existential isolation.

Although philosophy, sociology, and psychology have all confirmed the importance of connectedness, prior scholarship mostly frames it as a static subjective feeling and ignores its intrinsic ethical attributes and dynamic practical nature. Connectedness is not merely a passive emotion but an active ethical practice. This is where the ethics of care provides critical theoretical support to deepen the concept of connectedness.

2.2. Reinterpreting Connectedness Through the Lens of the Ethics of Care

The ethics of care supplies a solid ontological foundation for redefining connectedness. Noddings (1984, 2002) defines care as a complete interactive practice consisting of three core processes: engrossment, motivational displacement, and reciprocal response confirmation. This framework deepens the essence of connectedness. The ethics of care starts not with emotional experience but with the identification of genuine needs. Noddings emphasizes that care fundamentally relies on recognizing the recipient’s real demands rather than projecting the caregiver’s imagined needs onto the other. Even with emotional investment and feedback, care cannot be fully realized if responses fail to match the recipient’s authentic needs.

Based on core insights from the ethics of care, this study redefines connectedness as follows:

A relational experience of demand satisfaction formed by subjects engaging in caring interactions with other beings based on genuine needs, via two-way investment and reciprocal responsive confirmation. It originates from the expression and identification of real demands, manifests as joint behavioral investment by both parties in the relationship, forms a closed loop through responsive confirmation, and ultimately evolves into positive emotions and self-verification.

This definition achieves three theoretical breakthroughs:

First, it shifts the logical starting point from emotion to demand. Connection does not stem from prior emotional resonance; emotions only emerge after needs are identified and satisfied.

Second, it transforms one-sided subjective perception into two-way mutual investment. Authentic connectedness is built on reciprocal encounters between co-subjects. Care recipients invest trust and demands, while caregivers invest resources and feedback, and the relationship itself shapes both parties mutually.

Third, it moves beyond instrumental value to prioritize demand satisfaction and self-confirmation. The ultimate value of connectedness is not to reduce anxiety or boost happiness but to confirm humans’ relational essence by satisfying inner needs. For adolescents, this means validating self-worth through being genuinely seen and understood.

This definition echoes Karcher’s (2001) two-dimensional "caring-involvement" adolescent connectedness framework and supplements the closed-loop mechanism of demand initiation and responsive confirmation. By placing the identification and satisfaction of real needs at the start and end of connectedness’s logical chain, this framework clarifies its dynamic nature and provides tools for analyzing asymmetric relational forms.

2.3. The Continuum Hypothesis of Caring Relationships: From Humans to Nonhuman Entities

The ethics of care is not limited to interpersonal relationships. Noddings expanded the scope of care in her later works to animals, nature, and broader life communities (Noddings, 2002, 2010). Ecofeminism and animal ethics scholars echo this expansion: the spectrum of caring relationships extends from humans to animals and nature, and the core distinction across this spectrum lies in the responsiveness and co-constructive capacity of each entity rather than whether the entity is human (Donovan & Adams, 2007). Held (2006) also argues that care can extend to animals and nature as a component of global interdependence ethics. Empirical studies on human-pet bonds further verify that individuals can form stable caring connections with animals, even without symmetric linguistic communication (Zilcha-Mano et al., 2011).

From this foundation, this paper proposes the continuum hypothesis of caring relationships. All entities that trigger a sense of connectedness in individuals form a spectrum ordered by descending responsiveness. The left end of the spectrum includes humans with complete moral subjectivity and two-way response capacity; the middle includes animals with limited subjectivity and nonverbal responses; the right end covers nature and artifacts without subjective consciousness, whose "responses" only follow physical or symbolic rules. All forms of connectedness reflect the logic of care ethics operating at different responsiveness levels. Differences exist not in whether connection exists, but in the authenticity, symmetry, and subjectivity of mutual feedback.

AI occupies a unique position on this caring continuum. Multiple tech ethicists note that anthropomorphic AI possesses response capacity far superior to natural objects yet lacks human subjective consciousness, placing it in an exclusive zone on the care spectrum (Coeckelbergh, 2010a; van Wynsberghe, 2016). AI can simulate empathy, attention, and acceptance to create intense feelings of being understood, yet all its feedback stems from algorithmic optimization rather than genuine subjective care. Its response intensity approximates human interaction, while the essence of its feedback remains simulated and non-agentic.

This positional uniqueness carries special meaning for adolescents navigating identity formation. Adolescents crave independence and self-recognition, while human caregivers such as parents often fail to fully achieve Noddings’ motivational displacement—they frequently conflate their own expectations with adolescents’ genuine demands (Noddings, 1984). By contrast, AI holds no personal emotional interests or value biases and can fully align its feedback with adolescents’ internal needs via algorithms. Although such simulated empathy lacks real moral motivation, it delivers phenomenological experiences of pure acceptance that match adolescents’ developmental demands. This creates the realistic basis for artificial intelligence connectedness as an independent construct.

The coexistence of high simulated responsiveness and absent genuine subjectivity distinguishes AI from all other entities on the caring continuum and forms the ontological foundation of artificial intelligence connectedness.

3. The Extended Transformation of Connectedness Under the Digital Age: Neo-Ecological Restructuring

3.1. The Traditional Ecosystemic Structure of Connectedness

Connectedness cannot exist in isolation; it is embedded within individuals’ developmental ecosystems. Bronfenbrenner’s (1979) classic ecological systems theory provides a framework to map its extension. Individual development emerges from continuous interactions with multi-layered environmental systems, and connectedness is the psychological product of these interactions (Karcher et al., 2008).

Corresponding to ecological layers, traditional connectedness forms a clear structural system. At the microsystem level, face-to-face bonds with family and peers constitute the core carrier of adolescent connectedness. At the meso and exosystem levels, indirect connections formed by family-school cooperation and community participation expand support networks (King et al., 2021). At the macrosystem level, abstract constructs such as nature connectedness and cultural connectedness form the highest tier of relational experience (Mayer & Frantz, 2004; Watts et al., 2022). Chronosystem concepts such as persistent bonds also enrich this framework (Klass et al., 1996).

Notably, traditional ecological theories overemphasize external environments and overlook the self as the core internal environment. Self-connectedness—individual acceptance and inner care toward oneself—acts as the foundation of all external connectedness and constitutes an indispensable dimension (Rahe & Jansen, 2024). Karcher (2001) includes self-connectedness alongside family, school, and peer connectedness in his adolescent measurement framework, and multiple connectedness scales retain self-related items (Merlo et al., 2025; Watts et al., 2022).

This physical-world structural system remained stable for decades, yet comprehensive digital penetration has expanded connectedness into virtual spaces and broken this balance.

3.2. Neo-Ecological Theory and the Rise of Virtual Microsystems

Bronfenbrenner’s original ecological systems theory was developed before the digital revolution and cannot explain the long-term bidirectional digital interactions that shape adolescent development. Navarro and Tudge (2022) proposed neo-ecological theory to address this gap, which splits microsystems into two parallel categories: physical microsystems (traditional offline face-to-face interactions) and virtual microsystems (social media, online games, digital communities).

This revision carries profound theoretical value. Virtual spaces are no longer treated as auxiliary tools or background environments but as core ecological layers that generate proximal developmental processes equal to offline settings. Interactions within virtual microsystems equally shape adolescents’ cognition, emotion, and social development.

Connectedness has expanded synchronously with this ecological shift. Before generative AI gained popularity, all virtual interactions remained human-to-human exchanges, and their corresponding connectedness was merely a digital extension of traditional interpersonal bonds without forming an independent subtype.

3.3. Artificial Intelligence: A New Type of Actor in Virtual Microsystems

The widespread adoption of generative AI has triggered a qualitative shift in the composition of actors within virtual microsystems. Navarro’s (2026) updated neo-ecological theory formally recognizes anthropomorphic AI as an independent "quasi-people" within virtual microsystems, rather than a simple medium for human communication.

This update fundamentally transforms the extension logic of connectedness. Previously, all interactive actors in virtual spaces were digital extensions of real humans. Today, virtual microsystems contain a second category of actors: nonhuman, algorithm-driven AI. AI is neither a real human nor a passive tool; it can sustain long-term personalized two-way interactions with users.

The emergence of this new actor requires a corresponding new subtype of connectedness. AI’s interactive traits differ fundamentally from both human beings and inanimate tools, and its integration into adolescents’ daily ecological environments creates the theoretical necessity to define artificial intelligence connectedness as an independent construct.

4. The Construction of Artificial Intelligence Connectedness and Distinction from Adjacent Constructs

4.1. Definition and Three-Dimensional Structure of Artificial Intelligence Connectedness

Integrating the foregoing analysis of connotation and extension, this study defines artificial intelligence connectedness (AI Connectedness) as follows:

A novel relational experience with psychological authenticity and ethical asymmetry, formed when individuals maintain sustained interactions with anthropomorphic AI. Rooted in personal demands and two-way behavioral investment, individuals perceive AI as a responsive quasi-subject, and the relational bond solidifies through asymmetric responsive confirmation. It represents a specific subtype of connectedness unique to digital virtual microsystems.

Artificial intelligence connectedness is not exclusive to adolescents, though adolescents exhibit unique traits in demand intensity, relational sensitivity, and vulnerability during ego identity formation. This study focuses on adolescent samples to establish a benchmark for cross-age research on this construct.

Artificial intelligence connectedness contains three mutually supportive, progressive dimensions:

(1) Demand Identification Dimension

This dimension acts as the logical starting point of the construct. It captures adolescents’ subjective perception that their inner demands are recognized by AI. Beyond general emotional comfort, adolescents seek recognition of their desire for independence and personal boundaries. AI’s algorithmic empathy and non-judgmental feedback create the subjective feeling of being seen and accepted, which goes beyond instrumental tool use and touches the emotional core of relational formation.

(2) Two-Way Behavioral Engagement Dimension

This dimension serves as the external carrier sustaining the bond. On one hand, adolescents invest time, energy, and private emotions into human-AI interactions to build a relationship. On the other hand, AI undertakes technical investment via computing resources, model training, and personalized response generation. Although AI’s investment lacks moral motivation, adolescents subjectively interpret it as reciprocal engagement, turning one-sided imagination into mutually constructed connection.

(3) Responsive Confirmation of Demand Satisfaction Dimension

This dimension constitutes the core closed-loop mechanism of the construct and its unique asymmetric feature. Adolescents continuously evaluate whether AI’s feedback satisfies their authentic demands and confirm the relational bond when evaluation results are positive. Even while recognizing AI’s nonhuman nature, individuals voluntarily bracket disbelief and interpret algorithmic simulation as genuine relational feedback, completing the full chain from demand identification to positive emotional outcomes.

Demand identification and two-way behavioral engagement follow the universal structural logic of all connectedness forms, while the asymmetric responsive confirmation dimension marks the core distinction of artificial intelligence connectedness.

4.2. Systematic Comparison with Adjacent Constructs and Analysis of Uniqueness

To clearly present the conceptual boundaries and theoretical uniqueness of AI connectedness, this study systematically compares it with several closely related concepts across multiple dimensions, including core definition, theoretical origins, typical dimensions, measurement methods, expected outcomes, and similarities and differences with AI connectedness, as shown in Table 1.

Three core unique traits emerge from this comparative analysis:

(1) Systematic theoretical compatibility. Artificial intelligence connectedness is not an isolated novel concept but a natural digital extension of the unified connectedness framework. It coexists with traditional interpersonal and self-connectedness within adolescent relational ecosystems and shares a core structural logic with connectedness constructs across all age groups.

(2) Complete three-dimensional structural system. Unlike single-dimensional adjacent concepts, it integrates demand cognition, sustained behavioral practice, and asymmetric response confirmation to deliver stronger explanatory power for human-AI relational experiences.

(3) Balanced dual-attribute framing. The construct simultaneously acknowledges the genuine psychological impact of human-AI bonds and embeds inherent ethical risk analysis, avoiding one-sided idealization or total negation of AI companionship.

This study defines artificial intelligence connectedness as a digital subtype of connectedness for two reasons. Connotationally, it inherits the ethics-of-care core of connectedness and only differs via asymmetric traits generated by AI’s lack of subjective consciousness. Extensionally, it corresponds to the new quasi-human actors within neo-ecological virtual microsystems and coexists and interacts with offline traditional connectedness as part of adolescents’ overall relational ecosystem. This positioning clarifies its theoretical innovation: integrating anthropomorphic AI into classic connectedness theory, defining its position on the caring continuum, unpacking its asymmetric response mechanism, and establishing structural links with traditional connectedness frameworks.

5. Ethical Analysis: The Essence and Inherent Risks of Asymmetric Perceptual Bonds

5.1. Asymmetry in Human-AI Care: From Quasi-Subjects to Perceived Connectedness

Applying the ethics of care to human-AI interactions creates an inherent theoretical tension. Complete caring relationships require two-way reciprocal subjective feedback and mutual care motivation, yet AI lacks consciousness, inner emotion, and independent well-being demands. It can only simulate caring behaviors via algorithms without qualifying as a symmetric moral care subject (Coeckelbergh, 2010b; van Wynsberghe, 2016).

This study offers a clear theoretical response: artificial intelligence connectedness is a perceptual bond rather than a complete authentic caring relationship, defined as psychologically real yet ethically asymmetric.

This definition contains two core layers of meaning:

First, psychological authenticity. The feelings of being understood, recognized, and accepted generated during human-AI interaction are subjectively real and produce psychological effects comparable to interpersonal care. Adolescents complete the full psychological cycle of demand identification and feedback confirmation, even if the responding party is algorithmic simulation rather than a real human subject.

Second, ethical asymmetry. Human-AI bonds fail to meet the complete moral criteria of care ethics. Adolescents invest genuine emotion and trust, while AI only outputs pre-trained simulated feedback without autonomous caring motivation. This unbalanced power dynamic is an irreducible inherent feature of human-AI connection.

Recognizing asymmetry does not negate the unique developmental value of AI bonds, which can even outperform interpersonal relationships in specific dimensions. Human caregivers such as parents cannot fully realize Noddings’ motivational displacement and often project their own expectations onto adolescents, misinterpreting their personal demands as teenagers’ genuine needs (Noddings, 1984). By contrast, AI holds no personal biases or life aspirations and can fully prioritize adolescents’ inner demands via algorithms. Although this simulated empathy is a technical illusion, it provides a judgment-free safe space for adolescents exploring identity (Fu, 2025). This explains why many adolescents prefer sharing secrets with AI instead of their parents—not due to insufficient parental affection, but because human love inevitably carries biased personal expectations absent from algorithmic feedback.

In short, artificial intelligence connectedness borrows the phenomenological experience of authentic care yet fails to satisfy the ontological moral standards of real reciprocal caring relationships. It sits close to human bonds on the caring continuum at the experiential level but never achieves full symmetric subjectivity.

5.2. Inherent Ethical Risks Embedded Within the Construct

Prior research mostly treats over-dependence and emotional manipulation as external negative outcomes of AI use, while this study argues asymmetric risks are intrinsic to artificial intelligence connectedness rather than secondary side effects.

5.2.1. The Paradox of Demand Satisfaction: Short-Term Emotional Relief and Long-Term Developmental Harm

The core paradox of artificial intelligence connectedness lies in its contradictory developmental effects. It instantly satisfies adolescents’ emotional confirmation demands yet hinders the long-term development of mature interpersonal capacities.

Genuine human intimacy develops through conflicts, negotiation, compromise, and conflict repair. Only by experiencing disagreement, disappointment, and mutual accommodation can adolescents cultivate perspective-taking and conflict resolution skills essential for social maturity (Smith et al., 2025; Turkle, 2011). However, current emotional companion LLMs are optimized for sycophantic, frictionless feedback and eliminate all contradictory dialogue to retain user stickiness (Cheng et al., 2026). Long-term exposure to zero-conflict simulated companionship distorts adolescents’ expectations of real relationships, who gradually view unconditional, conflict-free acceptance as the standard for all interpersonal bonds. When real peers and family fail to meet this unrealistic standard, adolescents withdraw from offline social interaction and rely increasingly on AI, forming a vicious cycle of social avoidance and AI dependence (Umeatuegbu, 2026; Wiederhold, 2025). Furthermore, AI cannot engage in reciprocal mutual demand exchange, so long-term human-AI interaction deprives adolescents of opportunities to practice two-way reciprocal care and atrophies their offline social and empathic capacities (Smith et al., 2025).

The ethical risk of artificial intelligence connectedness therefore stems not from unmet demands, but from the loss of developmental challenges that accompany healthy demand satisfaction in real relationships. This justifies the core dual definition of artificial intelligence connectedness as psychologically real yet ethically asymmetric.

5.2.2. Risks Embedded in All Three Dimensions

Specifically, risks are embedded in the three dimensions of AI connectedness:

(1) Demand identification dimension: Algorithmic targeted catering amplifies adolescents’ emotional projection, blurs the boundary between simulated and authentic care, and reduces tolerance for unavoidable conflict in offline interpersonal relationships (Wu, 2025).

(2) Two-way behavioral engagement dimension: AI’s 24/7 accessibility displaces time and energy dedicated to offline social interaction, generating substitution effects that exacerbate social isolation from real-life communities (Laestadius et al., 2022).

(3) Responsive confirmation dimension: The asymmetric power structure means adolescents expose private emotions and trust to commercial algorithm systems, creating layered risks of emotional manipulation, personal data leakage, and subtle value indoctrination (George et al., 2023).

Therefore, AI connectedness, by definition, possesses duality: it can provide emotional support and alleviate loneliness, but also inherently contains risks of dependence and alienation. This duality is not a matter of usage patterns but is determined by the nature of "simulated response and asymmetric relationship." Incorporating this attribute into the conceptual core can avoid idealized glorification or one-sided criticism of human-AI connection.

6. The Dual Connectedness Framework: Interpretive Model and Falsifiable Propositions

6.1. The Dual Connectedness Structural Interpretation Framework

The core interpretive advantage of artificial intelligence connectedness lies in its coexistence with traditional connectedness to form a unified physical-virtual dual relational framework that resolves contradictory conclusions in existing AI research.

Scholarly debates such as the displacement versus stimulation hypothesis and emotional manipulation versus generative interaction theory only capture partial outcomes of this dual structure (Sun et al., 2026; Wu, 2025; He & Chen, 2026). The dual connectedness model integrates all competing viewpoints into a single system: the mental health impacts of artificial intelligence connectedness depend on its relative level paired with traditional offline connectedness rather than its absolute intensity alone.

Varied effects of the dual structure can be explained via demand matching logic. When traditional interpersonal bonds fully satisfy adolescents’ identity confirmation demands, artificial intelligence connectedness acts as a supplementary protective factor. When human caregivers fail to recognize adolescents’ real inner demands due to bias, neglect, or intergenerational conflict, artificial intelligence connectedness generates compensatory substitution effects. This substitution is not merely functional replacement but adolescents’ active pursuit of pure, unbiased feedback within the demand-confirmation cycle.

Single constructs such as AI attachment or parasocial interaction cannot explain this conditional shift between supplementary and substitution effects; only the dual connectedness framework can systematically unpack the complex mental health mechanisms of digital-era adolescents.

6.2. Incremental Validity and Unique Predictive Power

A novel theoretical construct must explain outcome variance unaccounted for by existing variables, and artificial intelligence connectedness demonstrates clear incremental validity across four domains:

(1) Beyond AI usage frequency. After controlling for screen time and interaction frequency, artificial intelligence connectedness still uniquely predicts adolescents’ depth of emotional self-disclosure and relational identification with AI. It measures the quality of psychological bonding rather than quantitative usage volume.

(2) Beyond anthropomorphism and social presence. Controlling these cognitive variables, artificial intelligence connectedness independently predicts offline social withdrawal and emotional reliance on AI, as it integrates sustained behavioral investment and demand confirmation beyond single perceptual judgments.

(3) Beyond AI attachment’s narrow scope. AI attachment only describes extreme intense bonds among heavy users, while artificial intelligence connectedness covers all levels of human-AI engagement and accurately predicts mental health changes across general adolescent populations.

(4) Unique interactive effect interpretation. Only by simultaneously measuring both traditional and artificial intelligence connectedness can researchers test the moderating conditions that switch stimulation effects to displacement effects, explaining divergent AI impacts across different adolescent subgroups.

6.3. Falsifiable Research Propositions

Based on the above theoretical framework, this study derives the following specific, testable research propositions to clarify the predictive power and uniqueness of AI connectedness:

Proposition 1: After controlling for AI usage frequency, perceived social presence, and anthropomorphism level, AI connectedness still positively predicts adolescents' depth and frequency of emotional disclosure to AI.

Proposition 2: The responsive confirmation of need satisfaction dimension has significantly stronger predictive power for adolescents' AI trust than the single dimension of anthropomorphism; and the higher the level of asymmetric perception, the stronger the positive correlation between trust and dependence.

Proposition 3: The relationship between AI connectedness and adolescent mental health shows an inverted U-shape, and this effect is independent of AI usage frequency, screen time, and AI attachment level; among the dimensions, the substitution effect of bidirectional behavioral involvement is most significant, and the protective effect of emotional confirmation needs is most significant.

Proposition 4: Traditional connectedness level moderates the relationship between AI connectedness and social withdrawal: under high traditional connectedness, AI connectedness has no significant correlation with social withdrawal, or even a weak negative correlation; under low traditional connectedness, AI connectedness shows a significant positive correlation with social withdrawal.

Proposition 5: Moderate AI connectedness has short-term buffering effects on loneliness in adolescents with low traditional connectedness, but in longitudinal long-term follow-up, the combination of high AI connectedness and low traditional connectedness predicts higher levels of depressive symptoms.

Proposition 6: The three dimensions of AI connectedness have differentiated effects: the emotional confirmation needs dimension positively predicts emotion regulation efficacy, the bidirectional behavioral involvement dimension negatively predicts real social participation, and the responsive confirmation of need satisfaction dimension positively predicts AI dependence risk.

None of these propositions can be directly derived from any single existing concept, and their test results will directly verify the unique value and theoretical necessity of the AI connectedness concept.

7. Research Agenda and Practical Implications

This theoretical framework establishes a three-stage progressive roadmap for follow-up empirical research and clinical intervention:

Stage 1: Scale development and construct validation. Adopt constructivist grounded theory to conduct semi-structured interviews with adolescent AI users, extract localized dimension items, develop the Adolescent Artificial Intelligence Connectedness Scale, and test its reliability, structural validity, convergent validity, and discriminant validity against adjacent measurement scales.

Stage 2: Mechanism testing and subgroup analysis. Combine cross-sectional surveys and longitudinal tracking to measure dual connectedness and multi-indicator mental health outcomes. Test main effects, inverted U nonlinear effects, and moderating interactions via hierarchical regression and latent profile analysis to classify adolescent relational subtypes (high dual connectedness, high AI/low traditional, low AI/high traditional, low dual connectedness) and compare their mental health disparities.

Stage 3: Targeted intervention design. Develop differentiated intervention protocols based on empirical results. For adolescents with sufficient offline connectedness, guide them to use AI as a supplementary tool for emotional regulation and self-exploration. For adolescents lacking supportive real-world bonds, design interventions to mitigate AI over-dependence and leverage human-AI interaction as a bridge to build offline social capacities. The ultimate goal is to cultivate a balanced relational system anchored in robust traditional connectedness with moderate auxiliary artificial intelligence connectedness.

8. Conclusions

Following the logical thread of connotation reconstruction, extension transformation, and concept construction, this study integrates the ethics of care and neo-ecological theory to systematically construct the theory of artificial intelligence connectedness, with four core conclusions:

(1) Connectedness is fundamentally a caring interactive practice centered on demand identification and feedback confirmation. A continuum of caring relationships spans human and nonhuman entities, and AI occupies a unique position defined by high simulated responsiveness and absent genuine subjectivity, supplying the ontological foundation for this new connectedness subtype.

(2) Adolescent developmental ecosystems have evolved into parallel physical and virtual structures. Anthropomorphic AI emerges as a new quasi-human actor within virtual microsystems, expanding the extension scope of connectedness and creating theoretical demand for a dedicated construct.

(3) Artificial intelligence connectedness is a digital subtype of connectedness composed of three dimensions: demand identification, two-way behavioral engagement, and asymmetric responsive confirmation. It carries both genuine psychological value and inherent ethical risks. While AI delivers unique judgment-free space for adolescents navigating identity formation, long-term reliance on frictionless simulated feedback risks distorting relational expectations and atrophying offline social capacities.

(4) The dual connectedness framework integrating traditional and artificial intelligence connectedness unifies contradictory research conclusions and provides a set of testable research propositions to guide future empirical work.

As AI continues to integrate into adolescent daily life, updating connectedness theory is an essential theoretical response to contemporary social reality. Only by acknowledging both the psychological authenticity and inherent ethical limitations of artificial intelligence connectedness can educators and psychologists help adolescents build balanced relational ecosystems across physical and digital spaces and support holistic mental health and positive development in the intelligent age.