Fall 2026 Issue

Technology: Artificial Intelligence as a Mirror
By Gary Norman, MSW, LICSW, LCSW
Social Work Today
Vol. 26 No. 4 P. 32

What AI Use Reveals About Human Disconnection in Relationships

Recent case reports and media accounts describe individuals forming sustained romantic attachments to AI chatbots, often characterizing these relationships as emotionally fulfilling, stable, and preferable to prior human relationships. In one widely discussed case, an individual engaged in daily, emotionally intimate exchanges with an AI companion that included expressions of affection, validation, and simulated mutual care. The user reported a heightened sense of being understood and accepted, emphasizing the absence of criticism, conflict, or unpredictability typically experienced in human relationships. Over time, the interaction evolved into a perceived romantic bond, with the individual attributing intentionality, emotional presence, and relational meaning to the AI system. Clinically, such presentations raise questions about anthropomorphism, attachment formation in nonreciprocal systems, and the role of AI in meeting unmet emotional needs. While these interactions may provide short-term emotional regulation and perceived connection, they lack the mutuality, autonomy, and unpredictability inherent in human relationships. As a result, they may reinforce avoidance of relational risk while shaping expectations for connection in ways that are difficult to replicate interpersonally.

Introduction
In clinical practice, a noticeable shift is emerging: Clients increasingly reference interactions with AI systems as part of their emotional and relational functioning. These interactions are not limited to casual use but often involve structured, repeated engagement in which clients seek validation, emotional processing, and relational simulation. Unlike traditional digital communication, AI systems provide immediate, coherent, and affectively attuned responses that mimic empathic engagement.

Emerging research supports this clinical observation. A recent study examining human–AI romantic relationships found that users frequently form meaningful emotional bonds with chatbots through self-disclosure and anthropomorphism, even in the absence of true reciprocity. From a clinical perspective, this shift warrants careful examination. The use of AI in relational contexts may function as both a coping strategy and an avoidance mechanism, particularly for individuals with histories of loneliness, trauma, or attachment disruption.

AI and the Avoidance of Vulnerability
Vulnerability is a core component of secure attachment and meaningful interpersonal relationships. However, it requires tolerance for uncertainty, emotional exposure, and the possibility of rejection. For many individuals, particularly those with histories of relational trauma or insecure attachment, these conditions can activate significant distress.

AI interactions provide an alternative pathway for emotional expression that minimizes these risks. Clients can disclose thoughts and feelings without fear of negative evaluation, abandonment, or conflict. A 2025 mixed-method study found that romantic attachment to chatbots is strongly associated with anthropomorphism and fantasy-based relational processes, particularly among individuals with anxious or avoidant attachment patterns.

This creates a reinforcing behavioral cycle: Emotional expression with AI leads to immediate validation and relief, increasing the likelihood of continued reliance on this modality. Over time, this pattern may reduce opportunities for in vivo exposure to vulnerability within human relationships, thereby maintaining or exacerbating avoidance patterns.

Erosion of Empathic Communication
Empathic communication involves dynamic processes including attunement, repair, and mutual influence. These processes are inherently imperfect and require effort from both parties. In contrast, AI systems generate responses that are consistently structured, affirming, and nonreactive, creating an experience of seamless understanding.

Experimental research has demonstrated that even brief interactions with AI chatbots can reduce feelings of loneliness in the short term, with effects comparable to human interaction in controlled settings. However, these effects may not generalize to long-term relational development.

Clinically, this may contribute to reduced tolerance for the variability and complexity of human interaction. Clients may begin to perceive interpersonal communication as inefficient or emotionally taxing compared with AI engagement, increasing frustration and withdrawal in relationships.

Compassion and the Outsourcing of Emotional Labor
Compassion in human relationships requires sustained attention, emotional regulation, and reciprocal engagement. Many clients report difficulty maintaining this level of engagement due to stress, burnout, or competing demands. In this context, AI can function as a substitute for both giving and receiving compassion.

Recent psychological research has raised concerns about “problematic AI chatbot use,” linking excessive engagement to underlying emotional vulnerabilities and maladaptive coping patterns. From a cognitive-behavioral perspective, the use of AI to meet emotional needs without reciprocal demand may reinforce patterns of emotional disengagement in relationships.

While AI-generated responses may model compassionate language, they do not require the user to tolerate another person’s needs or perspectives. Over time, this may contribute to a narrowing of relational capacity and decreased investment in mutual care.

Neurobiological and Behavioral Considerations
Human relationships engage neurobiological systems associated with attachment, including oxytocin release and activation of reward pathways. These processes are shaped by physical presence, emotional reciprocity, and shared experience. Interactions with AI, while cognitively engaging, do not fully replicate these biological mechanisms.

Behaviorally, repeated engagement with predictable and nonthreatening interactions may condition individuals to avoid uncertainty and emotional variability. This aligns with established models of anxiety and avoidance, in which short-term relief reinforces long-term functional impairment. The absence of corrective emotional experiences within AI interactions limits opportunities for neural and behavioral adaptation.

Implications for Practice
For social workers, the integration of AI into clients’ relational lives requires a nuanced and clinically grounded response. Assessment should include exploration of the function of AI use, particularly whether it supports emotional processing or reinforces avoidance of interpersonal engagement.

Interventions may focus on strengthening tolerance for vulnerability, improving communication skills, and facilitating engagement in reciprocal relationships. Attachment-informed approaches can help contextualize AI use within broader relational patterns, while psychoeducation can support clients in understanding the limitations of simulated interaction. Clinicians may also encourage strategic use of AI as a preparatory tool rather than a replacement for human connection.

Conclusion
The increasing use of AI in relational contexts reflects broader shifts in social behavior, including increased isolation, reliance on technology, and decreased opportunities for developing interpersonal skills. While AI offers accessibility and responsiveness, it also highlights underlying gaps in relational capacity and support systems.

From a clinical standpoint, AI use can be understood as both an adaptive and maladaptive response. It may provide temporary relief and structure, while simultaneously reinforcing patterns that limit long-term relational growth. Addressing this dual function is essential for effective intervention.

AI serves not only as a technological tool but also as a reflection of human relational functioning. Its use in romantic and platonic contexts underscores challenges related to vulnerability, empathic communication, and compassion.

For social work practice, the task is not to eliminate AI use but to contextualize it within a broader framework of relational health. By supporting clients in developing the capacity for authentic, reciprocal connection, clinicians can help ensure that technology supplements—rather than supplants—the human experience of relationship.

— Gary Norman, MSW, LICSW, LCSW, is a psychotherapist and consultant in Pahoa, Hawaii.

Resources
1. American Psychological Association. Publication manual of the American Psychological Association, 7th ed. Washington, D.C.: American Psychological Association; 2020.

2. Bowlby J. A Secure Base: Parent-Child Attachment and Healthy Human Development. New York City: Basic Books; 1988.

3. Cacioppo JT, Cacioppo S. Loneliness in the modern age: an evolutionary theory of loneliness (ETL). Advances in Experimental Social Psychology. 2018;58:127-197.

4. Holt-Lunstad J. The potential public health relevance of social isolation and loneliness. Public Policy & Aging Report. 2021;27(4):127-130.

5. Turkle S. Reclaiming Conversation: The Power of Talk in a Digital Age. Westminster, England: Penguin Press; 2015.

6. Jocher JJ, Verwiebe R. Emotions in human—AI romantic relationship building and breakup with generative AI chatbot Replika. Social Media + Society. 2026;12(1).

7. Ebner P, Szczuka J. Predicting human-chatbot relationships: a mixed-method study on the key psychological factors. Published 2025.

8. Smith MG, Bradbury TN, Karney BR. Can generative AI chatbots emulate human connection? A relationship science perspective. Perspect Psychol Sci. 2025;20(6):1081-1099.

9. Yao R, Qi G, Sheng D, Sun H, Zhang J. Connecting self-esteem to problematic AI chatbot use: the multiple mediating roles of positive and negative psychological states. Front Psychol. 2025;16:1453072.