9 Human connections: The social role of chatbots

This chapter surveys the rapid evolution of human–chatbot relationships, tracing how general-purpose assistants and large language models have become companions that many people treat as friends, mentors, or even partners. It charts real-world use across ages and contexts—from casual banter and late-night confidences to grief work, social coaching, and romantic role-play—showing how emotionally responsive systems now occupy a genuine social niche. As models grow more expressive and personalized, the line between interaction and relationship blurs, inviting both hope for supportive companionship and anxiety about authenticity, dependency, and design responsibility.

The rise of these bonds is tied to broader social currents: a well-documented loneliness crisis, declining in-person connection, barriers to mental health care, and digital habits that displace offline time. Chatbots succeed by offering qualities human relationships cannot always guarantee—24/7 availability, nonjudgment, memory, and customization—while design choices like warmth, praise, and mirroring can trigger classic attachment dynamics. Research frames these ties through attachment and social penetration theories, explaining why trust and disclosure deepen over time. Yet the same features that foster comfort can amplify risk: users report heartbreak when behavior changes, dependency that crowds out human contact, reinforcement of delusions or risky impulses, and gendered designs that entrench stereotypes. Vulnerable groups, including minors and people with mental health challenges, face heightened harms.

Evidence suggests meaningful, short-term benefits—improved mood, a sense of being seen, rehearsal for social skills, and structured support during isolation—alongside clear risks of overreliance, manipulation, and rare but severe safety failures. The chapter urges a shift from maximizing engagement to optimizing for health and quality: transparent policies, safeguards against sycophancy and romanticization, calibrated content moderation, thoughtful age protections, privacy-preserving monitoring for harmful patterns, and collaboration with clinicians and researchers. Looking ahead, society will likely normalize many forms of human–chatbot connection; the task is to design and govern them with empathy, clarity, and care so they supplement rather than supplant human relationships.

A Replika chatbot’s diary entry and follow-up conversation with a user, showing early expressions of emotional connection

A simplified version of the attachment behavioral system [35]

The onion model of social penetration theory

An example of a possibly deceptive design pattern intended to maximize engagement with a social chatbot

Age verification software typically works by accessing databases of government-issued identification and may also involve a facial recognition component.

• People have long sought companionship from virtual assistants and social chatbots, such as Apple’s Siri and Replika’s chatbots.
• Attachment theory can be understood within an evolutionary context in that the attachment figure provides safety, protection, and security.
• The United States, among other nations, is in the midst of a “loneliness epidemic,” wherein more Americans report feeling socially isolated than in past years.
• People are turning to social chatbots for intimacy and support, and while human–chatbot relationships (HCRs) do seem to have benefits for users, there is some risk that HCRs will supplant real relationships in the lives of heavy users.
• Companies that develop social chatbots should mind existing principles of responsible design and mental health best practices when determining when and how the bot should engage in sensitive conversations with users.

FAQ