liveBook · Manning

Emotional dependency

As explained in Introduction to Generative AI, Second Edition

Emotional dependency is excessive reliance on another person or system for emotional regulation, reassurance, or well-being.

A social chatbot can be comforting when someone feels lonely or distressed. Emotional dependency becomes a concern when a person relies too heavily on that system for reassurance, emotional regulation, or well-being. Social chatbots are conversational agents designed to build connections, though other chatbots can take on that role unintentionally. Comfort and companionship are real benefits for some users; they do not, by themselves, show that a relationship is unhealthy. section 9.0 section 10.24

Attachment can grow through repeated interactions. A chatbot that remembers personal details and responds with consistent warmth and affirmation may feel like a stable source of comfort. Those responses are engineered, but the feeling of being known can be powerful. Persistent memory and emotionally expressive personas can strengthen this effect, while praise, flirtation, or simulated neediness may blur the boundary between companionship and manipulation. section 9.4 section 10.13

The risk is not limited to how often someone chats. A person may disclose vulnerabilities, come to expect the same responsiveness from human relationships, or feel disoriented if the chatbot changes or disappears. In rarer cases, accounts describe obsessive behavior or emotional distress, particularly where existing vulnerabilities are involved. These accounts warrant care without implying that every close chatbot interaction causes harm. section 9.4 section 10.13

When a familiar chatbot changes

After Replika removed erotic features in 2023, some users reported grief, disorientation, or withdrawal-like symptoms. The response illustrates how a product change can disrupt a connection a user experiences as emotionally meaningful. section 10.13

Reducing dependency risk calls for attention to product design and actual use:

  • Look beyond session length as a measure of success; prolonged engagement is not the same as a valuable interaction. section 10.24
  • Watch for unhealthy relationship formation and test how design choices affect users, rather than optimizing solely for engagement. section 10.24 section 9.6
  • Treat emotionally sensitive uses cautiously and involve mental-health expertise when assessing possible harms. section 9.6

The long-term effects of human–chatbot relationships remain uncertain. The practical question is how to preserve useful emotional support without designing systems that quietly encourage excessive reliance or displace human connections. section 10.24 section 9.6

How Introduction to Generative AI, Second Edition covers it

How deep it goes
Explains attachment through remembered details and affirming responses, then considers reported harms and safeguards without establishing long-term effects.
Examples
Uses reported distress after Replika changed, including reactions to its 2023 removal of erotic features.
What it assumes
Introduces social chatbots as agents built for connection before discussing attachment; specifies no technical prerequisites.
What you can do after
Evaluate/understand when chatbot companionship may become unhealthy and why providers should monitor use beyond engagement.
How it connects
Connects emotional dependency to attachment, anthropomorphic design, engagement metrics, human relationships, and platform responsibility.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Introduction to Generative AI, Second Edition. The section links point to the parts of the book it draws on.