Understanding Risk and Dependency in AI Chatbot Use from User Discourse

Self-regulation difficulties are the most common psychological risk users experience with AI chatbots. The study found that across two Reddit communities dedicated to AI-related harms, users most frequently reported struggling to control their own usage patterns—suggesting that dependency issues may be a more immediate concern than technical failures or misuse scenarios.

Users' fears about AI center on loss of autonomy and control rather than existential risks. The research identified that emotional responses concentrated on concerns about personal autonomy and control, indicating that everyday users worry most about how AI might diminish their own agency rather than abstract long-term scenarios often discussed in AI safety literature.

Real-world user experiences reveal psychological harms that laboratory studies may miss. By analyzing authentic user discourse from 2023-2025, the study provides empirical evidence of how AI safety concerns actually manifest in lived experience outside controlled research settings, offering insights that speculative or theoretical approaches cannot capture.

Five distinct dimensions of psychological risk emerge from user experiences. The analysis identified 14 recurring themes that cluster into five higher-order patterns of how users experience AI-related distress, providing a structured framework for understanding the variety of harms people report rather than treating AI risk as monolithic.

Emotional patterns vary significantly across different types of AI-related concerns. The study's emotion labeling analysis revealed that different dimensions of risk produce distinct emotional profiles, suggesting that interventions and safety measures may need to be tailored to address specific types of user distress rather than applying one-size-fits-all approaches.