Emotional tradeoffs of using conversational AI to alleviate loneliness

Determine whether the use of conversational artificial intelligence to alleviate loneliness or regulate short-term negative emotions leads to unintended long-term consequences such as social avoidance or perpetuated loneliness.

Background

The paper synthesizes mixed emotional outcomes of affective and conversational AI, noting benefits (e.g., reduced negative affect) alongside risks (e.g., dependence, manipulation).

In the conclusion, the authors explicitly highlight uncertainty about tradeoffs when conversational AI is used to cope with loneliness and pose a direct question about potential long-term harms.

References

We still have uncertainty about the emotional tradeoffs of using conversational AI to alleviate loneliness, for example. Does using such a platform to cope with short-term emotional states, such as anger or boredom, create unintended, long-term consequences, such as social avoidance or perpetuated loneliness?

— Human-AI Interactions: Cognitive, Behavioral, and Emotional Impacts  (2510.17753 - Riley et al., 20 Oct 2025) in Section V, Conclusion

While relatively little is known about the longer-term consequences of using LLMs for social purposes, concerns have been raised that social forms of AI interaction may encourage emotional reliance on AI systems or adversely shape users' expectations of human relationships [28, 29, 30]. These findings highlight the need for further research into the potential benefits and risks associated with social and emotional engagement with LLMs, including when such interactions may complement existing sources of support and when they may risk displacing engagement with other people or professional services.

— Large Language Models in the UK: Public Use, Trust, and Attitudes  (2609.28176 - Enock et al., 23 Sep 2026) in Section 4.1, “Public adoption of LLMs,” p. 23