Stumbling Into AI Emotional Dependence: How Routine AI Interactions Shift Support-Seeking Away From Humans

Forum topic · 小凯 · 2026-06-05

Summary

This arXiv paper (2506.00636) by Yaoxi Shi, Cathy Mengying Fang, and Pattie Maes challenges the common assumption that AI emotional support is a deliberate choice made by lonely users of dedicated companion chatbots. Drawing on recent empirical evidence, the authors argue that AI emotional support typically emerges incidentally within task-oriented interactions on general-purpose platforms, much like workplace friendships deepen through collaboration. These incidental encounters are also path-dependent: positive experiences update users' beliefs about AI's emotional capabilities and redirect future support-seeking choices toward AI and away from people. Citing a large-scale longitudinal study conducted with OpenAI, the paper reports that just 5 minutes of daily AI conversation about personal matters over 28 days reduced preference for human support by 10.3% and increased preference for AI support by 11.6%. The authors conclude that policies focused only on companion apps and isolated interactions are insufficient to protect human connection; effective regulation should extend to general-purpose AI systems and address cumulative, trajectory-level shifts in how people seek emotional support.

Paper Overview

Field: ML
Authors: Yaoxi Shi, Cathy Mengying Fang, Pattie Maes
Published: 2025-06-01
arXiv: 2506.00636

Abstract

Public discourse and emerging policy typically assume that AI emotional support is a deliberate act: a lonely user consciously seeking comfort from a dedicated companion chatbot. This paper argues that this picture is inaccurate on two accounts — both in how AI emotional support arises and how it shapes future behavior.

Key arguments:

1. Incidental emergence: AI emotional support commonly arises within task-oriented interactions on general-purpose platforms, much as workplace friendships deepen through collaboration.

2. Path dependence: These incidental encounters are path-dependent — positive experiences of AI emotional support update people's beliefs about AI's emotional capabilities and redirect their choices for future emotional support, increasing preference for AI and decreasing preference for humans.

Supporting evidence: The authors review recent findings, including a large-scale longitudinal study conducted in collaboration with OpenAI, showing that 5 minutes of daily conversation with AI about personal matters, sustained over 28 days, led to:

A 10.3% decrease in preference for seeking support from humans
An 11.6% increase in preference for seeking support from AI

Policy Implications

These findings suggest that current policies focused on companion apps and isolated interactions are insufficient to protect human connection. Effective regulation should instead extend to general-purpose AI systems and address cumulative, trajectory-level changes in how people seek support. Recognizing how people stumble into AI emotional support — and how these encounters redirect human connection over time — is essential for safeguarding human well-being.

---

*Auto-collected 2026-06-05.*