[论文解读] Increasing happiness through conversations with artificial intelligence

Joseph Heffner, Chen Qin+5 more|ArXiv.org|Apr 2, 2025

Resilience and Mental Health被引用 4

一句话总结

本文比较与 AI 聊天机器人对话后的幸福感与日记记录，显示对话后幸福感更高，特别是在负性话题上，并通过情感建模解释其因 AI 回应的对齐与积极偏差所致。

ABSTRACT

Chatbots powered by artificial intelligence (AI) have rapidly become a significant part of everyday life, with over a quarter of American adults using them multiple times per week. While these tools offer potential benefits and risks, a fundamental question remains largely unexplored: How do conversations with AI influence subjective well-being? To investigate this, we conducted a study where participants either engaged in conversations with an AI chatbot (N = 334) or wrote journal entires (N = 193) on the same randomly assigned topics and reported their momentary happiness afterward. We found that happiness after AI chatbot conversations was higher than after journaling, particularly when discussing negative topics such as depression or guilt. Leveraging large language models for sentiment analysis, we found that the AI chatbot mirrored participants' sentiment while maintaining a consistent positivity bias. When discussing negative topics, participants gradually aligned their sentiment with the AI's positivity, leading to an overall increase in happiness. We hypothesized that the history of participants' sentiment prediction errors, the difference between expected and actual emotional tone when responding to the AI chatbot, might explain this happiness effect. Using computational modeling, we find the history of these sentiment prediction errors over the course of a conversation predicts greater post-conversation happiness, demonstrating a central role of emotional expectations during dialogue. Our findings underscore the effect that AI interactions can have on human well-being.

研究动机与目标

• Motivate investigation into how AI conversations impact subjective well-being.
• Compare happiness after AI conversations with journaling on the same topics.
• Explore whether AI sentiment and response patterns influence user happiness.

提出的方法

• Conduct a randomized study with AI chatbot conversations (N=334) and journaling (N=193) on the same topics.
• Use large language model–based sentiment analysis to assess sentiment alignment between user and AI.
• Fit computational models to test whether sentiment prediction errors predict happiness changes.
• Identify whether discussing negative topics (e.g., depression, guilt) modulates happiness effects.
• Analyze the role of the AI's positivity bias in sustaining user happiness.

实验结果

研究问题

• RQ1 Do AI chatbot conversations increase momentary happiness compared to journaling on equivalent topics?
• RQ2 How does topic negativity influence the happiness effect of AI conversations?
• RQ3 What role do sentiment alignment and prediction errors play in post-conversation happiness?
• RQ4 Does the AI's positivity bias contribute to increased happiness during negative-topic discussions?
• RQ5 Can computational models predict post-conversation happiness from sentiment dynamics over a conversation?

主要发现

• AI chatbot conversations yield higher happiness after the interaction than journaling, with stronger effects for negative topics.
• AI mirrors user sentiment but maintains a positivity bias.
• Users shift their sentiment to align with the AI's positivity during negative-topic discussions, contributing to higher happiness.
• A history of sentiment prediction errors during the conversation predicts greater post-conversation happiness.