被理解的错觉：情感识别AI中的拟人化回应与传播误读机制研究

Illusion of Being Understood: Anthropomorphic Emotion Responses and Misinterpretation in AI–Human Interaction

摘要: 在人机交互中，用户的许多行为都透露着丰富而微妙的情感信息，通过与用户进行文本/语音对话，AI在与人的对话交互中考虑情感信息，使机器具备理解、回应人的情感的能力，由此用户与AI互动中形成“被理解”的心理幻觉，本研究聚焦于人工智能情感识别系统中的“被理解的错觉”现象，旨在揭示算法通过拟人化语言与共情模板如何在传播实践中建构用户的情感感知。研究采用情境模拟与跨平台文本分析方法，选取六类典型情绪场景(正向与负向各三类)，并对三个主流AI平台的回应进行编码与比较。结果表明，当前情感识别AI普遍通过情绪标签化反馈、模板化共情句式与第一人称拟人化表达，营造“情感在场感”，进而引导用户产生被理解感与媒介信任。然而，这种情感反馈模式易导致情感误读、依赖与责任归属模糊等风险。研究指出，未来算法情感设计需在亲和力与真实性之间取得平衡，避免过度情感拟态，并提出情绪生成机制透明化与用户媒介素养提升等治理建议。

Abstract: In human-computer interaction, users’ behaviors often convey rich and subtle emotional cues. Through text and voice dialogues, artificial intelligence incorporates emotional information during interactions, enabling machines to understand and respond to human emotions. This process can induce a psychological illusion of “being understood” in user-AI interactions. This study focuses on the phenomenon of the “illusion of being understood” within AI-based emotion recognition systems and seeks to reveal how algorithms, through anthropomorphic language and templated empathic expressions, construct users’ perception of emotional understanding in communicative practice. Employing scenario simulation and cross-platform textual analysis, six representative emotional contexts (three positive and three negative) were examined, and the responses from three major AI platforms were coded and compared. The results demonstrate that current affective AI systems commonly utilize emotion-labeling feedback, formulaic empathetic statements, and first-person anthropomorphic expressions to create a sense of “emotional presence”, thereby fostering perceived understanding and media trust. However, this mode of emotional feedback also entails risks, including emotional misinterpretation, user dependency, and blurred responsibility attribution. The study highlights the necessity of balancing sociability and authenticity in the future design of algorithmic affective systems, avoiding excessive emotional mimicry, and advocates for greater transparency in emotion-generation mechanisms as well as enhanced user media literacy.