电商平台智能客服用户负面态度何以形成——基于行为现实主义的归因研究

How Shapes Negative User Attitudes toward Intelligent Customer Service in E-Commerce Platform—A Attribution Study Based on Behavioral Realism

摘要: 在电商消费场景中，智能客服已实现规模化应用，但用户对其普遍持有负面态度。现有研究多聚焦形式现实主义特征，忽视行为现实主义视角下的功能属性影响机制。本文基于行为现实主义，以微博、抖音平台的用户评论为数据源，采用BERTopic主题建模识别电商平台智能客服的核心属性，结合RoBERTa属性级情感分析量化用户情感极性，通过回归分析检验各属性对负面态度的影响。人机协同转接障碍显著加剧用户负面态度，响应速度、服务可及性与智能水平则显著抑制负面态度，职业替代焦虑与问题解决成本影响不显著。本文创新性地将行为现实主义引入电商平台智能客服研究，融合深度学习实现细粒度情感分析，但存在数据源覆盖局限、未充分考虑用户特征与服务情境的调节作用等问题，未来需结合多源数据与纵向设计深化研究。

Abstract: In the context of e-commerce consumption, intelligent customer service has been widely implemented, yet users generally hold a negative attitude toward it. Existing studies mostly focus on formal realism characteristics, neglecting the influence mechanism of functional attributes from the perspective of behavioral realism. Based on behavioral realism, this study takes user comments from Weibo and Douyin platforms as data sources. It adopts BERTopic modeling to identify the core attributes of intelligent customer service in e-commerce platforms, combines RoBERTa-based aspect-level sentiment analysis to quantify user sentiment polarity, and employs regression analysis to test the impact of each attribute on negative attitudes. Human-machine collaborative transfer barriers significantly exacerbate users’ negative attitudes, while response speed, service accessibility, and intelligence level significantly inhibit negative attitudes. The impacts of occupational displacement anxiety and problem-solving costs are not significant. This study innovatively introduces behavioral realism into research on intelligent customer service in e-commerce platforms and integrates deep learning to achieve fine-grained sentiment analysis. However, it has limitations such as restricted data source coverage and insufficient consideration of the moderating effects of user characteristics and service scenarios. Future research should be deepened by combining multi-source data and longitudinal design.