基于AIDUA拓展模型下宫颈癌AI筛查接受意愿的实证研究

An Empirical Study on the Acceptance Intention of AI-Assisted Cervical Cancer Screening Based on an Extended AIDUA Model

摘要: 目的：随着宫颈癌AI筛查逐步普及，本研究基于人工智能设备接受模型(AIDUA)，从认知与情绪交互视角，探讨社会影响、享乐动机、拟人化与绩效期望、努力期望及消极情绪的链式路径之间的关联，以及这些路径与女性对宫颈癌AI筛查服务的使用意愿之间的关系。方法：采用便利与滚雪球抽样，线上面向18岁以上女性开展问卷调查，共回收有效问卷436份(有效回收率93.2%)。量表基于成熟工具改编并适配情境，通过结构方程模型进行路径分析与Bootstrap链式中介检验。结果：研究量表信效度与整体模型拟合效果良好，无严重共同方法偏差。社会影响、享乐动机、拟人化均与绩效期望与努力期望呈显著正相关。两类期望与消极情绪呈显著负相关，消极情绪进而显著负向影响使用意愿。全部九项假设均获支持。结论：AIDUA模型在宫颈癌AI筛查场景中具有良好适配性。社会认可、新奇体验与人性化交互设计与绩效及努力期望的提升相关，以及女性对AI的焦虑与不信任感的降低相关，从而与使用意愿的提升存在关联。研究提示，技术推广不应仅关注算法准确率，还需同步加强社会宣传、优化交互设计、重视初次使用的情绪体验。后续研究可扩大样本地域与教育层次覆盖，并开展多群组比较分析。

Abstract: Objective: As AI screening for cervical cancer becomes increasingly prevalent, this study draws on the Artificial Intelligence Device User Acceptance (AIDUA) model. From the perspective of cognitive-emotional interaction, it explores the serial pathways linking social influence, hedonic motivation, anthropomorphism, performance expectancy, effort expectancy, and negative emotions, as well as their relationships with women’s intention to use AI-based cervical cancer screening services. Methods: Using convenience and snowball sampling, an online survey was conducted among women aged 18 and above. A total of 436 valid questionnaires were collected (valid response rate: 93.2%). The measurement scales were adapted from established instruments and tailored to the specific context. Path analysis and bootstrap serial mediation tests were performed using structural equation modeling (SEM). Results: The reliability and validity of the scales, as well as the overall model fit, were satisfactory, with no severe common method bias. Social influence, hedonic motivation, and anthropomorphism were all significantly positively correlated with both performance and effort expectancies. Both types of expectancies were significantly negatively correlated with negative emotions, which in turn had a significant negative impact on the intention to use. All nine hypotheses were supported. Conclusion: The AIDUA model demonstrates good applicability in the context of AI screening for cervical cancer. Social approval, novel experiences, and humanized interaction design are associated with enhanced performance and effort expectancies, as well as reduced anxiety and distrust toward AI, thereby increasing the intention to use. These findings suggest that technology promotion should not focus solely on algorithmic accuracy; it must also be accompanied by strengthened public awareness campaigns, optimized interaction design, and careful attention to users’ emotional experiences during their first use. Future research should expand the geographic and educational coverage of the sample and conduct multi-group comparative analyses.