Ethics of algorithmic invasiveness in beauty apps: an online experimental survey of public perspectives

The rise of beauty apps with artificial intelligence (AI) capabilities has transformed how users perceive, evaluate, and engage with their and others’ appearance. While such apps offer entertainment, concerns persist about their ethical implications and psychological effects. In this study, we focus on the concept of algorithmic invasiveness – defined as the tendency of AI algorithms to increasingly influence and integrate into our daily lives. Using a conjoint survey experiment, we empirically assess public attitudes and social acceptability of beauty app functionalities with varying degrees of invasiveness, from removing blemishes to collecting medical data. Results show that functionalities with high degree of algorithmic invasiveness, such as AI-generated attractiveness rankings, are viewed as significantly less socially acceptable than less invasive functionalities. However, acceptability depends on context: functionalities tied to commercial applications, like personalised product recommendations, are often deemed less problematic despite similar data-use practices. User demographics and attitudes further shape perceptions. Older adults display greater acceptance of highly invasive functionalities compared to younger users. Participants with higher self-objectification or positive views of AI show increased tolerance of invasive functionalities. These findings highlight how both technological design and user predisposition influence both ethical attitudes and preferences in beauty-related technologies. This study showcases how we can elicit moral intuitions about new technologies by exploring how the public form preferences around specific functionalities. Moreover, it provides novel insight about how individuals make normative distinction between degrees of invasiveness, contributing to broader discourse about responsible AI design and digital well-being.具备人工智能（Artificial Intelligence，AI）功能的美妆类应用兴起，彻底改变了用户感知、评估自身与他人外貌，并与之互动的方式。尽管此类应用可提供娱乐体验，但人们仍对其伦理影响与心理效应持续存在担忧。本研究聚焦于算法侵入性（algorithmic invasiveness）这一概念——其被定义为人工智能算法日益渗透并融入日常生活的趋势。研究采用联合分析调查实验（conjoint survey experiment）法，从实证层面评估公众对不同侵入程度的美妆应用功能的态度与社会接受度，功能覆盖范围从去除面部瑕疵直至收集医疗数据等场景。结果显示，算法侵入程度较高的功能（如AI生成的外貌吸引力排名），相较低侵入性功能，被认为社会接受度显著更低。但社会接受度会因场景而异：与商业应用绑定的功能（如个性化商品推荐），即便采用了类似的数据使用模式，通常也被认为争议性更低。用户的人口统计学特征与自身态度会进一步塑造其认知判断。与年轻用户相比，老年群体对高侵入性功能的接受度更高。自我客体化（self-objectification）程度较高，或对AI持正面态度的参与者，对侵入性功能的容忍度也更高。上述研究结果揭示，技术设计与用户固有倾向，均会影响美妆相关技术领域的伦理态度与偏好选择。本研究阐明了，通过探究公众针对特定功能的偏好形成机制，我们可如何激发公众对新技术的道德直觉判断。此外，本研究还为个体如何对侵入性程度做出规范性区分提供了全新视角，为围绕负责任人工智能设计与数字福祉的广泛学术讨论贡献了有益见解。