The lonely algorithm problem: the relationship between algorithmic personalization and social connectedness on TikTok
通过两项预注册研究（在线问卷和两周日记调查），发现TikTok的推荐算法推送频繁且正面的身份相关内容，能提升用户感知的算法响应性，进而增强社会连接感，但算法不敏感性无此作用。
Abstract Many major social media companies have claimed that their social feed algorithms were designed to promote relational well-being and enhance social connection. This project tested part of these claims by investigating if algorithmic personalization predicted social connectedness on TikTok, where connecting with identity-relevant media content is prioritized over interpersonal interaction. Drawing from the algorithm responsiveness process, we identified perceived algorithm responsiveness (PAR) and perceived algorithm insensitivity (PAI) as perceptual mechanisms of how algorithmic personalization relates to social connectedness. In two preregistered studies, an online survey and a 2-week daily diary survey, results suggested that the TikTok For You algorithm curating frequent and positively valenced identity-relevant content was associated with higher PAR and less PAI. In turn, higher levels of PAR predicted more social connectedness, but PAI did not. These findings highlight the sociotechnical dynamics of algorithms, suggesting human–algorithm interaction predicted social connectedness not only an algorithm’s code.