Seminal studies on social influence (Asch, 1956; Sherif, 1935) were based on face-to-face interactions between humans. Nowadays, computer-mediated communication is steadily becoming ubiquitous, and we are increasingly influenced by non-human agents, such as algorithms, robots, and chatbots. The present research aimed to answer two important questions: Can non-human agents exert social influence in a persistent manner and, thus, contribute to the emergence of social norms? And if this is the case, is social influence exerted by non-human agents mediated by the same or by different cognitive mechanisms as social influence exerted by human agents? To answer these questions, we used an online version of an established paradigm in research on social norm learning. To examine the cognitive underpinnings of social influence, we used a diffusion model approach. Our results provide strong evidence for the notion that non-human agents can induce persistent social influence outside an immediate group context and, hence, can contribute to the emergence of social norms. Furthermore, results from our diffusion model analyses support the notion that social influence exerted by non-human agents is mainly mediated by the same cognitive mechanisms as social influence exerted by human agents.

关于社会影响的开创性研究（Asch，1956；Sherif，1935）基于人与人之间的面对面互动。如今，以计算机为媒介的通信正稳步变得无处不在，我们越来越受到算法、机器人和聊天机器人等非人类代理的影响。本研究旨在回答两个重要问题：非人类代理人能否以持久的方式施加社会影响，从而促进社会规范的出现？如果是这样，非人类主体施加的社会影响与人类主体施加的社会影响是由相同的认知机制还是由不同的认知机制介导的？为了回答这些问题，我们在社会规范学习研究中使用了已建立范式的在线版本。为了检验社会影响的认知基础，我们使用了扩散模型方法。我们的结果为以下观点提供了强有力的证据，即非人类代理人可以在直接的群体环境之外引起持续的社会影响，因此可以促进社会规范的出现。此外，我们的扩散模型分析结果支持这样一种观点，即非人类主体施加的社会影响主要由与人类主体施加的社会影响相同的认知机制介导。