Algorithms and especially recommendation algorithms play an important role online, most notably on YouTube. Yet, little is known about the network communities that these algorithms form. We analyzed the channel recommendations on YouTube to map the communities that the social network is creating through its algorithms and to test the network for homophily, that is, the connectedness between communities. We find that YouTube’s channel recommendation algorithm fosters the creation of highly homophilous communities in the United States (n = 13,529 channels) and in Germany (n = 8,000 channels). Factors that seem to drive YouTube’s recommendations are topics, language, and location. We highlight the issue of homophilous communities in the context of politics where YouTube’s algorithms create far-right communities in both countries.

算法，尤其是推荐算法，在网上（尤其是在YouTube上）起着重要的作用。然而，对于这些算法形成的网络社区知之甚少。我们分析了YouTube上的频道推荐，以映射社交网络通过其算法创建的社区，并测试网络的同质性，即社区之间的连通性。我们发现YouTube的频道推荐算法可促进在美国（n  = 13,529个频道）和德国（n = 8,000个频道）。似乎会影响YouTube建议的因素包括主题，语言和位置。在政治背景下，我们着重强调了同性恋社区的问题，YouTube的算法在这两个国家创建了最右翼的社区。