Homophily - the attraction of similarity - profoundly influences social interactions, affecting associations, information disclosure, and the dynamics of social exchanges. Organizational studies reveal that when professional and personal boundaries overlap, individuals from minority backgrounds often encounter a dilemma between authenticity and inclusion due to these homophily-driven dynamics: if they disclose their genuine interests, they risk exclusion from the broader conversation. Conversely, to gain inclusion, they might feel pressured to assimilate. How might the nature and design of social media platforms, where different conversational contexts frequently collapse, and the recommender algorithms that are at the heart of these platforms, which can prioritize content based on network structure and historical user engagement, impact these dynamics? In this paper, we employ agent-based simulations to investigate this question. Our findings indicate a decline in the visibility of professional content generated by minority groups, a trend that is exacerbated over time by recommendation algorithms. Within these minority communities, users who closely resemble the majority group tend to receive greater visibility. We examine the philosophical and design implications of our results, discussing their relevance to questions of informational justice, inclusion, and the epistemic benefits of diversity.

翻译：同质性——即相似性产生的吸引力——深刻影响着社会互动，包括人际关联、信息表露以及社会交换的动态过程。组织研究表明，当专业与个人边界发生重叠时，来自少数群体的个体常因这种同质性驱动机制而陷入真实性与包容性之间的两难困境：若表露真实兴趣，可能面临被主流对话排斥的风险；反之，为获得包容，又可能被迫同化。社交媒体平台中不同对话语境频繁交融的特性及其核心推荐算法（能依据网络结构和历史用户参与度优化内容排序），将如何影响这种动态机制？本文采用基于智能体的模拟方法对此展开研究。研究发现：少数群体生成的专业内容可见度呈下降趋势，且推荐算法会随时间推移加剧这一现象；在少数群体内部，与主流群体特征高度相似的用户往往获得更高可见度。我们进一步探讨了研究结果在哲学与设计层面的启示，并论证其与信息正义、包容性及多样性认知效益等议题的关联。