Scholarlink

<p>Group recommender systems (GRSs) generate recommendations for groups of individuals, making them useful in decision-making scenarios such as friends agreeing on a restaurant, colleagues selecting a team-building activity, or a family choosing a movie for a movie night. Unlike single-user recommender systems, GRS must balance multiple, often conflicting user preferences. A key challenge is ensuring fairness so that no user’s preferences are consistently overlooked. Due to the heterogeneous nature of group members’ preferences, achieving fairness within a single recommendation iteration—sometimes also referred to as a “session”—is not always feasible. However, group decision-making often occurs over a sequence of iterations, providing opportunities to balance fairness over time. In this work, we evaluate existing GRS approaches, applying them in the context of sequential group recommendation, where traditional GRS methods are used across sequence of multiple iterations. Our evaluation also incorporates simulated group choices to better reflect real-world usage scenarios. At the same time, we propose a novel proportionality-preserving approach that improves long-term fairness across iterations while maintaining overall group satisfaction. The source code and additional data are available in the public repository at: <a href="https://osf.io/m8z57">https://osf.io/m8z57</a>.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Long-term fairness in sequential group recommendations

  • 期刊论文

  • 发表日期：

    2026年1月5日

查看全文
  • Patrik Dokoupil,
  • Ladislav Peska

摘要

Group recommender systems (GRSs) generate recommendations for groups of individuals, making them useful in decision-making scenarios such as friends agreeing on a restaurant, colleagues selecting a team-building activity, or a family choosing a movie for a movie night. Unlike single-user recommender systems, GRS must balance multiple, often conflicting user preferences. A key challenge is ensuring fairness so that no user’s preferences are consistently overlooked. Due to the heterogeneous nature of group members’ preferences, achieving fairness within a single recommendation iteration—sometimes also referred to as a “session”—is not always feasible. However, group decision-making often occurs over a sequence of iterations, providing opportunities to balance fairness over time. In this work, we evaluate existing GRS approaches, applying them in the context of sequential group recommendation, where traditional GRS methods are used across sequence of multiple iterations. Our evaluation also incorporates simulated group choices to better reflect real-world usage scenarios. At the same time, we propose a novel proportionality-preserving approach that improves long-term fairness across iterations while maintaining overall group satisfaction. The source code and additional data are available in the public repository at: https://osf.io/m8z57.