ADAPT: Fairness & diversity for sequential group recommendations

In group recommendation systems, achieving a balance between fairness and diversity is a challenging yet crucial task, particularly in sequential settings where preferences evolve over multiple iterations. This paper introduces ADAPT, a novel framework designed to optimize fairness and diversity in sequential group recommendations. ADAPT employs two novel aggregation methods, FaDJO and DiGSFO, to equitably meet group members’ needs while promoting diverse content. In addition to the novel aggregation methods ADAPT introduces a novel definition for the inter-round diversity based on item-lists embeddings. Experimental results on two real datasets and different group formation demonstrate ADAPT’s ability to optimize user satisfaction, fairness, and diversity, outperforming baseline methods in two different metrics (f-score and NDCG) and highlighting the importance of balancing these critical factors in sequential group settings. • The ADAPT framework is presented, enabling sequential group recommendations with guaranteed satisfaction, intra-group fairness, and diversity across iterations. • A new metric for inter-round diversity is introduced, tailored specifically to the requirements of sequential recommendation systems. • Two novel aggregation techniques are proposed and implemented to prioritize diversity in sequential group recommendation processes while maintaining a balance with fairness. • Comprehensive experimental evaluation is performed using two real-world datasets, demonstrating ADAPT’s adaptability and performance in varied scenarios. • A detailed analysis is provided to explore the trade-offs between diversity, fairness, and user satisfaction, highlighting their interplay in group recommendation settings.