Mainstream bias in recommender systems is assessed through the NDCG@K metric, which averages utility across users, masking differences in individual user performance.
The debiasing objective is guided by the Rawlsian Max-Min fairness principle, focusing on maximizing utility for niche users while maintaining or improving mainstream users' performance.
To counteract mainstream bias, the study emphasizes improving recommendations for niche users without sacrificing the utility of the mainstream groups, ensuring a fair distribution of recommendations.
[
Collection
]
[
|
...
]