A Graph-based Approach for Mitigating Multi-sided Exposure Bias in Recommender Systems

@article{Mansoury2022AGA,
  title={A Graph-based Approach for Mitigating Multi-sided Exposure Bias in Recommender Systems},
  author={Masoud Mansoury and Himan Abdollahpouri and Mykola Pechenizkiy and Bamshad Mobasher and Robin D. Burke},
  journal={ArXiv},
  year={2022},
  volume={abs/2107.03415}
}
Fairness is a critical system-level objective in recommender systems that has been the subject of extensive recent research. A specific form of fairness is supplier exposure fairness, where the objective is to ensure equitable coverage of items across all suppliers in recommendations provided to users. This is especially important in multistakeholder recommendation scenarios where it may be important to optimize utilities not just for the end user but also for other stakeholders such as item… 
Unbiased Cascade Bandits: Mitigating Exposure Bias in Online Learning to Rank Recommendation
TLDR
A discounting factor is proposed and incorporated into these algorithms that controls the exposure of items at each time step and the experimental results show that the proposed method improves the exposure fairness for items and suppliers.

References

SHOWING 1-10 OF 50 REFERENCES
FairMatch: A Graph-based Approach for Improving Aggregate Diversity in Recommender Systems
TLDR
FairMatch is introduced, a general graph-based algorithm that works as a post-processing approach after recommendation generation for improving aggregate diversity and fair distribution of recommended items and can be adapted to other recommendation scenarios using different underlying definitions of fairness.
Addressing the Multistakeholder Impact of Popularity Bias in Recommendation Through Calibration
TLDR
This paper proposes the concept of popularity calibration which measures the match between the popularity distribution of items in a user's profile and that of the recommended items, and develops an algorithm that optimizes this metric and has a secondary effect of improving supplier fairness.
Fairness-Aware Ranking in Search & Recommendation Systems with Application to LinkedIn Talent Search
TLDR
This work presents a framework for quantifying and mitigating algorithmic bias in mechanisms designed for ranking individuals, typically used as part of web-scale search and recommendation systems, and is the first large-scale deployed framework for ensuring fairness in the hiring domain.
Challenging the Long Tail Recommendation
TLDR
Empirical experiments show that the proposed algorithms are effective to recommend long tail items and outperform state-of-the-art recommendation techniques.
Post Processing Recommender Systems for Diversity
TLDR
This work addresses the problem of increasing diversity in recom- mendation systems that are based on collaborative filtering that use past ratings to predict a rating quality for potential recommendations and defines a new flexible notion of diversity that allows a system designer to prescribe the number of recommendations each item should receive.
Maximizing Aggregate Recommendation Diversity: A Graph-Theoretic Approach
TLDR
The proposed graph-theoretic approach for maximizing aggregate recommendation diversity based on maximum flow or maximum bipartite matching computations demonstrates substantial improvements in both diversity and accuracy, as compared to the recommendation re-ranking approaches.
Multistakeholder recommendation with provider constraints
TLDR
A constraint-based integer programming optimization model is proposed, in which different sets of constraints are used to reflect the goals of the different stakeholders, so it can easily be added onto an existing recommendation system to make it multi-stakeholder aware.
Improving sales diversity by recommending users to items
TLDR
This work explores the inversion of the recommendation task as a means to enhance sales diversity - and indirectly novelty - by selecting which users an item should be recommended to instead of the other way around, and addresses the inverted task by inverting the rating matrix.
FairRec: Two-Sided Fairness for Personalized Recommendations in Two-Sided Platforms
TLDR
The proposed FairRec algorithm guarantees at least Maximin Share of exposure for most of the producers and Envy-Free up to One Good fairness for every customer and extensive evaluations show the effectiveness of FairRec in ensuring two-sided fairness while incurring a marginal loss in the overall recommendation quality.
Multistakeholder recommendation: Survey and research directions
TLDR
The multistakeholder perspective on recommendation is outlined, highlighting example research areas and discussing important issues, open questions, and prospective research directions.
...
1
2
3
4
5
...