FEIR : quantifying and reducing envy and inferiority for fair recommendation of limited resources
- Author
- Nan Li (UGent) , Bo Kang (UGent) , Jefrey Lijffijt (UGent) and Tijl De Bie (UGent)
- Organization
- Project
- Abstract
- In settings such as e-recruitment and online dating, recommendation involves distributing limited opportunities, calling for novel approaches to quantify and enforce fairness. We introduce inferiority, a novel (un)fairness measure quantifying a user’s competitive disadvantage for their recommended items. Inferiority complements envy, a fairness notion measuring preference for others’ recommendations. We combine inferiority and envy with utility, an accuracy-related measure of aggregated relevancy scores. Since these measures are non-differentiable, we reformulate them using a probabilistic interpretation of recommender systems, yielding differentiable versions. We combine these loss functions in a multi-objective optimization problem called FEIR (Fairness through Envy and Inferiority Reduction), applied as post-processing for standard recommender systems. Experiments on synthetic and real-world data demonstrate that our approach improves trade-offs between inferiority, envy, and utility compared to naive recommendations and the baseline methods.
Downloads
-
RecSysHR2023-paper 6.pdf
- full text (Published version)
- |
- open access
- |
- |
- 4.38 MB
Citation
Please use this url to cite or link to this publication: http://hdl.handle.net/1854/LU-01HEZ1F38MRECYGMENTCMM1P39
- MLA
- Li, Nan, et al. “FEIR : Quantifying and Reducing Envy and Inferiority for Fair Recommendation of Limited Resources.” RECSYS in HR 2023 : The 3rd Workshop on Recommender Systems for Human Resources (RecSys in HR 2023), Proceedings, edited by Mesut Kaya et al., vol. 3490, CEUR, 2023.
- APA
- Li, N., Kang, B., Lijffijt, J., & De Bie, T. (2023). FEIR : quantifying and reducing envy and inferiority for fair recommendation of limited resources. In M. Kaya, T. Bogers, D. Graus, C. Johnson, & J.-J. Decorte (Eds.), RECSYS in HR 2023 : the 3rd Workshop on Recommender Systems for Human Resources (RecSys in HR 2023), Proceedings (Vol. 3490). CEUR.
- Chicago author-date
- Li, Nan, Bo Kang, Jefrey Lijffijt, and Tijl De Bie. 2023. “FEIR : Quantifying and Reducing Envy and Inferiority for Fair Recommendation of Limited Resources.” In RECSYS in HR 2023 : The 3rd Workshop on Recommender Systems for Human Resources (RecSys in HR 2023), Proceedings, edited by Mesut Kaya, Toine Bogers, David Graus, Chris Johnson, and Jens-Joris Decorte. Vol. 3490. CEUR.
- Chicago author-date (all authors)
- Li, Nan, Bo Kang, Jefrey Lijffijt, and Tijl De Bie. 2023. “FEIR : Quantifying and Reducing Envy and Inferiority for Fair Recommendation of Limited Resources.” In RECSYS in HR 2023 : The 3rd Workshop on Recommender Systems for Human Resources (RecSys in HR 2023), Proceedings, ed by. Mesut Kaya, Toine Bogers, David Graus, Chris Johnson, and Jens-Joris Decorte. Vol. 3490. CEUR.
- Vancouver
- 1.Li N, Kang B, Lijffijt J, De Bie T. FEIR : quantifying and reducing envy and inferiority for fair recommendation of limited resources. In: Kaya M, Bogers T, Graus D, Johnson C, Decorte J-J, editors. RECSYS in HR 2023 : the 3rd Workshop on Recommender Systems for Human Resources (RecSys in HR 2023), Proceedings. CEUR; 2023.
- IEEE
- [1]N. Li, B. Kang, J. Lijffijt, and T. De Bie, “FEIR : quantifying and reducing envy and inferiority for fair recommendation of limited resources,” in RECSYS in HR 2023 : the 3rd Workshop on Recommender Systems for Human Resources (RecSys in HR 2023), Proceedings, Singapore, Singapore, 2023, vol. 3490.
@inproceedings{01HEZ1F38MRECYGMENTCMM1P39,
abstract = {{In settings such as e-recruitment and online dating, recommendation involves distributing limited opportunities, calling for novel approaches to quantify and enforce fairness. We introduce inferiority, a novel (un)fairness measure quantifying a user’s competitive disadvantage for their recommended items. Inferiority complements envy, a fairness notion measuring preference for others’ recommendations. We combine inferiority and envy with utility, an accuracy-related measure of aggregated relevancy scores. Since these measures are non-differentiable, we reformulate them using a probabilistic interpretation of recommender systems, yielding differentiable versions. We combine these loss functions in a multi-objective optimization problem called FEIR (Fairness through Envy and Inferiority Reduction), applied as post-processing for standard recommender systems. Experiments on synthetic and real-world data demonstrate that our approach improves trade-offs between inferiority, envy, and utility compared to naive recommendations and the baseline methods.}},
articleno = {{6}},
author = {{Li, Nan and Kang, Bo and Lijffijt, Jefrey and De Bie, Tijl}},
booktitle = {{RECSYS in HR 2023 : the 3rd Workshop on Recommender Systems for Human Resources (RecSys in HR 2023), Proceedings}},
editor = {{Kaya, Mesut and Bogers, Toine and Graus, David and Johnson, Chris and Decorte, Jens-Joris}},
issn = {{1613-0073}},
language = {{eng}},
location = {{Singapore, Singapore}},
pages = {{10}},
publisher = {{CEUR}},
title = {{FEIR : quantifying and reducing envy and inferiority for fair recommendation of limited resources}},
volume = {{3490}},
year = {{2023}},
}