\n \n \n
\n
\n\n \n \n \n \n \n \n Estimation of fair ranking metrics with incomplete judgments.\n \n \n \n \n\n\n \n Kırnap, Ö.; Diaz, F.; Biega, A.; Ekstrand, M.; Carterette, B.; and Yilmaz, E.\n\n\n \n\n\n\n In
WWW '21, pages 1065–1075, New York, NY, USA, April 2021. Association for Computing Machinery\n
Journal Abbreviation: WWW '21\n\n
\n\n
\n\n
\n\n \n \n Paper\n \n \n\n \n \n doi\n \n \n\n \n link\n \n \n\n bibtex\n \n\n \n \n \n abstract \n \n\n \n\n \n \n \n \n \n \n \n\n \n \n \n \n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{kirnap_estimation_2021,\n\taddress = {New York, NY, USA},\n\ttitle = {Estimation of fair ranking metrics with incomplete judgments},\n\turl = {https://arxiv.org/abs/2108.05152},\n\tdoi = {10.1145/3442381.3450080},\n\tabstract = {There is increasing attention to evaluating the fairness of search system\nranking decisions. These metrics often consider the membership of items to\nparticular groups, often identified using protected attributes such as\ngender or ethnicity. To date, these metrics typically assume the\navailability and completeness of protected attribute labels of items.\nHowever, the protected attributes of individuals are rarely present,\nlimiting the application of fair ranking metrics in large scale systems.\nIn order to address this problem, we propose a sampling strategy and\nestimation technique for four fair ranking metrics. We formulate a robust\nand unbiased estimator which can operate even with very limited number of\nlabeled items. We evaluate our approach using both simulated and real\nworld data. Our experimental results demonstrate that our method can\nestimate this family of fair ranking metrics and provides a robust,\nreliable alternative to exhaustive or random data annotation.},\n\turldate = {2021-04-27},\n\tbooktitle = {{WWW} '21},\n\tpublisher = {Association for Computing Machinery},\n\tauthor = {Kırnap, Ömer and Diaz, Fernando and Biega, Asia and Ekstrand, Michael and Carterette, Ben and Yilmaz, Emine},\n\tmonth = apr,\n\tyear = {2021},\n\tnote = {Journal Abbreviation: WWW '21},\n\tkeywords = {evaluation, information retrieval, fair ranking, fairness},\n\tpages = {1065--1075},\n}\n\n
\n
\n\n\n
\n There is increasing attention to evaluating the fairness of search system ranking decisions. These metrics often consider the membership of items to particular groups, often identified using protected attributes such as gender or ethnicity. To date, these metrics typically assume the availability and completeness of protected attribute labels of items. However, the protected attributes of individuals are rarely present, limiting the application of fair ranking metrics in large scale systems. In order to address this problem, we propose a sampling strategy and estimation technique for four fair ranking metrics. We formulate a robust and unbiased estimator which can operate even with very limited number of labeled items. We evaluate our approach using both simulated and real world data. Our experimental results demonstrate that our method can estimate this family of fair ranking metrics and provides a robust, reliable alternative to exhaustive or random data annotation.\n
\n\n\n
\n\n\n
\n
\n\n \n \n \n \n \n \n Exploring author gender in book rating and recommendation.\n \n \n \n \n\n\n \n Ekstrand, M. D; and Kluver, D.\n\n\n \n\n\n\n
User Modeling and User-Adapted Interaction, 31(3): 377–420. July 2021.\n
\n\n
\n\n
\n\n
\n\n \n \n Paper\n \n \n\n \n \n doi\n \n \n\n \n link\n \n \n\n bibtex\n \n\n \n \n \n abstract \n \n\n \n \n \n 12 downloads\n \n \n\n \n \n \n \n \n \n \n\n \n \n \n\n\n\n
\n
@article{ekstrand_exploring_2021,\n\ttitle = {Exploring author gender in book rating and recommendation},\n\tvolume = {31},\n\tissn = {0924-1868},\n\turl = {https://md.ekstrandom.net/pubs/bag-extended},\n\tdoi = {10.1007/s11257-020-09284-2},\n\tabstract = {Collaborative filtering algorithms find useful patterns in rating and consumption data and exploit these patterns to guide users to good items. Many of the patterns in rating datasets reflect important real-world differences between the various users and items in the data; other patterns may be irrelevant or possibly undesirable for social or ethical reasons, particularly if they reflect undesired discrimination, such as discrimination in publishing or purchasing against authors who are women or ethnic minorities. In this work, we examine the response of collaborative filtering recommender algorithms to the distribution of their input data with respect to a dimension of social concern, namely content creator gender. Using publicly-available book ratings data, we measure the distribution of the genders of the authors of books in user rating profiles and recommendation lists produced from this data. We find that common collaborative filtering algorithms differ in the gender distribution of their recommendation lists, and in the relationship of that output distribution to user profile distribution.},\n\tnumber = {3},\n\turldate = {2020-06-05},\n\tjournal = {User Modeling and User-Adapted Interaction},\n\tauthor = {Ekstrand, Michael D and Kluver, Daniel},\n\tmonth = jul,\n\tyear = {2021},\n\tpages = {377--420},\n}\n
\n
\n\n\n
\n Collaborative filtering algorithms find useful patterns in rating and consumption data and exploit these patterns to guide users to good items. Many of the patterns in rating datasets reflect important real-world differences between the various users and items in the data; other patterns may be irrelevant or possibly undesirable for social or ethical reasons, particularly if they reflect undesired discrimination, such as discrimination in publishing or purchasing against authors who are women or ethnic minorities. In this work, we examine the response of collaborative filtering recommender algorithms to the distribution of their input data with respect to a dimension of social concern, namely content creator gender. Using publicly-available book ratings data, we measure the distribution of the genders of the authors of books in user rating profiles and recommendation lists produced from this data. We find that common collaborative filtering algorithms differ in the gender distribution of their recommendation lists, and in the relationship of that output distribution to user profile distribution.\n
\n\n\n
\n\n\n\n\n\n