Societal Biases in Retrieved Contents: Measurement Framework and Adversarial Mitigation for BERT Rankers

TL;DR

This paper introduces a new framework to measure societal biases in retrieval systems and proposes AdvBert, an adversarial model that reduces bias while maintaining relevance, validated on passage retrieval datasets.

Contribution

The work presents a novel fairness measurement framework and a bias mitigation method for deep ranking models, specifically adapting adversarial training for IR tasks.

Findings

Bias is prevalent in ranking models compared to ranker-agnostic baselines.

AdvBert significantly improves fairness of BERT rankers.

Fairness can be improved without sacrificing relevance.

Abstract

Societal biases resonate in the retrieved contents of information retrieval (IR) systems, resulting in reinforcing existing stereotypes. Approaching this issue requires established measures of fairness in respect to the representation of various social groups in retrieval results, as well as methods to mitigate such biases, particularly in the light of the advances in deep ranking models. In this work, we first provide a novel framework to measure the fairness in the retrieved text contents of ranking models. Introducing a ranker-agnostic measurement, the framework also enables the disentanglement of the effect on fairness of collection from that of rankers. To mitigate these biases, we propose AdvBert, a ranking model achieved by adapting adversarial bias mitigation for IR, which jointly learns to predict relevance and remove protected attributes. We conduct experiments on two passage retrieval collections (MSMARCO Passage Re-ranking and TREC Deep Learning 2019 Passage Re-ranking), which we extend by fairness annotations of a selected subset of queries regarding gender attributes. Our results on the MSMARCO benchmark show that, (1) all ranking models are less fair in comparison with ranker-agnostic baselines, and (2) the fairness of Bert rankers significantly improves when using the proposed AdvBert models. Lastly, we investigate the trade-off between fairness and utility, showing that we can maintain the significant improvements in fairness without any significant loss in utility.