Mitigating Bias in RAG: Controlling the Embedder

TL;DR

This paper investigates how biases from different components in retrieval augmented generation systems interact and demonstrates that controlling embedder bias effectively reduces overall system bias without sacrificing utility.

Contribution

It introduces the concept of bias conflict in RAG systems, characterizes it linearly across components, and shows how reverse-biasing the embedder mitigates overall bias.

Findings

Bias conflict can be modeled linearly among components.

Reverse-biasing the embedder reduces overall bias.

Controlling embedder bias maintains utility and fairness.

Abstract

In retrieval augmented generation (RAG) systems, each individual component -- the LLM, embedder, and corpus -- could introduce biases in the form of skews towards outputting certain perspectives or identities. In this work, we study the conflict between biases of each component and their relationship to the overall bias of the RAG system, which we call bias conflict. Examining both gender and political biases as case studies, we show that bias conflict can be characterized through a linear relationship among components despite its complexity in 6 different LLMs. Through comprehensive fine-tuning experiments creating 120 differently biased embedders, we demonstrate how to control bias while maintaining utility and reveal the importance of reverse-biasing the embedder to mitigate bias in the overall system. Additionally, we find that LLMs and tasks exhibit varying sensitivities to the embedder bias, a crucial factor to consider for debiasing. Our results underscore that a fair RAG system can be better achieved by carefully controlling the bias of the embedder rather than increasing its fairness.