The Impact of Ideological Discourses in RAG: A Case Study with COVID-19 Treatments

TL;DR

This study investigates how ideological texts retrieved during RAG influence LLM outputs, revealing that external ideological sources significantly shape model responses and emphasizing the need to address ideological bias and manipulation risks.

Contribution

It introduces a corpus linguistics framework using LMDA to identify ideologies in RAG, and demonstrates how retrieved ideological texts affect LLM responses in the context of COVID-19 treatments.

Findings

LLMs' responses align more with external ideological texts.

Enhanced prompts increase ideological influence on responses.

Highlighting risks of ideological bias and manipulation in RAG.

Abstract

This paper studies the impact of retrieved ideological texts on the outputs of large language models (LLMs). While interest in understanding ideology in LLMs has recently increased, little attention has been given to this issue in the context of Retrieval-Augmented Generation (RAG). To fill this gap, we design an external knowledge source based on ideological loaded texts about COVID-19 treatments. Our corpus is based on 1,117 academic articles representing discourses about controversial and endorsed treatments for the disease. We propose a corpus linguistics framework, based on Lexical Multidimensional Analysis (LMDA), to identify the ideologies within the corpus. LLMs are tasked to answer questions derived from three identified ideological dimensions, and two types of contextual prompts are adopted: the first comprises the user question and ideological texts; and the second contains the question, ideological texts, and LMDA descriptions. Ideological alignment between reference ideological texts and LLMs' responses is assessed using cosine similarity for lexical and semantic representations. Results demonstrate that LLMs' responses based on ideological retrieved texts are more aligned with the ideology encountered in the external knowledge, with the enhanced prompt further influencing LLMs' outputs. Our findings highlight the importance of identifying ideological discourses within the RAG framework in order to mitigate not just unintended ideological bias, but also the risks of malicious manipulation of such models.