Bias-Aware Agent: Enhancing Fairness in AI-Driven Knowledge Retrieval

TL;DR

The paper tackles bias in AI-driven knowledge retrieval by addressing fairness in dynamically retrieved content via large language models. It introduces the Bias-Aware Agent, a modular framework built on a ReAct-based agent that integrates a retriever (vector store) and a bias-detection tool (Dbias) to evaluate content before summarization, enabling transparency and responsible AI. The approach is evaluated on a curated news corpus with 40 queries, reporting a weighted $F1$-Score of $0.795$ for bias detection and revealing how bias signals influence final responses. This work advances fairer retrieval systems by making bias analysis an explicit, pluggable part of the retrieval-and-generation loop, with future directions toward bias mitigation and multi-agent collaboration to further reduce unfair representations.

Abstract

Advancements in retrieving accessible information have evolved faster in the last few years compared to the decades since the internet's creation. Search engines, like Google, have been the number one way to find relevant data. They have always relied on the user's abilities to find the best information in its billions of links and sources at everybody's fingertips. The advent of large language models (LLMs) has completely transformed the field of information retrieval. The LLMs excel not only at retrieving relevant knowledge but also at summarizing it effectively, making information more accessible and consumable for users. On top of it, the rise of AI Agents has introduced another aspect to information retrieval i.e. dynamic information retrieval which enables the integration of real-time data such as weather forecasts, and financial data with the knowledge base to curate context-aware knowledge. However, despite these advancements the agents remain susceptible to issues of bias and fairness, challenges deeply rooted within the knowledge base and training of LLMs. This study introduces a novel approach to bias-aware knowledge retrieval by leveraging agentic framework and the innovative use of bias detectors as tools to identify and highlight inherent biases in the retrieved content. By empowering users with transparency and awareness, this approach aims to foster more equitable information systems and promote the development of responsible AI.

Figures (6)

• Figure 1: High Level Architecture of Bias Aware Agent. The user submits a query to the agent. The agent processes and reasons on the query, retrieves relevant news chunks from the Vector Store, and ranks them based on relevance. The top-k chunk vectors are passed to the Bias Detection Tool, which analyzes potential biases in the content. Using the output of the Bias Detection Tool, the agent reasons and summarizes the retrieved news while appending a Bias Analysis to the final answer provided to the user. This ensures both relevance and fairness in the response.
• Figure 2: System prompt for the reasoner of the Bias Aware Agent.
• Figure 3: Reasoning loop of Bias Aware Agent which shows how the reasoner can interact with the tools available to the agent and iteratively does the bias analysis over the content retrieved from a vector store.
• Figure 4: Confusion Matrix showing the performance of the bias detection ability of the agent. Values shown are from queries resulting in analysis of 1 article used a source for a response to a query.
• Figure 5: Graph showcasing the correlation between Bias Probability (Confidence) and the correlating Predicted/Actual Result. The Trendline in the graph is auto generated from the data points