Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

From Personalization to Prejudice: Bias and Discrimination in Memory-Enhanced AI Agents for Recruitment

Himanshu Gharat, Himanshi Agrawal, Gourab K. Patro

TL;DR

This study investigates how memory-enhanced personalization in AI agents can introduce and amplify bias in high-stakes recruitment tasks. Using a memory-enabled agent and the Bias in Bios dataset, it analyzes bias emergence across pre-retrieval, retrieval, and re-ranking stages, linking memory patterns to discriminatory outcomes. The findings show that personalization amplifies gender-related biases, especially during re-ranking, and that scrubbing explicit gender cues only partially mitigates this effect, highlighting gaps in current safeguards. The work argues for robust guardrails and targeted mitigation strategies to preserve personalization benefits while preventing discrimination in agentic systems. It further suggests extending the analysis to other domains and multi-turn interactions.

Abstract

Large Language Models (LLMs) have empowered AI agents with advanced capabilities for understanding, reasoning, and interacting across diverse tasks. The addition of memory further enhances them by enabling continuity across interactions, learning from past experiences, and improving the relevance of actions and responses over time; termed as memory-enhanced personalization. Although such personalization through memory offers clear benefits, it also introduces risks of bias. While several previous studies have highlighted bias in ML and LLMs, bias due to memory-enhanced personalized agents is largely unexplored. Using recruitment as an example use case, we simulate the behavior of a memory-enhanced personalized agent, and study whether and how bias is introduced and amplified in and across various stages of operation. Our experiments on agents using safety-trained LLMs reveal that bias is systematically introduced and reinforced through personalization, emphasizing the need for additional protective measures or agent guardrails in memory-enhanced LLM-based AI agents.

From Personalization to Prejudice: Bias and Discrimination in Memory-Enhanced AI Agents for Recruitment

TL;DR

This study investigates how memory-enhanced personalization in AI agents can introduce and amplify bias in high-stakes recruitment tasks. Using a memory-enabled agent and the Bias in Bios dataset, it analyzes bias emergence across pre-retrieval, retrieval, and re-ranking stages, linking memory patterns to discriminatory outcomes. The findings show that personalization amplifies gender-related biases, especially during re-ranking, and that scrubbing explicit gender cues only partially mitigates this effect, highlighting gaps in current safeguards. The work argues for robust guardrails and targeted mitigation strategies to preserve personalization benefits while preventing discrimination in agentic systems. It further suggests extending the analysis to other domains and multi-turn interactions.

Abstract

Large Language Models (LLMs) have empowered AI agents with advanced capabilities for understanding, reasoning, and interacting across diverse tasks. The addition of memory further enhances them by enabling continuity across interactions, learning from past experiences, and improving the relevance of actions and responses over time; termed as memory-enhanced personalization. Although such personalization through memory offers clear benefits, it also introduces risks of bias. While several previous studies have highlighted bias in ML and LLMs, bias due to memory-enhanced personalized agents is largely unexplored. Using recruitment as an example use case, we simulate the behavior of a memory-enhanced personalized agent, and study whether and how bias is introduced and amplified in and across various stages of operation. Our experiments on agents using safety-trained LLMs reveal that bias is systematically introduced and reinforced through personalization, emphasizing the need for additional protective measures or agent guardrails in memory-enhanced LLM-based AI agents.

Paper Structure

This paper contains 15 sections, 1 figure, 2 tables.

Table of Contents

  1. Introduction
  2. Personalized Agents: Design and Operation
  3. Recruitment Agent Experiments
  4. Experimental Settings
  5. Dataset:
  6. Agent Configuration:
  7. Experiments:
  8. Evaluation Metrics:
  9. Experimental Results
  10. Assessing Personalization: Utility Gains vs. Bias Risks
  11. Bias in Retrieval and Re-Ranking Stages:
  12. Bias Amplification in Re-Ranking Stage:
  13. Do Re-ranking Adjustments Reflect Merit or Gender Bias?
  14. Bias in Pre-Retrieval Stage:
  15. Discussion

Figures (1)

  • Figure 1: Diagram illustrating how bias emerges and amplifies in a memory-enhanced personalized recruitment agent.