Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Your AI Bosses Are Still Prejudiced: The Emergence of Stereotypes in LLM-Based Multi-Agent Systems

Jingyu Guo, Yingying Xu

TL;DR

It is suggested that stereotype formation in AI systems may arise as an emergent property of multi-agent interactions, rather than merely from training data biases.

Abstract

While stereotypes are well-documented in human social interactions, AI systems are often presumed to be less susceptible to such biases. Previous studies have focused on biases inherited from training data, but whether stereotypes can emerge spontaneously in AI agent interactions merits further exploration. Through a novel experimental framework simulating workplace interactions with neutral initial conditions, we investigate the emergence and evolution of stereotypes in LLM-based multi-agent systems. Our findings reveal that (1) LLM-Based AI agents develop stereotype-driven biases in their interactions despite beginning without predefined biases; (2) stereotype effects intensify with increased interaction rounds and decision-making power, particularly after introducing hierarchical structures; (3) these systems exhibit group effects analogous to human social behavior, including halo effects, confirmation bias, and role congruity; and (4) these stereotype patterns manifest consistently across different LLM architectures. Through comprehensive quantitative analysis, these findings suggest that stereotype formation in AI systems may arise as an emergent property of multi-agent interactions, rather than merely from training data biases. Our work underscores the need for future research to explore the underlying mechanisms of this phenomenon and develop strategies to mitigate its ethical impacts.

Your AI Bosses Are Still Prejudiced: The Emergence of Stereotypes in LLM-Based Multi-Agent Systems

TL;DR

It is suggested that stereotype formation in AI systems may arise as an emergent property of multi-agent interactions, rather than merely from training data biases.

Abstract

While stereotypes are well-documented in human social interactions, AI systems are often presumed to be less susceptible to such biases. Previous studies have focused on biases inherited from training data, but whether stereotypes can emerge spontaneously in AI agent interactions merits further exploration. Through a novel experimental framework simulating workplace interactions with neutral initial conditions, we investigate the emergence and evolution of stereotypes in LLM-based multi-agent systems. Our findings reveal that (1) LLM-Based AI agents develop stereotype-driven biases in their interactions despite beginning without predefined biases; (2) stereotype effects intensify with increased interaction rounds and decision-making power, particularly after introducing hierarchical structures; (3) these systems exhibit group effects analogous to human social behavior, including halo effects, confirmation bias, and role congruity; and (4) these stereotype patterns manifest consistently across different LLM architectures. Through comprehensive quantitative analysis, these findings suggest that stereotype formation in AI systems may arise as an emergent property of multi-agent interactions, rather than merely from training data biases. Our work underscores the need for future research to explore the underlying mechanisms of this phenomenon and develop strategies to mitigate its ethical impacts.

Paper Structure

This paper contains 31 sections, 4 equations, 6 figures, 1 table.

Table of Contents

  1. Introduction
  2. Workplace Discrimination and Stereotypes
  3. Theoretical Framework of Stereotypes
  4. Stereotypical Behaviors in LLM
  5. LLM-based Agents and Multi-agent Systems
  6. Research Findings
  7. Methods and Experimental Design
  8. Multi-Agent Simulation Environment
  9. Task Execution and Interaction
  10. Hierarchical Extension
  11. Evaluation
  12. Experiments Setup
  13. Job Classification
  14. Bias Control
  15. Ablation Study on Bias Control
  16. ...and 16 more sections

Figures (6)

  • Figure 1: Experimental procedure: (1) Each agent acts as an individual. (2) They complete a randomly assigned task in each stage. (3) Everyone engages in conversation with others. (4) An agent was added as the boss, assigning tasks instead of possibility distribution function. (5) Agents evaluate each other.
  • Figure 2: An example of interactions between agents.
  • Figure 3: Person-job stereotype formation comparison: single experiment run vs all runs vs all runs with background condition
  • Figure 4: Meta-Analysis of Stereotype Formation.
  • Figure 5: Random task assignment vs AI boss assignment.
  • ...and 1 more figures