Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Gender Bias of LLM in Economics: An Existentialism Perspective

Hui Zhong, Songsheng Chen, Mian Liang

Abstract

Large Language Models (LLMs), such as GPT-4 and BERT, have rapidly gained traction in natural language processing (NLP) and are now integral to financial decision-making. However, their deployment introduces critical challenges, particularly in perpetuating gender biases that can distort decision-making outcomes in high-stakes economic environments. This paper investigates gender bias in LLMs through both mathematical proofs and empirical experiments using the Word Embedding Association Test (WEAT), demonstrating that LLMs inherently reinforce gender stereotypes even without explicit gender markers. By comparing the decision-making processes of humans and LLMs, we reveal fundamental differences: while humans can override biases through ethical reasoning and individualized understanding, LLMs maintain bias as a rational outcome of their mathematical optimization on biased data. Our analysis proves that bias in LLMs is not an unintended flaw but a systematic result of their rational processing, which tends to preserve and amplify existing societal biases encoded in training data. Drawing on existentialist theory, we argue that LLM-generated bias reflects entrenched societal structures and highlights the limitations of purely technical debiasing methods. This research underscores the need for new theoretical frameworks and interdisciplinary methodologies that address the ethical implications of integrating LLMs into economic and financial decision-making. We advocate for a reconceptualization of how LLMs influence economic decisions, emphasizing the importance of incorporating human-like ethical considerations into AI governance to ensure fairness and equity in AI-driven financial systems.

Gender Bias of LLM in Economics: An Existentialism Perspective

Abstract

Large Language Models (LLMs), such as GPT-4 and BERT, have rapidly gained traction in natural language processing (NLP) and are now integral to financial decision-making. However, their deployment introduces critical challenges, particularly in perpetuating gender biases that can distort decision-making outcomes in high-stakes economic environments. This paper investigates gender bias in LLMs through both mathematical proofs and empirical experiments using the Word Embedding Association Test (WEAT), demonstrating that LLMs inherently reinforce gender stereotypes even without explicit gender markers. By comparing the decision-making processes of humans and LLMs, we reveal fundamental differences: while humans can override biases through ethical reasoning and individualized understanding, LLMs maintain bias as a rational outcome of their mathematical optimization on biased data. Our analysis proves that bias in LLMs is not an unintended flaw but a systematic result of their rational processing, which tends to preserve and amplify existing societal biases encoded in training data. Drawing on existentialist theory, we argue that LLM-generated bias reflects entrenched societal structures and highlights the limitations of purely technical debiasing methods. This research underscores the need for new theoretical frameworks and interdisciplinary methodologies that address the ethical implications of integrating LLMs into economic and financial decision-making. We advocate for a reconceptualization of how LLMs influence economic decisions, emphasizing the importance of incorporating human-like ethical considerations into AI governance to ensure fairness and equity in AI-driven financial systems.

Paper Structure

This paper contains 13 sections, 14 equations, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Literature Review
  3. Methodology
  4. Mathematical Proof of Statistical Gender Bias in LLMs
  5. Comparison with Human Decision-making in Rational Process
  6. Experiment
  7. Experimental Principle
  8. Experimental Design and Methodology
  9. Results
  10. Analysis Using Existentialism
  11. Gender Bias in Terms of Reality and Truth
  12. Gender Bias in Terms of Intersubjectivity
  13. Conclusion

Figures (3)

  • Figure 1: Compare with human agent decision making process
  • Figure 2: shows the effect size values for the seven models. On the left side are the general-purpose models, with parameter sizes increasing from top to bottom. On the right side are the models fine-tuned for finance and economics tasks. The effect size for the LLaMAFin model in Chinese is 0.
  • Figure 3: displays the p-value values from the WEAT test. The left side represents the English test results, while the right side shows the Chinese test results. A median-fitting curve is applied to both sets of data.