Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Investigating the Impact of LLM Personality on Cognitive Bias Manifestation in Automated Decision-Making Tasks

Jiangen He, Jiqun Liu

TL;DR

This study investigates how Big Five personality traits shape cognitive biases in LLM-driven decision-making across multiple model architectures and bias types. It leverages two datasets, Student Admission and BiasEval, to induce personality prompts and measure eight biases, evaluating debiasing via a zero-shot awareness prompt and reversed-trait prompts. The findings reveal that six biases are prevalent while sunk cost and group attribution show minimal impact, with mitigation effectiveness highly dependent on both personality trait and model architecture; conscientiousness and agreeableness often enhance debiasing. Overall, the work highlights the necessity of architecture-aware, personality-informed mitigation strategies to improve fairness, reliability, and safety in AI-assisted decision-making tasks.

Abstract

Large Language Models (LLMs) are increasingly used in decision-making, yet their susceptibility to cognitive biases remains a pressing challenge. This study explores how personality traits influence these biases and evaluates the effectiveness of mitigation strategies across various model architectures. Our findings identify six prevalent cognitive biases, while the sunk cost and group attribution biases exhibit minimal impact. Personality traits play a crucial role in either amplifying or reducing biases, significantly affecting how LLMs respond to debiasing techniques. Notably, Conscientiousness and Agreeableness may generally enhance the efficacy of bias mitigation strategies, suggesting that LLMs exhibiting these traits are more receptive to corrective measures. These findings address the importance of personality-driven bias dynamics and highlight the need for targeted mitigation approaches to improve fairness and reliability in AI-assisted decision-making.

Investigating the Impact of LLM Personality on Cognitive Bias Manifestation in Automated Decision-Making Tasks

TL;DR

This study investigates how Big Five personality traits shape cognitive biases in LLM-driven decision-making across multiple model architectures and bias types. It leverages two datasets, Student Admission and BiasEval, to induce personality prompts and measure eight biases, evaluating debiasing via a zero-shot awareness prompt and reversed-trait prompts. The findings reveal that six biases are prevalent while sunk cost and group attribution show minimal impact, with mitigation effectiveness highly dependent on both personality trait and model architecture; conscientiousness and agreeableness often enhance debiasing. Overall, the work highlights the necessity of architecture-aware, personality-informed mitigation strategies to improve fairness, reliability, and safety in AI-assisted decision-making tasks.

Abstract

Large Language Models (LLMs) are increasingly used in decision-making, yet their susceptibility to cognitive biases remains a pressing challenge. This study explores how personality traits influence these biases and evaluates the effectiveness of mitigation strategies across various model architectures. Our findings identify six prevalent cognitive biases, while the sunk cost and group attribution biases exhibit minimal impact. Personality traits play a crucial role in either amplifying or reducing biases, significantly affecting how LLMs respond to debiasing techniques. Notably, Conscientiousness and Agreeableness may generally enhance the efficacy of bias mitigation strategies, suggesting that LLMs exhibiting these traits are more receptive to corrective measures. These findings address the importance of personality-driven bias dynamics and highlight the need for targeted mitigation approaches to improve fairness and reliability in AI-assisted decision-making.

Paper Structure

This paper contains 30 sections, 2 figures, 11 tables.

Table of Contents

  1. Introduction
  2. Cognitive Biases in Decision-Making
  3. Personality Traits and Cognitive Biases
  4. Personality and Bias Impact in LLMs
  5. Research Gap
  6. Methodology
  7. Personality Traits in LLMs
  8. Cognitive Biases
  9. Datasets
  10. Student Admission Dataset
  11. BiasEval Dataset
  12. Experimental Setup
  13. Results
  14. Personality Traits and Cognitive Biases
  15. Reversed Personality Traits
  16. ...and 15 more sections

Figures (2)

  • Figure 1: Personality-Bias Framework.
  • Figure 2: A visualization of the extent to which biases are mitigated across different LLMs and personality traits when applying the awareness debiasing approach. The green-shaded values indicate effective bias reduction, whereas red-shaded values denote instances where the bias increased despite mitigation attempts.