Challenging the Appearance of Machine Intelligence: Cognitive Bias in LLMs and Best Practices for Workplace Adoption

By Alaina N. Talboy et al.

Read the original document by opening this link in a new tab.

Table of Contents

Abstract
Keywords
Introduction
An Introduction to Heuristics and Cognitive Biases
Bias in Machine Intelligence
Testing for Cognitive Bias in LLMs

Summary

This paper challenges the appearance of machine intelligence by exploring cognitive biases in large language models (LLMs) and proposing best practices for workplace adoption. It discusses the presence of cognitive biases in LLMs beyond traditional discriminatory biases, emphasizing the need for education, risk management, and continued research. The paper highlights the potential negative impact of biased reasoning in AI technologies and suggests caution in their widespread adoption. The authors provide evidence of several cognitive biases in LLM outputs and advocate for a holistic understanding of bias in machine intelligence. The study tests prominent cognitive biases in LLMs and examines their susceptibility to biases related to reasoning, representativeness, insensitivity to sample size, base rate neglect, and value selection bias.