Large Language Newsvendor: Decision Biases and Cognitive Mechanisms

TL;DR

This paper investigates how large language models exhibit and amplify human-like decision biases in supply chain management tasks, revealing that more advanced models can be more irrational and emphasizing the need for oversight and structured prompts.

Contribution

It provides the first systematic analysis of decision biases in LLMs within a supply chain context, highlighting architectural constraints as the source of biases and offering practical mitigation strategies.

Findings

LLMs replicate classic ordering biases like 'Too Low/Too High'

More sophisticated models like GPT-4 show greater irrationality

Efficiency-optimized models perform near-optimally in some tasks

Abstract

Problem definition: Although large language models (LLMs) are increasingly integrated into business decision making, their potential to replicate and even amplify human cognitive biases cautions a significant, yet not well-understood, risk. This is particularly critical in high-stakes operational contexts like supply chain management. To address this, we investigate the decision-making patterns of leading LLMs using the canonical newsvendor problem in a dynamic setting, aiming to identify the nature and origins of their cognitive biases. Methodology/results: Through dynamic, multi-round experiments with GPT-4, GPT-4o, and LLaMA-8B, we tested for five established decision biases. We found that LLMs consistently replicated the classic ``Too Low/Too High'' ordering bias and significantly amplified other tendencies like demand-chasing behavior compared to human benchmarks. Our analysis uncovered a ``paradox of intelligence'': the more sophisticated GPT-4 demonstrated the greatest irrationality through overthinking, while the efficiency-optimized GPT-4o performed near-optimally. Because these biases persist even when optimal formulas are provided, we conclude they stem from architectural constraints rather than knowledge gaps. Managerial implications: First, managers should select models based on the specific task, as our results show that efficiency-optimized models can outperform more complex ones on certain optimization problems. Second, the significant amplification of bias by LLMs highlights the urgent need for robust human-in-the-loop oversight in high-stakes decisions to prevent costly errors. Third, our findings suggest that designing structured, rule-based prompts is a practical and effective strategy for managers to constrain models' heuristic tendencies and improve the reliability of AI-assisted decisions.