Adnan KhanAdnanProduct & AI
San Francisco Bay Area
Get in touch

Source note3 min read

Reproducing human biases in route choice using large language models: Toward scalable behavioral modeling

Source

arXiv

arxiv.org/abs/2607.11632v1

Read the source

Source

Plain-Language Brief

This research explores the ability of large language models (LLMs) to reproduce human biases in route choice decisions, traditionally described using Cumulative Prospect Theory (CPT). The study finds that LLMs can generate decisions that match human behavioral patterns without explicitly specifying prospect-theoretic parameters.

What the Document Says

  • Human choice behavior exhibits biases not fully covered by full rationality assumptions.
  • Cumulative Prospect Theory (CPT) is used to characterize these biases but requires individual-level CPT parameters, a challenge due to limitations in conventional calibration methods.
  • The paper investigates if LLMs can reproduce human behavioral biases in choice-making without explicitly specifying prospect-theoretic parameters.
  • Route choice is used as a representative scenario for this investigation.
  • A behavioral evaluation framework is designed, and LLM-generated decisions are compared with established CPT-predicted human behavioral patterns.
  • Results show that LLMs can reproduce non-rational human choice biases and exhibit behaviors consistent with prospect-theoretic effects under uncertainty.
  • The findings suggest LLMs may offer a scalable alternative for modeling human decision processes and be beneficial for next-generation large-scale agent-based simulation and AI-driven behavioral research.

Key Concepts Mentioned

  • Human choice behavior
  • Route choice
  • Cumulative Prospect Theory (CPT)
  • Large language models (LLMs)
  • Generative AI models

Claims Made by the Document

  • Human choice behavior deviates from full rationality assumptions.
  • CPT is effective for characterizing human behavioral patterns but requires individual-level CPT parameters.
  • Conventional methods for calibrating CPT parameters are difficult to generalize and fail to capture diversity in human decision-making.
  • LLMs can reproduce non-rational human choice biases and exhibit behaviors consistent with prospect-theoretic effects under uncertainty.
  • Generative AI models may provide a scalable alternative for modeling human decision processes.

Evidence Used

  • The study presents experimental results comparing LLM-generated decisions with established CPT-predicted human behavioral patterns.

Methods / Approach

  • Design of a behavioral evaluation framework
  • Comparison of LLM-generated decisions with established human behavioral patterns predicted by CPT

Limitations Stated by the Authors

  • The study focuses on route choice as a representative scenario, but the findings may not be universally applicable to all decision-making scenarios.

Limitations Noted by the Agent

Agent interpretation — not from the source.

  • The study does not explore potential biases or limitations in the LLM itself.
  • The study does not consider the potential impact of different LLM architectures on the reproduction of human behavioral patterns.

Open Questions

  • How applicable are the findings to other decision-making scenarios beyond route choice?
  • How do different LLM architectures affect the reproduction of human behavioral patterns?

Notes for Future Q&A

May include interpretation. Labeled where it does.

  • The study does not specify which LLM was used.
  • The study does not provide details on how the behavioral evaluation framework was designed.

Raw Abstract / Excerpt

Human choice behavior, including route choice, exhibits systematic behavioral biases that deviate from the assumptions of full rationality. Cumulative prospect theory (CPT) has been widely recognized as an effective framework for characterizing such behavioral patterns. However, its large-scale application, particularly in simulation and agent-based modeling, critically depends on specifying individual-level CPT parameters, which remain a major bottleneck. Conventional approaches typically rely on surveys and controlled experiments to calibrate CPT parameters, yet these methods are difficult to generalize and often fail to capture the full diversity of human decision-making. To address this challenge, this paper investigates whether large language models (LLMs) can reproduce human behavioral biases in choice-making without explicit specification of prospect-theoretic parameters. Using route choice as a representative scenario, we design a behavioral evaluation framework and systematically compare LLM-generated decisions with established human behavioral patterns predicted by CPT. Experimental results demonstrate that LLMs are capable of reproducing non-rational human choice biases and can exhibit decision behaviors consistent with prospect-theoretic effects under uncertainty. These findings suggest that generative AI models may provide a scalable alternative for modeling human decision processes and offer a promising foundation for next-generation large-scale agent-based simulation and AI-driven behavioral research.