Reproducing human biases in route choice using larg...

Reproducing human biases in route choice using larg... | AI Research

Key Takeaways

Reproducing human biases in route choice using large language models: Toward scalable behavioral modeling explores whether Large Language Models (LLMs) can a...
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.
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.

Paper AbstractExpand

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.

Reproducing human biases in route choice using large language models: Toward scalable behavioral modeling explores whether Large Language Models (LLMs) can act as realistic, scalable proxies for human decision-making. Traditional methods for modeling how people choose routes—such as Cumulative Prospect Theory (CPT)—often rely on expensive, small-scale surveys that are difficult to generalize. This research investigates if LLMs can naturally replicate the "non-rational" biases humans exhibit, such as loss aversion and risk sensitivity, without needing complex manual parameter tuning.

Simulating Human Behavior

To test this, the researchers developed a framework that treats LLMs as individual "agents." Each agent is assigned a specific profile—including occupation, personality traits, and travel goals—using standardized prompts. These agents are then placed into simulated traffic environments where they must choose between different routes under varying conditions of risk and uncertainty. By structuring these interactions, the researchers can transform the LLMs' natural language decisions into data that can be analyzed statistically.

Replicating Human Biases

The study found that LLMs are capable of exhibiting decision-making patterns that mirror human behavior. Specifically, the models demonstrated characteristics consistent with Cumulative Prospect Theory, such as evaluating gains and losses relative to a reference point and showing different risk preferences depending on whether they are facing a potential gain or loss. This suggests that the internal reasoning capabilities of LLMs can capture the subjective, boundedly rational nature of human choices in complex, uncertain environments.

A Scalable Future for Research

The findings suggest that generative AI could significantly lower the barrier to entry for behavioral research. By using LLMs to simulate large, diverse populations of agents, researchers can conduct virtual experiments that are more scalable and flexible than traditional laboratory studies. This approach offers a promising foundation for next-generation agent-based simulations, potentially allowing for more accurate predictions in fields like traffic management and urban planning without the limitations of traditional, small-sample data collection.

Important Considerations

While these results are promising, the researchers note that this is an emerging field. The study highlights that the robustness of these LLM-generated behaviors needs continued verification. Because LLMs are sensitive to how they are prompted, creating standardized, reliable frameworks for these virtual experiments remains a key challenge. Future work will need to further refine how these models estimate specific behavioral parameters to ensure they remain consistent across even larger and more diverse simulated populations.