AI neural network predicting travel choices over a cityscape

Unlock Travel Insights: How AI is Revolutionizing Choice Modeling

Beau Callahan in Tech & Innovation December 2025 • 4 min read.

"Discover how AI neural networks are transforming travel behavior analysis, offering more accurate and economically consistent predictions than traditional methods."

Understanding why people make certain travel choices is crucial for urban planning, policy development, and transport appraisal. Traditionally, this has been done using Random Utility Maximization (RUM) models, which are fundamental in discrete choice modeling. However, specifying the utility function in these models can be tricky and significantly impacts the outcomes, especially when interpreting welfare measures.

Enter artificial neural networks (ANNs), a type of AI that's changing the game. ANNs offer a way to create more flexible models that can adapt to complex data without the rigid specifications of traditional methods. One such model, the Alternative-Specific and Shared weights Neural Network (ASS-NN), balances flexibility with economic consistency. It respects core economic principles like the fungibility of money, ensuring that the model's outputs are economically sound.

This article explores how the ASS-NN model works, its advantages over traditional models, and its potential applications in transport appraisal and policymaking. We'll delve into how it uses data to understand travel choices, providing a more nuanced and accurate picture of traveler behavior.

Why Traditional Travel Models Fall Short

AI neural network predicting travel choices over a cityscape

Traditional RUM models require analysts to predefine the utility function, often through a trial-and-error process. This involves estimating several competing models with different functional specifications based on prior knowledge and economic theory. The final specification is chosen based on behavioral intuition, goodness-of-fit, and information criteria.

However, the true utility functional form is often unknown and is assumed a priori by the analyst. This choice has a considerable impact on interpretable measures such as willingness to pay, making the selection process critical for policymaking and appraisal.

Specification Challenges: Traditional models struggle with accurately specifying utility functions, leading to potential inaccuracies.
Impact on Measures: The chosen utility specification significantly impacts derived measures like willingness to pay.
Limited Flexibility: These models may not capture complex interactions and nonlinear effects in traveler behavior.

Machine learning (ML) models, particularly ANNs, offer an alternative by learning patterns and approximating mathematical functions directly from data. In discrete choice modeling, ANNs have gained traction due to their ability to model complex interactions without needing predefined specifications.

The Future of Travel Modeling with AI

The ASS-NN model represents a significant step forward in travel behavior modeling. By combining the flexibility of neural networks with core economic principles, it offers a more accurate, consistent, and interpretable approach to understanding travel choices. As AI continues to evolve, its applications in transport planning and policy development will likely expand, leading to more informed and effective decision-making.

About this Article -

This article was crafted using a human-AI hybrid and collaborative approach. AI assisted our team with initial drafting, research insights, identifying key questions, and image generation. Our human editors guided topic selection, defined the angle, structured the content, ensured factual accuracy and relevance, refined the tone, and conducted thorough editing to deliver helpful, high-quality information.See our About page for more information.

This article is based on research published under:

DOI-LINK: https://doi.org/10.48550/arXiv.2404.13198,

Title: An Economically-Consistent Discrete Choice Model With Flexible Utility Specification Based On Artificial Neural Networks

Subject: stat.ml cs.lg econ.em

Authors: Jose Ignacio Hernandez, Niek Mouter, Sander Van Cranenburgh

Published: 19-04-2024

Everything You Need To Know

What are the main challenges of using Random Utility Maximization (RUM) models in travel behavior analysis?

Traditional RUM models face several specification challenges. Analysts must predefine the utility function, often relying on trial and error. This process can be subjective and influence the results. The choice of utility function significantly impacts key metrics like willingness to pay, which is critical for policy decisions. Moreover, RUM models may struggle to capture the complex and nonlinear interactions inherent in traveler behavior, limiting their accuracy and predictive power.

How does the Alternative-Specific and Shared weights Neural Network (ASS-NN) improve upon traditional travel modeling methods?

The ASS-NN model offers several advantages. Unlike Random Utility Maximization (RUM) models, it leverages the flexibility of artificial neural networks (ANNs) to learn patterns directly from data, reducing the need for predefined utility functions. This allows it to capture complex interactions more effectively. Crucially, ASS-NN incorporates economic consistency by respecting principles such as the fungibility of money. This ensures that its outputs are economically sound, leading to more reliable and interpretable results for transport appraisal and policymaking.

What role do Artificial Neural Networks (ANNs) play in the evolution of travel choice modeling?

Artificial Neural Networks (ANNs) are at the forefront of revolutionizing travel choice modeling. They provide a flexible alternative to traditional methods by learning from data without requiring predefined specifications. ANNs, like the ASS-NN model, can model intricate relationships within travel data, improving the accuracy of predictions and offering deeper insights into consumer behavior. This advancement enables more informed decision-making in urban planning, policy development, and transport appraisal.

Explain the concept of 'willingness to pay' and why the choice of utility function in Random Utility Maximization (RUM) models affects it.

Willingness to pay (WTP) represents the maximum amount a traveler is prepared to spend for a particular benefit, such as a shorter travel time or improved comfort. In Random Utility Maximization (RUM) models, the choice of the utility function, which describes the preferences of travelers, directly affects the calculation of WTP. Different utility function specifications can lead to significantly different WTP values, making the selection process crucial for accurate and reliable policy recommendations. The subjectivity in defining the utility function can lead to inaccuracies in WTP estimations, thereby undermining the model's reliability.

How can AI and the ASS-NN model be used to improve transport planning and policy development?

AI, particularly the ASS-NN model, offers significant improvements for transport planning and policy development. By providing more accurate and economically consistent predictions of travel behavior, the ASS-NN model enables policymakers to make more informed decisions. It can help assess the impact of different policies, such as changes in fares, infrastructure investments, or congestion pricing, by understanding how travelers will react. This leads to more effective resource allocation, improved transport systems, and better outcomes for both travelers and the community. Furthermore, by offering greater interpretability and understanding of traveler behavior, the ASS-NN can assist in designing policies that align with societal objectives more effectively.