Objectives: Taboo trade-offs can explain some of the (moral) difficulties in healthcare decision-making. The moral psychology literature suggests that individuals are averse to making trade-offs between attributes belonging to different values, such as (sacred) human lives versus (secular) money. We demonstrate and empirically test a discrete choice model designed to capture Taboo Trade-off Aversion (TTOA) behaviors in the healthcare domain. Methods: The linear-additive Random Utility Maximization (RUM) model is extended to capture TTOA behaviors by including penalties for taboo trade-offs. Using two Discrete Choice Experiments (DCEs) focusing on taboo trade-offs in public health policies, we empirically compare conventional linear-additive RUM models with TTOA models to explore differences in model and behavioral results. Results: We observe TTOA in both DCEs. In one DCE, the TTOA model separates TTOA effects from attribute-related parameters, showing inflated parameters in conventional RUM models when TTOA behavior is present. This discrepancy affected Willingness-To-Pay (WTP) estimates, with WTP to save an incremental patient life approximately 3.5 times higher in conventional RUM models compared to the TTOA models. The presence and magnitude of TTOA varied considerably across respondents. Latent Class (LC) models reveal that some respondent groups perceive trade-offs as taboo significantly, while others do not. Conclusions: Accounting for TTOA in RUM models may lead to more accurate behavioral information when choice behaviors are affected by taboo trade-offs. Researchers and policymakers can use TTOA models to obtain a more nuanced understanding of public acceptability in morally salient policy decisions – ultimately helping to navigate, rather than avoid, taboo trade-offs.

The effectiveness of neural network models largely relies on learning meaningful latent patterns from data, where self-supervised learning of informative representations can enhance model performance and generalisability. However, self-supervised representation learning for spatially characterised time series, which are ubiquitous in transportation domain, poses unique challenges due to the necessity of maintaining fine-grained spatio-temporal similarities in the latent space. In this study, we introduce two structure-preserving regularisers for the contrastive learning of spatial time series: one regulariser preserves the topology of similarities between instances, and the other preserves the graph geometry of similarities across spatial and temporal dimensions. To balance the contrastive learning objective and the need for structure preservation, we propose a dynamic weighting mechanism that adaptively manages this trade-off and stabilises training. We validate the proposed method through extensive experiments, including multivariate time series classification to demonstrate its general applicability, as well as macroscopic and microscopic traffic prediction to highlight its particular usefulness in encoding traffic interactions. Across all tasks, our method preserves the similarity structures more effectively and improves state-of-the-art task performances. This method can be integrated with an arbitrary neural network model and is particularly beneficial for time series data with spatial or geographical features. Furthermore, our findings suggest that well-preserved similarity structures in the latent space indicate more informative and useful representations. This provides insights to design more effective neural networks for data-driven transportation research. Our code is made openly accessible with all resulting data at this https URL: https://github.com/yiru-jiao/spclt

We present a method to classify street networks using only geo-tagged street-level imagery. By combining pre-trained image embeddings with unsupervised clustering, it produces visually coherent street typologies without supervised training or labeled data and requires only minimal data curation. The approach is lightweight, scalable, and, in principle, transferable across urban contexts. In a Delft (Netherlands) case study, we classify approximately 2,000 road sections using over 70,000 images. Our method recovers distinct street types such as residential, arterial, and historic ones. These results show that pre-trained visual embeddings alone can support effective street classification from visual inputs, offering a practical tool for urban planning, transport analysis, and mobility research.

Accurate and timely alerts for drivers or automated systems to unfolding collisions remains a challenge in road safety, particularly in highly interactive urban traffic. Existing approaches require labour-intensive annotation of sparse risk, struggle to consider varying contextual factors, or are useful only in the scenarios they are designed for. To address these limits, this study introduces the generalised surrogate safety measure (GSSM), a new approach that learns exclusively from naturalistic driving without crash or risk labels. GSSM captures the patterns of normal driving and estimates the extent to which a traffic interaction deviates from the norm towards unsafe extreme. Utilising neural networks, normal interactions are characterised by context-conditioned distributions of multi-directional spacing between road users. In the same interaction context, a spacing closer than normal entails higher risk of potential collision. Then a context-adaptive risk score and its associated probability can be calculated based on the theory of extreme values. Any measurable factors, such as motion kinematics, weather, lighting, can serve as part of the context, allowing for diverse coverage of safety-critical interactions. Multiple public driving datasets are used to train GSSMs, which are tested with 2,591 real-world crashes and near-crashes reconstructed from the SHRP2 NDS. A vanilla GSSM using only instantaneous states achieves AUPRC of 0.9 and secures a median time advance of 2.6 seconds to prevent potential collisions. Additional data and contextual factors provide further performance gains. Across various interaction types such as rear-end, merging, and crossing, the accuracy and timeliness of GSSM consistently outperforms existing baselines. GSSM therefore establishes a scalable, context-aware, and generalisable foundation to proactively quantify collision risk in traffic interactions.

Several studies examined what drives citizens’ support for COVID-19 measures, but no works have addressed how the effects of these drivers are distributed at the individual level. Yet, if significant differences in support are present but not accounted for, policymakers’ interpretations could lead to misleading decisions. In this study, we use XGBoost, a supervised machine learning model, combined with SHAP (Shapley Additive eXplanations) to identify the factors associated with differences in policy support for COVID-19 measures and how such differences are distributed across different citizens and measures. We use secondary data from a Participatory Value Evaluation (PVE) experiment, in which 1,888 Dutch citizens answered which COVID-19 measures should be imposed under four risk scenarios. We identified considerable heterogeneity in citizens’ support for different COVID-19 measures regarding different age groups, the weight given to citizens’ opinions and the perceived risk of getting sick of COVID-19. Data analysis methods employed in previous studies do not reveal such heterogeneity of policy support. Policymakers can use our results to tailor measures further to increase support for specific citizens/measures.

Visual imagery is indispensable to many multi-attribute decision situations. Examples of such decision situations in travel behaviour research include residential location choices, vehicle choices, tourist destination choices, and various safety-related choices. However, current discrete choice models cannot handle image data algorithmically and thus cannot incorporate information embedded in images into their representations of choice behaviour. This gap between discrete choice models’ capabilities and the real-world behaviour it seeks to model leads to incomplete and, possibly, misleading outcomes. To solve this gap, this study proposes “Computer Vision-enriched Discrete Choice Models” (CV-DCMs). CV-DCMs can handle choice tasks involving numeric attributes and images by integrating computer vision and traditional discrete choice models. Moreover, because CV-DCMs are grounded in random utility maximisation principles, they maintain the solid behavioural foundation of traditional discrete choice models. We demonstrate the proposed CV-DCM by applying it to data obtained through a novel stated choice experiment involving residential location choices. In this experiment, respondents faced choice tasks with trade-offs between commute time, monthly housing cost and street-level conditions, presented using images. We find that CV-DCMs can offer novel insights into preferences regarding features presented in images, such as what street-level conditions people find most and least attractive and how these preferences vary across age groups.

The emergence of social media offers unprecedented opportunities to map social unrest with high spatiotemporal resolution. This study leverages geolocated social media footage to analyze the spatiotemporal distribution of the 2023 ‘Nahel Merzouk’ riots in France. Using a fine-tuned computer vision model, we detect riot-related content in visual data and validate our approach by comparing the spatiotemporal patterns of detected posts with rioting events reported in the press. Our method yields a spatial resolution of 300 300 m, thereby facilitating a detailed analysis of riot distributions at unprecedented scale. By applying density-based clustering, we map riots across seven French cities, revealing their highly localized and bursty dynamics. This study opens pathways for future research on the causes and dynamics of social unrest, enabling a deeper understanding of urban riots and their potential mitigation.

This research introduces a new method for constructing and training an Urban Space Embedding Model (USEM) by integrating human perceptions and street-level images (SLI) into its formulation. Traditional urban embedding models often overlook subjective human experiences, such as perceptions of safety or attractiveness. To address this gap, our method leverages similarity judgments from over 1500 participants, who compared different urban spaces based on SLI. These human judgments were then used as a supervision signal in training the USEM, allowing the model to capture both visual and perceptual information about urban spaces. The method is implemented across the Netherlands, using around one million geo-tagged SLI, and demonstrated in Rotterdam. This approach represents a significant advancement in urban computing by incorporating human-centered data into urban modeling. It offers new opportunities for city planners and policymakers to better understand how urban spaces are perceived and to consider these perceptions in efforts to design more livable and inclusive environments.

BACKGROUND: Long-term noise annoyance can be expected to have worse outcomes than short-term annoyance. This study investigates noise annoyance over time, its association with personality traits and potential reciprocal effects between health outcomes and noise annoyance. METHODS: Firstly, we conducted a Longitudinal Latent Class Analysis to identify noise annoyance profiles. We further analysed the effect of Big Five personality traits on the likelihood of belonging to these annoyance profiles. Secondly, we used Cross-lagged Panel Models to analyse whether changes in noise annoyance precede changes in health outcomes or vice versa. For both analyses, we used 8 years of data from the Dutch Longitudinal Internet Studies for the Social Sciences (LISS) panel. Between 2708 and 11,068 subjects were included (this varies between models). RESULTS: We found three profiles of noise annoyance, namely, chronically, occasionally and never annoyed. Among all participants, 12% were chronically annoyed by neighbour noise and 6% by street noise. Extraversion and emotional stability decreased the chance of belonging to the cluster of chronically annoyed, while openness had the opposite effect. Chronic noise annoyance showed a significant effect on self-reported heart complaints and sleeping problems, while the effects of noise annoyance profiles on high blood pressure and heart attacks were insignificant. Some potential indications for a reverse effect from health outcomes on noise annoyance were found. CONCLUSION: Noise annoyance was relatively stable over time possibly because of its correlation with personality traits. Noise had a small negative effect on health outcomes, and some health outcomes affected noise annoyance. Further research should be conducted to collect dedicated panel data.

Sound source classification is a valuable addition to noise monitoring, providing ‘further insights into local soundscapes. For privacy preservation, this classification often must be conducted on the edge, i.e., in real time on noise sensors. This puts constraints on the size and complexity of the classification models that can be used. Furthermore, there is a trade-off between accuracy and efficiency, which needs to be balanced on battery or solar powered sensors. However, little is known about this trade-off under consideration of constraints imposed by such sensors. In this paper, we explore the scope of sound classification models that can run efficiently on low-cost sound sensors. Specifically, we investigate the Pareto frontiers between model accuracy and computational complexity, providing insights into the trade-off necessary for deploying such models on very constrained hardware. Building on these findings, we train new classification models optimized for edge devices. The models are trained on publicly available audio samples and a new Dutch Urban Sounds dataset specifically collected to enhance the accuracy of sound source classification in urban environments. The models and implementation are open source, enabling researchers and practitioners to adopt, adapt, and build upon our work.

We investigate the evolution of residential segregation patterns in the Netherlands, with a focus on the population with a non-western migration background. Unlike previous research relying on predefined spatial structures, this study employs a regionalization approach to track the evolution of social enclaves in 82 municipalities from 2015 to 2020. Enclaves have become more mixed in municipalities with historically homogeneous social enclaves whereas in the other municipalities, they have become more homogeneous. In addition, we find a positive association between the increase in the share of population with a non-western migration background at the municipality level and the spatial growth of social enclaves. These insights contribute to a deeper understanding of residential segregation in the Netherlands, offering a valuable foundation for informed policymaking.

In this article, the affiliation details for author Jose Ignacio Hernandez were incorrectly given as ‘Center of Economics for Sustainable Development (CEDES), Faculty of Economics and Government, Universidad San Sebastian, Lientur 1457, Concepción, Chile ' but should have been ‘Center of Economics for Sustainable Development (CEDES), Faculty of Economics and Government, Universidad San Sebastian, Concepción, Chile. The affiliation ‘Transport and Logistics Group, Faculty of Technology, Policy and Management, Delft University of Technology, Delft, Netherlands’ for authors Jose Ignacio Hernandez and Niek Mouter was missing. The original article has been corrected.

Efficient experimental designs aim to maximise the information obtained from stated choice data to estimate discrete choice models' parameters statistically efficiently. Almost without exception efficient experimental designs assume that decision-makers use a Random Utility Maximisation (RUM) decision rule. When using such designs, researchers (implicitly) assume that the decision rule used to generate the design has no impact on respondents' choice behaviour. This study investigates whether the decision rule assumption underlying an experimental design affects respondents' choice behaviour. We use four stated choice experiments on coastal adaptation to climate change: Two are based on experimental designs optimised for utility maximisation and two are based on experimental designs optimised for a mixture of RUM and Random Regret Minimisation (RRM). Generally, we find that respondents place value on adaptation measures (e.g., dykes and beach nourishments). We evaluate the models' fits and investigate whether some choice tasks particularly invoke RUM or RRM decision rules. For the latter, we develop a new sampling-based approach that avoids the confounding between preference and decision rule heterogeneity. We find no evidence that RUM-optimised designs invoke RUM-consistent choice behaviour. However, we find a relationship between some of the attributes and decision rules, and compelling evidence that some choice tasks invoke RUM consistent behaviour while others invoke RRM consistent behaviour. This implies that respondents’ choice behaviour and choice modelling outcomes are not exogenous to the choice tasks, which can be particularly critical when information on preferences is used to inform actual decision-making on a sensitive issue of common interest as climate change.

This paper introduces systems theory and system safety concepts to ongoing academic debates about the safety of Machine Learning (ML) systems in the public sector. In particular, we analyze the risk factors of ML systems and their respective institutional context, which impact the ability to control such systems. We use interview data to abductively show what risk factors of such systems are present in public professionals' perceptions and what factors are expected based on systems theory but are missing. Based on the hypothesis that ML systems are best addressed with a systems theory lens, we argue that the missing factors deserve greater attention in ongoing efforts to address ML systems safety. These factors include the explication of safety goals and constraints, the inclusion of systemic factors in system design, the development of safety control structures, and the tendency of ML systems to migrate towards higher risk. Our observations support the hypothesis that ML systems can be best regarded through a systems theory lens. Therefore, we conclude that system safety concepts can be useful aids for policymakers who aim to improve ML system safety.

Noise annoyance and its relation to health outcomes have been studied extensively. The vast majority of studies in this field use cross-sectional data. Such data does not allow investigation of temporal effects or the direction of these effects. It is reasonable to expect that the effects of noise on health build up over time. Moreover, noise may not only impact health outcomes, but health outcomes may also impact the sensitivity to noise – hence flipping the direction of the effect. Further adding to the complexity of the relationship between noise and health outcomes is the fact that personality traits may influence re-ported noise annoyance. This study aims to shed light on the accumulative effects of noise annoyance on health, as well as the bidirectional relationships between noise annoyance and health. To do so, we analyze eight years of data from the Dutch Longitudinal Internet Studies for the Social Sciences (LISS) panel. Specifically, we conduct a Longitudinal Latent Class Analysis to identify annoyance profiles and analyze the effect of Big Five personality traits on the likelihood of belonging to the different annoyance profiles. Furthermore, we use Cross-Lagged Panel Models to analyze whether changes in noise annoy-ance precede changes in health outcomes or vice versa. We find three different profiles of noise annoy-ance, namely chronically annoyed people, occasionally annoyed people, and people who are generally not annoyed. Noise annoyance was found to be relatively stable over time. Regarding personality traits, we find that extraversion and emotional stability decrease the chance of belonging to the cluster of chronically annoyed persons, while openness has the opposite effect. Finally, chronic annoyance shows a significant effect on self-reported heart complaints and sleeping problems, while the effects of noise annoyance profiles on high blood pressure and heart attacks are insignificant.

Integrating Automated Vehicles (AVs) into existing traffic systems holds the promise of enhanced road safety, reduced congestion, and more sustainable travel. Effective integration of AVs requires understanding the interactions between AVs and Human-driving Vehicles (HVs), especially during the transition period in which AVs and HVs coexist in a mixed traffic environment. Numerous recent empirical studies find reduced headways of human drivers following an AV compared to following an HV, and attribute this reduction to behavioural changes of drivers when they follow AVs. However, more factors may be at play due to the inherent inconsistencies between the comparison conditions of HV-following-AV and HV-following-HV. This study scrutinises three alternative explanations for the observed reduction in headways: (1) systematic differences in car-following states during data collection, (2) systematic differences in driving variability between leading AVs and HVs, and (3) systematic differences in driving characteristics of leading AVs versus HVs. We use a large-scale dataset extracted from Lyft AV motion data and examine each of these explanations through data stratification and simulation. Our results show that all three mechanisms contribute to the observed reduction in headways of human drivers following AVs. In addition, our findings highlight the importance of driving homogeneity and stability in achieving reliably shorter headways. Thereby, this study offers a more comprehensive understanding on the difference between HV–AV and HV–HV interactions in mixed traffic, and is expected to promote more effective integration of AVs into human traffic.

Traffic conflict detection is essential for proactive road safety by identifying potential collisions before they occur. Existing methods rely on surrogate safety measures tailored to specific interactions (e.g., car-following, side-swiping, or path-crossing) and require varying thresholds in different traffic conditions. This variation leads to inconsistencies and limited adaptability of conflict detection in evolving traffic environments, particularly as the integration of autonomous driving systems adds complexity. Consequently, there is an increasing need for consistent detection of traffic conflicts across interaction contexts. To address this need, we propose a unified probabilistic approach in this study. The proposed approach establishes a unified framework of traffic conflict detection, where traffic conflicts are formulated as context-dependent extreme events of road user interactions. The detection of conflicts is then decomposed into a series of statistical learning tasks: representing interaction contexts, inferring proximity distributions, and assessing extreme collision risk. The unified formulation accommodates diverse hypotheses of traffic conflicts and the learning tasks enable data-driven analysis of factors such as motion states of road users, environment conditions, and participant characteristics. Jointly, this approach supports consistent and comprehensive evaluation of the collision risk emerging in road user interactions. We demonstrate the proposed approach by experiments using real-world trajectory data. A unified metric for indicating conflicts is first trained with lane-change interactions on German highways, and then compared with existing metrics using near-crash events from the U.S. 100-Car Naturalistic Driving Study. Our results show that the unified metric provides effective collision warnings, generalises across distinct datasets and traffic environments, covers a broad range of conflict types, and captures a long-tailed distribution of conflict intensity. In summary, this study provides an explainable and generalisable approach that enables traffic conflict detection across varying interaction contexts. The findings highlight its potential to enhance the safety assessment of traffic infrastructures and policies, improve collision warning systems for autonomous driving, and deepen the understanding of road user behaviour in safety–critical interactions.

Two-attribute-two-alternative stated choice experiments are widely used to infer the Value-of-Travel-Time (VTT) distribution. Two-attribute-two-alternative stated choice experiments have the advantage that their data can be analysed using nonparametric models, which allow for the inference of the VTT distribution without having to impose assumptions on its shape. However, a software package that enables researchers to estimate nonparametric models promptly is currently lacking. As a result, nonparametric models are underused. This paper aims to fill this software void. It presents NP4VTT, a Python package that enables researchers to estimate and compare nonparametric models in a fast and convenient way. It comprises five nonparametric models for estimating the VTT distribution from data coming from two-attribute-two-alternative stated choice experiments. We illustrate the use of NP4VTT by applying it to the Norwegian 2009 VTT data. We hope this software package will help researchers studying the VTT make more informed decisions concerning the shape of the VTT distribution and encourages the use and development of nonparametric models for choice behaviour analyses.

We propose three procedures based on association rules (AR) learning and random forests (RF) to support the specification of a portfolio choice model applied in data from complex choice experiment data, specifically a Participatory Value Evaluation (PVE) choice experiment. In a PVE choice experiment, respondents choose a combination of alternatives, subject to a resource constraint. We combine a methodological-iterative (MI) procedure with AR learning and RF models to support the specification of parameters of a portfolio choice model. Additionally, we use RF model predictions to contrast the validity of the behavioural assumptions of different specifications of the portfolio choice model. We use data of a PVE choice experiment conducted to elicit the preferences of Dutch citizens for lifting COVID-19 measures. Our results show model fit and interpretation improvements in the portfolio choice model, compared with conventional model specifications. Additionally, we provide guidelines on the use of outcomes from AR learning and RF models from a choice modelling perspective.