Beyond the Wheel: How Infrastructure Designs Steer Human Behavior

A new study reveals how specific elements of the built environment substantially influence driving patterns and road safety. Nicolosi and colleagues analyzed extensive driving data with machine learning to reveal connections between infrastructure and driver conduct.
These findings inform urban planners, traffic engineers, and policymakers on how to improve road safety through environmental design.
The Hidden Influence of Infrastructure
While most road safety discussions focus on individual driver responsibility, this research shows that external environmental factors play a crucial role in shaping how people drive. Certain infrastructure elements consistently correlate with more aggressive driving patterns.
Locations near schools, hospitals, bridges, parking areas, subway stations, bicycle lanes, and pedestrian walkways showed increased likelihood of dangerous driving behaviors. This finding indicates that complex road environments may heighten driver stress and contribute to erratic decisions.
Advanced Methodology Yields Clear Results
This study stands out for its merging of onboard diagnostic (OBD) data with OpenStreetMap geographic information. OBD systems monitor various vehicle functions through electronic control units, providing real-time data on speed, acceleration, engine performance, and fuel consumption. When combined with detailed geographic context, this creates a comprehensive picture of how environmental factors influence driving decisions.
The research team utilized the DAVIS Driving Dataset 2020 (DDD20), which contains 51 hours of recordings across approximately 4,000 kilometers of diverse driving scenarios. This extensive dataset provided the foundation for training powerful neural network models capable of classifying driver behavior into three categories: safe, normal, and dangerous.
The researchers implemented Multi-Layer Perceptron Neural Networks to process this complex data. Three different models were developed. One combining OBD and geographic data, another using only geographic information, and a third employing a reduced set of geographic features. A simplified model using only select geographic features achieved similar accuracy to more complex versions, suggesting that environmental context can predict driving behavior patterns.
Key Environmental Factors
Through Shapley Additive Explanations (SHAP) analysis, the study identified which specific infrastructure elements most strongly influence driver behavior. This technique, based on game theory, quantifies the contribution of each environmental feature to aggressive driving predictions.
The presence of educational institutions emerged as a significant factor, likely due to increased pedestrian activity and complex traffic patterns during school hours. Healthcare facilities showed similar effects, possibly reflecting the stress and urgency often associated with medical visits. Bridges consistently correlated with more aggressive driving, likely due to reduced visibility, lane restrictions, or elevation changes.
Parking areas, subway stations, and cycling infrastructure also contributed to more erratic driving patterns. These locations typically involve complex interactions between different types of road users, creating scenarios where drivers must navigate around pedestrians, cyclists, and other vehicles making unpredictable movements.
Implications for Urban Planning
These findings have major implications for how cities design and manage their transportation infrastructure. Rather than simply adding more safety signage or enforcement in problematic areas, planners might consider how the built environment itself contributes to dangerous driving patterns.
The research suggests areas with multiple demands on driver attention require careful design. Simple interventions like improved sight lines near schools, dedicated turning lanes near hospitals, or better separation between vehicular and pedestrian traffic could help reduce the cognitive load on drivers and decrease aggressive behaviors.
Fuel Consumption and Safety Connections
The study also explores the relationship between driving behavior and fuel efficiency, finding that eco-friendly driving patterns typically align with safer driving practices. Aggressive driving characterized by rapid acceleration and frequent speed changes not only increases safety risks but also substantially increases fuel consumption.
Researchers found that smooth, predictable driving patterns could reduce fuel usage to approximately 4.83 liters per 100 kilometers, while aggressive driving styles increased consumption to 6.15 liters per 100 kilometers. This connection between environmental responsibility and road safety suggests that interventions promoting fuel-efficient driving could yield dual benefits.
Future Applications
The methodology developed in this study offers a scalable approach for analyzing driver behavior across large geographic areas. By relying primarily on publicly available geographic data rather than expensive sensor installations, transportation authorities could implement this framework to identify problematic locations and prioritize safety improvements.
The research team’s ability to achieve accurate behavior prediction using only environmental features opens new possibilities for proactive traffic management. Cities could potentially identify high-risk locations before accidents occur, allowing for preventive measures rather than reactive responses to safety problems.
Advancing Road Safety Through Data Science
Geographic Data Alone Proves Sufficient
Perhaps the most striking finding from this research is that environmental context alone can effectively predict driving behavior patterns. When researchers tested a model using only geographic information without any vehicle sensor data, it achieved over 91% accuracy in classifying dangerous driving behaviors.
This discovery has profound practical implications. Traditional approaches to monitoring driver behavior require expensive onboard diagnostic equipment in every vehicle. The new methodology shows that publicly available geographic data can identify high-risk driving situations with comparable accuracy, making large-scale safety analysis far more feasible and cost-effective.
Streamlined Model Maintains Performance
The research team took their analysis one step further by creating a simplified version using only nine key environmental features identified through SHAP analysis. This reduced model maintained accuracy above 90% across all driving behavior categories while requiring significantly less computational power.
The nine critical factors include proximity to schools, hospitals, terrain slope, bridge locations, parking areas, subway stations, cycling infrastructure, pedestrian pathways, and local wind conditions. This streamlined approach offers several advantages: reduced memory and computing requirements, lower energy consumption, enhanced scalability for widespread deployment, and decreased overall system environmental impact.
Practical Applications and Future Directions
The ability to predict dangerous driving patterns using only geographic data opens new avenues for proactive safety interventions. Urban planners could use this methodology to assess proposed infrastructure changes before construction begins, identifying potentially problematic areas and modifying designs accordingly.
Traffic authorities could apply the framework to existing road networks, systematically identifying locations where environmental factors contribute to aggressive driving. This would allow for targeted interventions such as improved signage, modified traffic flow patterns, or infrastructure adjustments that reduce cognitive load on drivers.
The research team’s success in predicting behavior using only environmental features opens new avenues for proactive traffic management. Smartphone integration with basic vehicle diagnostic systems could provide a cost-effective alternative to expensive sensor installations, making comprehensive driver behavior monitoring accessible to a much broader range of vehicles and drivers.
Connecting Behavior to Safety Outcomes
Future research directions include linking aggressive driving patterns directly to crash data to establish clearer causal relationships between environmental factors and actual safety outcomes. The methodology could be particularly valuable for understanding interactions between drivers and vulnerable road users like pedestrians and cyclists.
The framework’s scalability makes it suitable for analysis across different geographic regions and cultural contexts. Expanding the analysis to European cities or other international locations could reveal whether the identified environmental factors have universal effects on driving behavior or vary by local driving culture and infrastructure design.
A New Approach to Road Safety
This study constitutes an important advancement in understanding how our built environment shapes driving behavior. By proving that environmental context alone can reliably predict dangerous driving patterns, the research provides transportation authorities with a powerful new tool for improving road safety.
The findings suggest that road safety extends beyond individual driver education and enforcement to encompass thoughtful infrastructure design. The research reveals how specific environmental elements consistently influence driver behavior, offering concrete guidance for creating safer transportation systems.
As urban areas continue expanding and traffic patterns become more complex, this type of data-driven analysis will become increasingly valuable. The methodology provides a scientific foundation for making informed decisions about infrastructure design, helping create road environments that naturally encourage safer driving behaviors for all users.
Citation: Nicolosi, V., Mameli, M., Shiralizadeh, S., Coltea, I.G., D’Apuzzo, M., & Cappelli, G. (2025). How does the built environment affect driver behaviours? A methodological framework for large-scale analysis that combines Onboard Diagnostic and Geographical Data to promote Road Safety. Transportation Engineering, 100384. https://doi.org/10.1016/j.treng.2025.100384
