Introduction
AI is now used for forecasting, pricing, routing, and operations across transport and logistics.
It supports tasks such as demand prediction, fleet allocation, dynamic pricing, and real-time control. While these applications can improve efficiency, they can also reproduce or amplify existing inequalities, especially when algorithms learn from data sets that embed underlying biases [1,2].
Data bias feeding into algorithmic bias can, for instance, manifest in unequal bike-rebalancing service provision when models over-serve central or affluent areas while neglecting peripheral or low-income districts [3]. Other examples include hidden distributional effects in simulations and digital twins that rely on synthetic populations lacking key socio-demographic detail (for example, gender, ethnicity, disability), thereby masking inequities [4].
Transport is a domain already fraught with bias along gender, income, ethnicity, education, ability, and age dimensions, among others (see, for example, [5-7]). It is critical that any newly introduced technologies, including AI, tackle rather than perpetuate or worsen these inequities.
In transport research, algorithms—including optimisation and machine learning—are often treated as neutral. A key question, therefore, is whether fairness criteria are considered and, if not, how to incorporate them alongside predictive accuracy, efficiency, and reliability.
There is strong policy demand [8] for cities, regulators, and operators to demonstrate how inclusive AI-enabled mobility is and whether, and how, AI can be trusted. Yet practical guidance on designing and evaluating fair transport algorithms remains scarce. This PhD will address that gap by integrating fairness directly into the design and operation of algorithms, aiming to ensure that AI systems are not biased or discriminatory towards particular groups.
Aims and Direction
The aim is to develop and validate methods to diagnose and reduce unfairness in AI-driven transport while maintaining acceptable operational performance.
Specifically, it will explore:
To what extent existing allocation, pricing, or other AI-enabled methods reproduce or amplify spatial and socio-demographic inequities.
Which fairness definitions (equality, equity, need-based) are most appropriate for transport decision-making, and under what conditions.
How fairness-constrained optimisation, machine learning, or other methods can be designed and tested to preserve acceptable efficiency.
How different sociodemographic groups perceive fairness, and how those perceptions influence trust and adoption.
One possible direction is to create and test a fairness-aware framework for allocating shared mobility services. The framework would measure disparities, select a context-appropriate fairness definition, and embed it in optimisation as explicit constraints, demonstrating the resulting efficiency trade-offs. Other directions include a focus on fairness in surge pricing in ride-hailing (for example, Uber), routing in journey planners or demand-responsive buses, sensors in autonomous vehicles and traffic systems, or the design of simulation models and digital twins.
The project takes an interdisciplinary approach, combining operations research, computer science, and transport psychology. It links algorithmic fairness (detecting and mitigating bias in models) and fairness of outcomes (setting equity goals and assessing efficiency–equity trade-offs) with perceived fairness and acceptance (gathering users’ views).
References:
[1] Fermanian, J-D., Xidonas, P. & Corrente, S. (2025) Machine learning and fairness: an integrated multicriteria approach for the evaluation of supervised classifiers. J. Oper. Res. Soc. 0:1–13. https://doi.org/10.1080/01605682.2025.2554749
[2] Zhang, X., Ke, Q. & Zhao, X. (2024) Travel demand forecasting: a fair AI approach. IEEE Trans. Intell. Transp. Syst. 25:14611–14627. https://doi.org/10.1109/TITS.2024.3395061
[3] Villa-Zapata, L.M., Rodriguez-Roman, D., Flórez-Coronel, J.E., González-López, J.M. & Figueroa-Medina, A.M. (2024) Incorporating equity in the vehicle rebalancing operations of dockless micromobility services. Lat. Am. Transp. Stud. 2:100009. https://doi.org/10.1016/j.latran.2024.100009
[4] Nag, D., Brandel-Tanis, F., Pramestri, Z.A., Pitera, K. & Frøyen, Y.K. (2025) Exploring digital twins for transport planning: a review. Eur. Transp. Res. Rev. 17:15. https://doi.org/10.1186/s12544-025-00713-0
[5] Suel, E., Lynch, C., Wood, M., Murat, T., Casey, G. & Dennett, A. (2024) Measuring transport-associated urban inequalities: where are we and where do we go from here? Transp. Rev. 44(6):1235–1257. https://doi.org/10.1080/01441647.2024.2389800
[6] Parnell, K.J., Pope, K.A., Hart, S., Sturgess, E., Hayward, R., Leonard, P. & Madeira-Revell, K. (2022) ‘It’s a man’s world’: a gender-equitable scoping review of gender, transportation, and work. Ergonomics 65(11):1537–1553. https://doi.org/10.1080/00140139.2022.2070662
[7] Transport Committee (2025) Access denied: rights versus reality in disabled people’s access to transport. UK Parliament. Available at: https://committees.parliament.uk/publications/48336/documents/252972/default/ (accessed 25.09.2025)
[8] Department for Transport (2025) Transport Artificial Intelligence Action Plan: Transforming Ambitions. London: Department for Transport. Available at: https://assets.publishing.service.gov.uk/media/68519a48510376e43ffdb344/transport-ai-action-plan.pdf (accessed 23.09.2025)