From predictive to prescriptive digital twins
Traditional planning tools rely on predictive digital twins built on traffic and transport models. Stakeholders propose a handful of options, and models estimate impacts on accessibility, network performance, and other indicators. This approach has two big limitations: only a small set of designs can be explored, and a small group of decision‑makers tends to dominate the process, raising the risk of unintentional bias.
Our lab focuses on prescriptive digital twins, which can generate and recommend design alternatives themselves. These tools can surface a far larger, more diverse solution set and ensure that all population groups are considered by explicitly incorporating stakeholder preferences and equity indicators into the objectives.
Public authorities and area developers are eager to use prescriptive digital twins to accelerate decisions on mobility and spatial design. Because these systems learn from user input and provide policy recommendations, they qualify as AI systems under the European AI Act and must be used in a compliant and responsible manner. Decisions taken with these tools can influence people’s ability to access essential services at scale; that makes careful risk assessment, transparency, and human oversight essential.
Integrating AI into digital twins raises important ethical, legal, and societal questions. For example:
Bias in recommendations could limit access to new services for marginalised communities.
Traffic management strategies must avoid systematically disadvantaging e.g. lower‑income neighbourhoods.
Mobility hub placement could be skewed toward affluent areas if equity is not explicitly considered.
Shared mobility operations may concentrate supply in high‑demand districts, neglecting others.