Research question/goal:

The project CAIUS studied the unintended consequences of AI-based technologies in smart city contexts at the theoretical, empirical, and applied level. Focusing on real-world applications in resource allocation and service pricing, we investigated the impact of AI-based decision-making on urban societies. To this end and to advance theory on the digitisation of society, we conducted social simulations modelling AI-based decision-making, citizen behaviour, and individual attitudes. The parameters for these simulations were empirically grounded through large-scale surveys of public fairness perceptions and survey experiments conducted in Germany. The project explored two real-world use cases: the dynamic pricing of urban parking places and the algorithmic matching of refugees to locations within their host country.

Overall, the project results showed that public fairness perceptions are highly context-dependent and that semi-automated decision-making is favoured over fully automated approaches. Moreover, the simulations for the first use case showed that dynamic parking pricing, as tested and implemented in numerous cities worldwide, generally has a negative effect on the fairness of parking spaces for different income groups. The simulations for the second use case further highlighted that the effects the algorithmic matching of refugees differ: compared to random allocation, estimated improvements in employment outcomes due to algorithmic location assignment varied geographically and demographically, which in turn reinforces existing structural disparities between regions and groups of refugees.[NH1.1][PH1.2] Overall, the project showed the value of interdisciplinary research at the intersection of computer science and social science in studying the emerging challenges of AI applications in urban environments.