CAIUS: Consequences of AI-Based Decision Making for Urban Societies

Research question/goal: 

The deployment of AI in decision-making processes has the potential to allocate resources efficiently and evaluate situations objectively. Drawing upon these advantages, municipalities employ sensors, cameras, and other AI-related technologies and applications to enhance their smart city infrastructures. However, apart from the desired improvements, such technologies may also have unintended consequences for urban societies: by exacerbating existing social inequalities or creating new ones, social solidarity of the urban society can be eroded. CAIUS aims to unveil such unintended consequences on a theoretical, empirical, and applied level: drawing upon real-world applications in the smart city context (resource allocation and service pricing), we investigate the impact of AI-based decision-making on individual citizen behaviour and human society at large. To this end and to advance theory on the digitization of society, we conduct social simulations modelling AI-based decision-making, citizen behaviour, and attitudes. The parameters for these simulations are empirically acquired through surveys and experiments. The gained insights are applied in two real-world use cases with local partners: (1) the choice of spots to install smart cameras for traffic law enforcement and (2) dynamic pricing of parking places. Ultimately, going beyond these specific use cases, we infer a general framework for the evaluation of AI applications in urban contexts.

Current stage: 

The project team has reviewed the potential of sociological research on the consequences of smart systems based on artificial intelligence in urban contexts. At the same time, we developed an agent-based simulation for modelling parking in the city of Mannheim, where parking opportunities are influenced by a smart (AI) system. For example, an intelligent system can influence parking behaviour by means of adapted pricing. Next steps involve the design and collection of data on individual preferences and utility functions to develop a more detailed simulation model, which may also serve as a framework for evaluating future systems regarding unintended consequences.

Fact sheet

Volkswagen Foundation
2019 to 2023
Geographic Space: 
Metropolitan Region Rhine-Neckar