Paolo Nesi

Artificial Intelligence is increasingly used to support decision-making across many domains. However, concerns related to transparency, reliability and human oversight indicate the need for improved human-centered AI (HCAI) approaches in decision support systems (DSSs). In this paper, a systematic review was conducted in accordance with PRISMA 2020 in the Web of Science database: ninety research articles published between 2015 and 2025 were analyzed to investigate how HCAI is applied within DSSs in multiple application domains. HCAI + DSS research outcomes were analyzed and explored, first identifying the main architectural designs and discussing the involved components integrating human interaction, generative AI models, data and knowledge management, decision logic, and orchestration mechanisms, then focusing on specific domains and highlighting impact achieved, technologies used, and validation strategies employed. In addition, alignment with United Nations’ Sustainable Development Goals (SDG) was considered, and the temporal evolution of the most relevant topics was studied to identify more interesting trends and less investigated areas. Finally, findings were summarized, current limitations were discussed, and future research directions for helping researchers and practitioners in developing more reliable, explainable, and human-aware decision support systems were outlined.

A systematic transition toward more sustainable transport solutions, included among the Sustainable Development Goals of the United Nations, is one of the main challenge modern cities must face. Increasing the usage of public transport and alternative mobility, like sharing services and micro-mobility, has been indicated as the most promising strategy since it can provide wide accessibility, reduce traffic congestion and consequently contain environmental impacts. Nevertheless, the citizen behavior shift toward such solutions in place of private motor mobility is hampered by several factors, among which quality of service emerge as one of the most relevant. To support decision makers in what-if analysis, tactical and strategic planning for improving of transport services, we present MODOM (MObility Demand-Offer Matching), a lightweight agent-based simulator that takes into account public transportation offer and sharing/micro-mobility solutions to evaluate the match with respect to the city users’ demand. MODOM supports dynamic updates of people’s travel plans to react to specific conditions and produces rich, decision-grade KPIs. The simulator runs city-scale scenarios in few minutes, avoids time-consuming configurations, and is integrated with a Smart City Digital Twin for rapid what-if analysis, Snap4City. When needed for tactical planning, measured traffic can be injected by altering public transport timetables. Several experiments demonstrate that MODOM solves current simulator shortcomings and provides detailed simulation results to support decision-making processes.

In the early 2000s, the rapid growth of data—driven by European Open Data policies, IoT/IoE, and digital platforms—created new opportunities for knowledge extraction. The DISIT Lab at the University of Florence shifted its research toward Big Data analytics, AI, and smart city systems. Key achievements include the Km4City knowledge model and the Snap4City platform, enabling data-driven decision support, smart mobility, urban analytics, and digital twin solutions for cities.