Collaborative explainable neuro-symbolic AI for Decision Support Assistant
CAI4DSA aims at exploiting AI for creating decision support systems (DSS) which can evolve with the collaboration among and with humans. The NGDSS should be capable to react to changes in the context, interact with humans, and learn from the occurrence. Symbolic and neural models could be also shared in communities to create a global knowledge and understanding of the problems and solutions. The main objectives will be to develop new knowledge and solutions regarding: Continuous incremental reinforced learning; human readable multidomain XAI representations; extract and formalize new knowledge from XAI, results/KPIs as facts, relationships, constraints, models (causals, maths, etc.); human NGDSS interaction which should be capable to work reconcile with humans; and validation of NGDSS in a number of critical infrastructure domains.
In this context, a significant relevance is attributed to: continuous incremental reinforced learning; human readable multidomain XAI representations to facilitate human understanding and trust; extract and formalize new knowledge from XAI, results/KPIs; human NGDSS interaction; validation of NGDSS in critical infrastructure domains. The research will be based on an iterative approach that focuses on developing, evaluating, and deploying artifacts that address specific problems or opportunities in the specific context. The iterative nature would allow for continuous improvement of the NGDSS based on feedback from decision makers in the target domain. A number of reference cases in different domains will be used for validation for which details and data are accessible: mobility and transport, medicine, complex manufacturing, and security. The innovative outcomes of this research project will be:
• New Algorithms or Models: Development of novel AI algorithms or models that can enhance decision-making processes.
• NGDSS Theories: Contribution to the theoretical understanding of how AI can be effectively integrated into decision support systems.
• Early Prototype of the NGDSS, providing a proof of concept for how AI can enhance decision-making in a number of target domains
The project results will be sustainable since the results will be proof with respect to real cases. This approach will also guarantee a strong impact in terms of: new knowledge on neuro-symbolic DSS, research publications and presentations, enlargement of research network, ethical and societal implications due to the domain of validation, future research directions, and finally also strong possibilities of the exploitation of results via some technology transfer and commercialization of results. The NGDSS integration will be performed in the context of the Snap4City.org platform which simplifies the data ingestion, data access, data and results representation making the integration and general assessment phases simpler and cheaper.
Keywords: decision support systems, XAI/AI, KPI-based XAI, neuro-symbolic approach, human in the loop.
Project totally performed by University of Florence, DINFO department: DISIT Lab, MICC, AI Lab.
Coordinator: Paolo Nesi, paolo.nesi@unifi.it, DISIT LAB
Reference platform: Snap4City
Project duration 14 months.
Founded by PNRR, PE FAIR: https://fondazione-fair.it/
