Explainable Artificial Intelligence for Agile Mediation Propensity Assessment

Abstract

Italian Justice has recently added mechanisms to exploit mediation process. One of the most critical aspects is a reliable identification of litigations which can be successfully mediated outside court procedures. The decision is under responsibility of a judge/court who has to read hundreds of pages and several documents, to be able to take a decision on the basis of few statements. This paper describes both an artificial intelligence solution and a tool to provide a decision support system which could process documents and be capable to: (i) produce reliable suggestions, (ii) produce circumstantiated motivations, thus highlighting statements which could support identified suggestion focusing the work of any judge/court on actual statements and documents with relevant facts, and (iii) provide a web based tool producing suggestions and motivations on demand at service of the involved court and judges, compliant with privacy and security, as to data. To this end, AI and eXplainable AI technologies have been used and a solution has been obtained which meets the above-mentioned objectives and many other detailed requirements. Such a solution has been developed in the context of the research project “Giustizia Agile”, funded by the Italian National PON Governance and Institutional Capacity, and validated against real cases. The solution has exploited the Snap4City framework for data and AI/XAI management.

Keywords

• Artificial intelligence
• explainable AI
• mediation propensity assessment
• decision support system
• shortening justice procedures