Testing software for non-discrimination: an updated and extended audit in the Italian car insurance domain

• DOI: 10.48550/arxiv.2502.06439
• Publication Date: 2025-02-10
• Authors: Marco Rondina; Antonio Vetrò; Riccardo Coppola; Oumaima Regragrui; Alessandro Fabris; Gianmaria Silvello; Gian Antonio Susto; Juan Carlos De Ma...
• Keywords: Software Engineering (cs.SE); FOS: Computer and information sciences; Computer Science - Software Engineering; Computer Science - Machine Learning; Artificial Intelligence (cs.AI); Computer Science - Artificial Intelligence; algorithmic bias;fairness;software audit;empirical methods; Computer Science - Human-Computer Interaction; Human-Computer Interaction (cs.HC); Machine Learning (cs.LG)

Context. As software systems become more integrated into society's infrastructure, the responsibility of professionals to ensure compliance with various non-functional requirements increases. These include security, safety, privacy, and, increasingly, non-discrimination. Motivation. Fairness in pricing algorithms grants equitable access basic services without discriminating on basis protected attributes. Method. We replicate a previous empirical study that used black box testing audit by Italian car insurance companies, accessible through popular online system. With respect study, we enlarged number tests and demographic variables under analysis. Results. Our work confirms extends findings, highlighting problematic permanence discrimination across time: significantly impact this day, birthplace remaining main discriminatory factor against individuals not born cities. also found driver profiles can determine quotes available user, denying equal opportunities all. Conclusion. The underscores importance for non-discrimination affect people's everyday lives. Performing algorithmic audits over time makes it possible evaluate evolution such algorithms. It demonstrates role engineering play making accountable.