Algorithmic bias mitigation is one of the most difficult conundrums for the data science community and machine learning (ML) experts. In recent years, researchers have devoted huge efforts to the field of fairness in ML. Although progress has been made toward identifying biases and designing fair algorithms, translating them into industrial practice remains a major challenge. In this seminar, we report the results of an open innovation project in the banking sector and we propose a general roadmap for fairness in ML. We also present the implementation of a toolkit called BeFair that helps to identify and mitigate bias. Results show – among other things – that training a model without explicit constraints may can exacerbate bias in the predictions.
Riccardo Crupi is a mathematical engineer and a graduate of Polytechnic University of Turin, who is employed as a data scientist at the Big Data Lab of Intesa Sanpaolo. He has been working on the development of the predictive Net Promoter Score model and on the implementation of fairness in AI algorithms. Before joining Intesa Sanpaolo, he worked for three years in Comau Innovation Hub (Turin) as a data scientist, and in the same period he obtained a master’s degree in industrial automation.
Daniele Regoli is a data scientist at the Big Data Lab of Intesa Sanpaolo. His background is in physics, with a PhD from the University of Bologna and a master’s degree in mathematical finance. Before joining Intesa Sanpaolo, he worked for several years as researcher, first at the Math Department of University of Padova and then at the Scuola Normale Superiore in Pisa, mainly on mathematical models applied to finance and economics.
This event is part of the Industry talks series organised by the Technological Change and Society cluster. These events feature industry or government representatives sharing the experience of their organizations with technological innovation. We discuss the impact of technologies from the viewpoint of the actors effectively developing and/or implementing them. We also intend to discuss how developers and adopters think about technological and societal risks arising from innovation. These meetings are open to EUI members only.
The event is open to EUI members and to externals upon invitation.