Dear Students, the following talk that will be held at the Department of Information Engineering is related to NLP and might be of interest to you.
Date: May 26th, 12:30, Aula Magna A. Lepschy
Speaker: Giuseppe Attanasio, Institute of Telecommunications Lisbon
Title: Technical Scaffolding for Ethical AI: Explaining, Building, and Scrutinizing Models Under Social Axes
Abstract: This talk reflects on three years of research conducted to scrutinize modern language, vision, and speech AI models and their social implications. It presents technical and practical tools to explain model outputs, enhance user safety and fairness across diverse social groups. Taking a meta-analytical approach, it explores the socio-technical pitfalls of evaluation metrics–one of the core drivers of AI advancement. Finally, drawing on experience developing both mono- and multimodal systems, it reflects on the key challenges in creating future AI that brings social groups together, rather than dividing them.
Short bio: Giuseppe is a postdoctoral researcher at the Institute of Telecommunications in Lisbon. He focuses on developing ethical and safe AI technologies. His work explores the intersections of technical design, societal biases, and AI safety. Giuseppe's research in language, vision, and speech systems has been published in leading machine learning and natural language processing venues and journals (*ACL, ICLR, ICASSP, INTERSPEECH, Nature, ACM Transactions).