Over 80 researchers, students and professionals gathered at NOVA IMS on 17 July for the 4th edition of the Data Research Meetup by MagIC, where leading experts discussed one of the most pressing questions facing artificial intelligence today: how can increasingly powerful AI systems become understandable, trustworthy and accountable?

Under the theme "Beyond the Black Box: Understanding, Interpreting and Explaining AI", the event brought together the Data Science community to explore recent advances in Interpretable and Explainable Artificial Intelligence (XAI), while fostering dialogue between researchers working across artificial intelligence, information systems and management.

The scientific programme opened with a keynote address by Professor Luca Longo, Associate Professor at University College Cork and Director of the Artificial Intelligence and Cognitive Load Research Lab, entitled "We Built AI, But Can We Understand It? The Explainability Imperative." Rather than presenting Explainable AI as a purely technical challenge, Longo argued that the field has rapidly evolved into an interdisciplinary research domain, bringing together expertise from computer science, engineering, psychology, neuroscience, philosophy and other disciplines. As AI systems become increasingly integrated into scientific research, healthcare, finance, education and public decision-making, understanding how these systems reach their conclusions is becoming as important as improving their predictive performance.

According to Longo, explainability should not be understood simply as generating explanations for algorithms, but as making AI systems understandable for the people who develop, regulate and ultimately rely on them. This shift requires moving beyond viewing AI as a "black box" towards designing systems that can communicate their reasoning in ways that are meaningful for different stakeholders.

Drawing on his recent work on Explainable Artificial Intelligence (XAI) 2.0, Longo outlined how the field is now confronting a series of open scientific challenges that extend well beyond existing explanation techniques. His research identifies 28 major research problems grouped into nine broad categories, ranging from developing explanations for generative AI and large language models, to evaluating explanation quality, adapting explanations to different users, mitigating misuse, and improving the societal impact of AI.

The keynote highlighted that future progress in AI transparency will depend not only on advances in machine learning, but also on integrating human-centred design, psychology, philosophy, governance and social sciences into the development of explainability methods. Rather than seeking a single universal solution, Longo advocated for explanation strategies that are tailored to different domains, contexts and audiences.

The discussion also explored emerging applications of Explainable AI across sectors including healthcare, finance, agriculture and education, illustrating how transparent AI systems can improve trust, accountability and informed decision-making in real-world settings.

– Professor Luca Longo

The keynote was followed by the roundtable "The Future of AI Transparency", moderated by Professor Vítor Santos (NOVA IMS), which brought together Professor Catarina Barata (Instituto Superior Técnico / ISR-Lisboa, LARSyS), Professor Nuno Lourenço (University of Coimbra, CISUC), Professor Cátia Pesquita (Universidade de Lisboa, LASIGE) and Luca Longo. The discussion explored current advances in Explainable AI, the importance of interdisciplinary collaboration and the future challenges of developing transparent AI systems.

The afternoon programme highlighted ongoing research conducted at MagIC through presentations by Karina Rebuli, Mijail Naranjo, Sofia Alves Pereira and Vítor Duarte Santos, illustrating how explainability and interpretability are being applied across machine learning, symbolic regression, digital government and emerging AI technologies.

The event also featured a PhD Poster Session, offering doctoral researchers the opportunity to present their work, exchange ideas with experts and receive scientific feedback. The Best PhD Poster Award was presented to Diogo Costa, recognising the quality and scientific contribution of his research.

Now in its fourth edition, the Data Research Meetup continues to establish itself as a platform where researchers from different disciplines can discuss emerging challenges at the intersection of Data Science and Management. By combining internationally recognised keynote speakers, national experts and research developed within MagIC, this year's edition reinforced the importance of interdisciplinary collaboration in addressing one of today's most important questions in Artificial Intelligence: how can increasingly complex AI systems become more transparent, understandable and accountable?