Frontiers in AI

Abstract: In this talk, I describe lessons learned in nearly 20 years of evaluating explanation interfaces with people. In my work, I have observed that tailoring the explanation to the user (e.g., a domain expert or a layperson) and task (e.g., decision support or model improvement), or context (e.g., under time pressure) is necessary for meaningful assessment of explanation quality (e.g., correct decisions or better understanding).
Learning from trends in empirical research methods (natural language processing, information retrieval, and machine learning), I further argue that this is an issue for both human-computer interaction and machine learning. Real-world factors influencing the performance of systems are at best implicitly encoded when inferring probabilities from observations, and at worst, no longer applicable to the settings for which they are used. In seeking a balanced approach between this reality and a pragmatic, data-driven and (by necessity) reductionist approach, I make constructive suggestions for evaluating the quality of explainable artificial intelligence.

Bio: Prof. Nava Tintarev is a Full Professor in Explainable AI at Maastricht University in the Department of Advanced Computing Sciences (DACS), where she also serves as the Director of Research. Her interdisciplinary work bridges computer science and human-centered evaluation, focusing on making AI systems more transparent and increasing user control. Prof. Tintarev specializes in developing interactive explanation interfaces for recommender systems and search technologies, emphasizing user empowerment and decision support. She is a lab director of the ICAI TAIM lab, working on trustworthy AI in media, and is an incoming board member of the Informatics Platform Nederlands. She was a founding principal investigator of ROBUST, a €87 million Dutch national initiative advancing trustworthy AI. Her research has also been supported by major organizations including IBM, Twitter, and the European Commission. Recognized as a Senior Member of the ACM in 2020, Prof. Tintarev has received multiple best paper awards for her research contributions. Her work underscores the importance of transparency and user agency, aiming to support better decision-making with AI.

Abstract: As machine learning systems are increasingly deployed in safety-critical settings, ensuring alignment with formal requirements has become a key challenge. Two main strategies have emerged to address this: a priori design, which integrates the requirements directly into the model’s architecture or training process, and a posteriori verification, which checks whether trained models satisfy the desired properties.
In this talk, I argue that these approaches are not at odds but mutually reinforcing. Enforcing constraints during training enables correctness by design and reduces the need for costly post hoc checks. Yet, verification remains essential when requirements are difficult to encode in a differentiable way. To support this perspective, I present a family of neuro-symbolic methods that embed logical requirements directly into the network topology, highlighting both their strengths and limitations.

Bio: Dr. Eleonora Giunchiglia is a Lecturer at Imperial College London’s I-X initiative and the Department of Electrical and Electronic Engineering. Her research focuses on enhancing the safety and trustworthiness of deep learning systems by integrating formal logical constraints into their design. Dr. Giunchiglia earned her PhD from the University of Oxford in 2022, where she explored the intersection of machine learning and logic. She subsequently held a postdoctoral position at TU Wien before joining Imperial College in 2024. Her work has led to the development of frameworks like PiShield and CCN+, which facilitate the creation of neural networks that comply with predefined constraints. These contributions have significant implications for fields such as autonomous driving and synthetic data generation. Beyond her research, Dr. Giunchiglia serves on the editorial board of the Neuro-Symbolic AI journal and is a junior member of the Future of Life Institute. Her commitment to AI safety underscores her dedication to developing technologies that are both innovative and aligned with human values.

Abstract: Benchmarking is an important driver of (both real and illusionary) progress in many subfields of machine learning (ML) research. In this talk we explore the challenges posed by benchmarking fairness-aware machine learning (fairML). FairML benchmarks inherit many of the difficulties of general ML benchmarks, but also pose some unique challenges. Most importantly, the inherent contextual and contestable nature of a social concept like fairness renders it difficult to quantify. Consequently, framing fairness as a black-box optimization problem of fairness and accuracy is misleading and can lead to undesirable side effects or even harm those fairML interventions are intended to protect. We therefore call for a shift in focus from competitive benchmarks to evaluations which facilitate a deeper understanding of what fairML interventions (do not) achieve and why.

Bio: Prof. Dr. Mykola Pechenizkiy is a Full Professor and Chair of the Data Mining group at the Department of Mathematics and Computer Science, Eindhoven University of Technology. His technical AI research focuses on predictive analytics and knowledge discovery from evolving data, with applications in industry, healthcare, and education. Their research was recognized with the IEEE ICDE 2023 Best Demo Award, IDA 2023 Runner-up Frontier Prize, IEEE DSAA 2022 Best Paper Award, LoG 2022 Best Paper Award, and ALA 2022 Best Paper Award. Since 2025 he serves as a founding Director of the Center for Safe AI where he leads the trustworthy and responsible AI interdisciplinary research program aiming to address both technical and socio-technical challenges.

Bio: Nadin Kokciyan is an Associate Professor (Reader) in Artificial Intelligence in the School of Informatics at the University of Edinburgh; and a Senior Research Affiliate at the Centre for Technomoral Futures, Edinburgh Futures Institute. Her research interests include human-centered AI, Privacy, Argument Mining, Responsible AI and AI Ethics. She currently leads the Human-Centered AI Lab (CHAI Lab) and is actively involved in initiatives that bridge technical AI development with ethical and societal considerations. Nadin is currently a Turing BridgeAI Associate Advisor at the Alan Turing Institute to provide specific guidance to organisations in Ethics of AI. Nadin earned her PhD in Computer Engineering from Boğaziçi University, where she developed an AI-based privacy management framework to assist humans in making decisions. She has held a postdoc position at King’s College London, where she worked on the development of an AI-based health management tool to assist users in managing their health by using argumentation techniques. Nadin regularly serves on the program committees for leading AI conferences such as AAMAS, IJCAI, AAAI and ECAI. She has dedicated over a decade of her career to the realm of academia. Her expertise is in using AI to develop decision-support tools for humans to make informed decisions. In 2021 and 2025, she served as a guest editor for special issues on ‘Sociotechnical Perspectives of AI Ethics and Accountability’ and ‘Humans Meets AI’ in IEEE Internet Computing.​