Abstract: In this talk, I would like to introduce a paradigm shift in the way we manage and share health data. With the emergence of the Data Governance, the Data Act and mainly the European Health Data Space regulations, the focus on health data in Europe has very much been on “Data altruism”, as regulated in the Data Governance Act. While Data altruism is good for the “common good” it does not bring anything to “me” as a patient, and it does not help my physician to develop preventive and personalized medicine for “me”.  I will argue that the best way to support the common good, is to start with “me”, with my data. By curating my data across multiple messy data sources into an interoperable and reusable longitudinal health record (a personal knowledge graph), I can then share high quality data for the common good while securing quality data for my clinical care. And this enables to replace the “curate health data anew for each clinical research question (and then dump them)” principle by the “curate once, use many times” paradigm. 

Bio: Isabelle de Zegher, MD, MS, is founder of b!loba, focusing on health data sharing solutions and currently the clinical coordinator of the AIDAVA Horizon Europe project. Isabelle is a Medical Doctor from the University of Louvain and a Master of Sciences in Medical Information Sciences from Stanford University. She has extensive experience in health information technology, data standards, data integration, semantic interoperability, and application of AI in clinical research. 

Abstract: Recommender systems offer users recommendations of items to buy. Recommender systems use algorithms that can be categorized as ranking algorithms, which result in an increase in the prominence of some information. Other information is filtered and not shown to people. However, it is often not clear to users whether the recommender system’s advice is suitable to be followed, e.g., whether the right information was taken into account. This keynote will describe how explanations can help bridge that gap, what we constitute as helpful explanations, as well as the considerations and external factors we must consider when determining the level of helpfulness. I will introduce the notion of interactive explanation interfaces for recommender systems, which support exploration rather than explaining recommendations post-hoc. I will also address the importance of user modeling in the design of explanation interfaces. Both the characteristics of the people and the situation place different requirements on which explanations are useful, also w.r.t. to different presentational choices.  Simply put, “One size does not fit all”.

Bio:  Nava Tintarev is a full professor of Explainable Artificial Intelligence at the Maastricht University and a visiting professor at TU Delft. She conducts research in the field of human-computer interaction in artificial advice giving systems. She has published over 80 peer-reviewed papers in top human-computer interaction and artificial intelligence journals and conferences.

Abstract: Natural Language Processing has discovered the trove of legal text corpora that are now being transformed into ‘law as data’. A wide variety of legal technologies is being developed purportedly to make the administration of justice and the practice of law more efficient and more effective. In this talk I will discuss the gap between the claims made on behalf of these technologies and the substantiation of such claims, while fleshing out how upstream design decisions may impact downstream deployment and its impact on law and the rule of law. The talk will refer to the web tool that was developed by a team of lawyers and computer scientists as a method, a mindset and a resource to help lawyers and developers to better understand what is at stake. This should allow lawyers and developers to develop a ‘by design’ approach to legal protection, thus warding off attempts to replace the force of law with the force of technology. 

Bio: Mireille Hildebrandt is a research professor on interfacing law and technology at Vrije Universiteit Brussels, and a part time chair at the CS department at Radboud in Nijmegen as a full professor. ​Her research is on the implications of automated decisions, machine learning and mindless artificial agency for law and the rule of law in constitutional democracies. She has published 5 scientific monographs, and over 100 chapters and articles in scientific journals and volumes.

Abstract: Big data has the potential to advance and facilitate clinical decision making for care providers. Data, however, is heterogenous and spread across different providers and cannot be easily shared due to patient privacy protection regulations. Various standardization and novel learning techniques have been proposed to overcome the diversity of medical data and the limitations related to its access, such as transfer learning and the different forms of distributed learning.

Bio: Fadila is a Clinical Research Scientist at Median Technologies. Her research interests are centered around the application of distributed learning to preserve medical data privacy and facilitate the use of AI to build new innovative imaging biomarkers.

Abstract: The one thing that has really changed, in recent years, is that the applications of AI are starting to have a real impact on business.  An increasing number of models are moving from the lab into the industry and new challenges are being revealed in this process. In this talk, we would like to share how the energy industry is using data to reinforce critical business processes and support the energy transition. We will focus on a few machine learning models that are used to optimize the maintenance and to maximize the production of wind power plants, and a new project, Long Term Forecast, which allows to make precise predictions of energy consumption and to offer fixed contracts to the clients (this way saving them a lot of money). In the end we would like to talk about the industrialization of these projects, highlighting the difference between running a model on a perfectly clean sample “toy” data set on a computer and running it on a huge real (usually very dirty) data set completely automatically on schedule. A lot of people are not aware that around 90% of effort is made in industrialization after the model has been created. Here we will explain why this is a big problem, what are the challenges and the steps needed to industrialize the model successfully.

Bio: Cristiana Pompei is IT Deputy Director Application Delivery working at Luminus, a Belgian Energy company based in Brussels, who spent the last 3 years industrializing data solutions with focus on AI, ML, Business Intelligence and Data Governance and she is now owner of Data Transformation Program. Her background is in Information Technology and she previously spent 15 years in Edison, an Italian Energy company based in Milan, covering managerial positions and supporting different business departments in digitalization and realization of complex IT systems. She was a candidate of EDF Leadership Program on Innovation and Entrepreneurship in INSEAD and she has a master degree in Industrial Engineering.

Bio: Jelena Grujic is a Data Scientist and Engineer working for a Data Engineering consulting company DataMinded, based in Leuven, Belgium who spent almost 2 years on a mission at Energy Company Luminus. Her background is in Physics and she previously spent 13 years in Academia, working in the area of Data Science, Complex Systems, Game Theory, Machine Learning etc. She worked at Imperial College London, Vrije Universiteit Brussel and has a PhD from Univerisdad Carlos III de Madrid. She is a founder of the Brussels chapter of the Data Science event series called DataBeers, running since 2016.

Bio: Anabel is a data scientist and performance engineer at the second largest Belgian energy company, Luminus. With a background in electromechanical engineering, she started her career as a young graduate trainee, which allowed her to experience many different aspects of the energy sector. She now works in the wind department where she monitors the performance of the fleet of wind

turbines. It is to have better insights in the operation of these assets that her and the team surrounding her develop models and tools based on machine learning.

Outside of business hours she loves to go for a run in the woods or to travel around Europe in her camping car.