9:00 - 9:10 Welcome by the chairs

9:10 - 10:10 Keynote 1: Sharing sensitive data for machine learning: a repository perspective

Speaker: Zosia Beckles

Ms Zosia Beckles
University of Bristol

10:10 - 11:00 Paper Session 1

Paper 1: A Comparative Analysis on Machine Learning Techniques for Research Metadata: the ARDUOUS Case Study. Presenter: Dipendra Yadav                                                    

Paper 2: A Methodology and System For Big-Thick Data Collection. Presenter: Ivan Kayongo                                                                     

Paper 3: Addressing Privacy in Passive Data Collection for Nursing Documentation. Presenter: Farnod Bahrololloomi                                                                         

Paper 4: Challenges in Data Preservation for AI and ML Systems. Presenter: Gregory Tourte                                                                    

Paper 5: Understanding and Addressing User Needs for Annotation of Simple Sensor Data

Dipendra Yadav
University of Greifswald

Gregory Tourte
University of Oxford, UK

11:00 - 11:30 Coffee break

11:30 - 12:30 Panel 1: Data management and preservation in applied machine learning systems

The recent popularity of artificial intelligence and machine learning methods in a broad range of contexts has led to a significant increase in the amount of data and models being generated. Similarly, the popularity of deep learning methods leads to increasingly large and complex models, with significant training and processing requirements. To train and validate models requires data labels, annotations: information about the expected performance or predictions of the model. 

In this panel, we will explore the management and preservation of machine learning systems and data. What information would machine learners require to effectively find and reuse the data? How should we tackle the long-term storage of and access to very large datasets? How do we account for the risks and requirements, such as transparency, associated to the reuse of data in contexts where the dataset involves personal data, or special characteristics such as healthcare status? Most importantly, where do the responsibilities lie, and how can we work effectively across disciplines - research, engineering, user experience, regulatory compliance, library and archive - to maximise the responsible reuse of large ML datasets?

Panelists: Zosia Beckles, Gregory Tourte, Dagmar Waltemath, Emma Tonkin

Ms Zosia Beckles
University of Bristol

Gregory Tourte
University of Oxford, UK

12:30 - 14:00 Lunch break

14:00 - 14:30 Paper Session 2

Paper 6: Variability of annotations over time. An experimental study in the dementia-related named entity recognition domain. Presenter: Teodor Stoev

Paper 7: Assessing Large Language Models for annotating data in Dementia-Related texts: A Comparative Study with Human Annotators. Presenter: Sumaiya Suravee         

Teodor Stoev
Universität Greifswald

sumaiya suravee
University of Greifswald

14:30 - 15:30 Panel 2: Research Data Management from the AI perspective: capturing data and annotation, processing, validation, maintenance and preservation in real-world applications

AI and ML are in practical use in many domains ranging from enterprise to academia, in areas as diverse as warehouse logistics, medicine, Internet of Things and ubiquitous computing, digitisation, education and library provision. This has led to a large number of demonstrators, which often result in mature toolsets in production environments, but just as often do not. This panel explores themes surrounding data management through the lens of experience in these application areas. 

In this panel, we will examine questions like: what application development process is used to create and maintain these tools throughout their lifecycle? What values and principles guide the development and maintenance process? How can FAIR data principles and open data concepts be translated into ML and AI contexts, and into contexts which involve significant privacy concerns? How can interdisciplinary knowledge transfer be promoted - what approaches and standards are useful (for example, data and model cards?) Interdisciplinary work may be viewed as key to enabling development processes that effectively combine engineering, participant, legal and privacy concerns: what practices can most effectively support the establishment of effective cross-disciplinary teamwork? Finally, what industry, academic or professional organisations have been useful in your journeys through this topic, and what would you recommend as a resource for a new starter in the area?

Panelists: Kristina Yordanova, Christopher Reining, Max Schröder, Frank Krüger

Max Schröder

15:30 - 16:00 Coffee break

16:00 - 17:00 Keynote 2: Navigating the EU AI Act. An introduction to the EU Regulation on Artificial Intelligence.

Speaker: Patrick Brunner

17:00 - 17:30 Summary and closing words