FOIS 2024Keynotes

Keynotes

FOIS 2024 will feature the following world-class keynote speakers.

Mieke Boon: "Where to locate the explainability of explainable machine learning?"

This keynote address at FOIS 2024 delves into the epistemological aspects of AI explainability. Drawing on the perspective of Logical Empiricism, the talk examines Carl Hempel's covering law theory of scientific explanation, which posits that laws derived from inductive reasoning explain specific events. However, this account faces challenges, such as distinguishing between accidental regularities and genuine laws, leading to the exploration of explainability in machine learning (ML). While ML models provide accurate predictions, their opacity prompts the need for explanations behind these predictions. Explainable ML (XML) aims to extract reasons for predictions but may struggle to provide genuine explanations akin to shortcomings in Hempel's theory. Instead, the talk advocates for a mechanistic view of explanation, suggesting that ML explainability should be sought outside the ML models, within a conceptual model (CM) of the real-world phenomenon. The CM guides the construction of ML models, ensuring relevance of variables and facilitating explanations of outcomes. Consequently, ML models serve as instruments for quantitative analysis, while the CM serves as the explanatory framework, grounding explanations in the understanding of the target phenomenon. This approach shifts the focus from the inner workings of ML models to the broader conceptual understanding of the phenomena they seek to explain, providing a more robust foundation for AI explainability.

  • About Mieke

    Mieke Boon is professor of philosophy of science in practice at the Department of Philosophy of the University of Twente. She holds a PhD in chemical engineering and biotechnology. In 2006, she established a new movement, dubbed the philosophy of science in practice. Her research in this field focuses on epistemology, with particular interest in how human reasoning (i.e., their reasoning strategies) enables the construction of epistemic results (theories, models, concepts and phenomena, laws) that meets stringent epistemic and pragmatic criteria. Topics she has been publishing on are scientific models, representation, scientific instruments, phenomena, paradigms of science, scientific methodology, and epistemological responsibility. Current topics include AI, interdisciplinarity, and science education.

Frank van Harmelen: "Ontologies for Machine Learning"

After the breakthroughs of machine learning (deep learning or otherwise) in the past decade, the shortcomings of machine learning are also becoming increasingly clear: unexplainable results, data hunger and limited generalisability are all becoming bottle necks. As a result, the combination of data-driven techniques from machine learning with symbolic techniques from knowledge representation is now recognised as one of the grand challenges of modern AI. Surprisingly, the role of ontologies has been largely neglected in many of the neuro-symbolic systems that have been published. Fortunately, this is has been changing recently, and I will give an overview of a number of proposals from the literature for neuro-symbolic systems where ontologies play a key role.

  • About Frank

    Frank van Harmelen is full professor at the Vrije Universiteit Amsterdam in the Learning and Reasoning group. He has been a leading contributor to widely used semantic web techniques, including the web ontology language OWL and one of the first scalable triple stores (Sesame), which earned him the 10-year impact award of the International Semantic Web Conference. He is one of the editors of the Handbook of Knowledge Representation, the standard reference in his field, and he is member of the Royal Netherlands Academy of Sciences and of Academia Europaea.

Øystein Linnebo: "Constructional ontology and criteria of identity"

Gödel and others have suggested that a set can be regarded as constructed from its elements. Given any objects, we can apply the “set of” operation to construct the set of these objects. Inspired by Gödel’s suggestion, this talk presents a far more general constructional approach to ontology. First, I clarify what it is for some objects to be “constructed” from others. The key is that all truths about the “new” objects that are constructed must reduce to truths about the “old” objects on which the construction is based. Then, I explain the central role of criteria of identity in the constructional approach. When constructing “new” objects, it is particularly important to stipulate what it takes for them to be identical or distinct. A variety of examples are provided, including mereological sums, ordered pairs, as well as cardinal and ordinal numbers. Finally, I present a logical framework in which this constructional approach can be developed. In this framework, a large and natural family of forms of construction can be proved to be consistent. 

  • About Øystein

    Øystein Linnebo is Professor of Philosophy at the University of Oslo, where he is PI of the ERC Advanced Grant Construction in the Formal Sciences (2023-27). His main research interests lie in the philosophy of logic and mathematics, metaphysics, and early analytic philosophy (especially Frege). He has published about 70 scientific articles and is the author of three books, most recently Thin Objects: An Abstractionist Account (2018) and The Many and the One: A Philosophical Study of Plural Logic (with Salvatore Florio, 2021), both Oxford University Press.

Barend Mons: "Stop Data Sharing"

(shared keynote with ICBO)

The rapid developments in the field of machine learning have also brought along some existential challenges, which are in essence all related to the broad concept of ‘trust’. Aspects of this broad concept include trust in the output of any ML proces (and the prevention of black boxes, hallucinations and so forth). The very trust in science is at stake, especially now that paper mills come up that also aggravate the perverse reward systems in current research environments, which are stuck in 20th (in fact 17th) century scholarly communication. The other side of the same coin is that ML, if nor properly controlled will also break through security and privacy barriers and violate GDPR and other Ethical, Legal and Societal barriers, including equitability. In addition, the ‘existence’ of data somewhere by no means implies its actual Reusability. This includes the by now well established four elements of the FAIR principles: Much data is not even Findable, if found, not Accessible under well defined conditions, and if accessed not Interoperable (understandable by third parties and machines) and this results in the vast majority of data and information not being Reusable without violation of copyrights, privacy regulations or the basic conceptual models that implicitly or explicitly underpin the query or the deep learning algorithm. This keynote will address how ‘data visiting’ as opposed to classical ‘data sharing’, which carries the connotation of data downloads, transport and loosing control, mitigates most, if not all, the unwanted side effects of classical ‘data sharing’. For federated data visiting, the data should be FAIR in an additional sense or perspective, they should be ‘Federated, AI-Ready’, so that visiting algorithms can answer questions related to Access Control, Consent, Format, and can read rich (FAIR) metadata about the data itself to determine whether they are ‘fit for purpose’ and machine actionable (i.e. FAIR digital Objects, or Machine Actionable Units). The ‘fitness for purpose’ concept goes way beyond (but includes) information about methods, quality, error bars etc. The ‘immutable logging’ of all operation of visiting algorithms is crucial, especially when self learning algorithms in ‘swarm learning’ are being used. Enough to keep us busy for a while.

  • About Barend

    Barend Mons (born 1957, The Hague) is a molecular biologist by training and a leading FAIR data specialist. The first decade of his scientific career he spent on fundamental research on malaria parasites and later on translational research for malaria vaccines. In the year 2000 he switched to advanced data stewardship and (biological) systems analytics. He is currently a professor at Leiden and is most known for innovations in scholarly collaboration, especially nanopublications, knowledge graph-based discovery and most recently the FAIR data initiative and GO FAIR. Since 2012 he is a Professor in biosemantics in the Department of Human Genetics at the Leiden University Medical Center (LUMC) and since May 2022 also at the Leiden Academic Centre for Drug Research (LACDR) in The Netherlands. In 2015 Barend was appointed chair of the High Level Expert Group on the European Open Science Cloud. In 2017 Barend started the International Support and Coordination office of the GO FAIR initiative. From 2018 to 2023, Barend was elected president of CODATA the standing committee on research data-related issues of the International Science Council. Since 2021, Barend is the Scientific Director of the GO FAIR Foundation. In 2023 he became a fellow of the International Science Council. Barend is a frequent keynote speaker about FAIR and open science around the world, and participates in various scientific advisory boards of international research projects.