RW 2025
The 21th Reasoning Web Summer School
İstanbul, Türkİye
25 - 28 September 2025
program
The program of RW 2025 will comprise eight tutorials delivered by lecturers who are experts in the field. Attendees will become acquainted with both foundational and practical aspects of declarative AI, covering a variety of different logic-based formalisms (Datalog, answer set programming, description logic ontologies...) and reasoning paradigms (temporal reasoning, inconsistency-tolerant reasoning, explanation...), as well as how declarative AI techniques can be combined with machine-learning approaches.
Please see below for details on the tutorial topics and lecturers!
Inconsistency-Tolerant Semantics Based on Preferred Repairs
Camille Bourgaux, CNRS, DI ENS, France
Abstract: Real-world datasets are plagued by data quality issues which may render the data inconsistent w.r.t. a set of constraints, be they given by database dependencies or ontologies. A prominent way to handle such inconsistent data is to use inconsistency-tolerant semantics to obtain meaningful answers to queries. Most of these semantics are based on some notion of repairs, which represent ways of restoring the data consistency. The most basic kind of repairs is that of subset repairs, which are maximal consistent subsets of the dataset. However, in many scenarios, one can define preferred repairs based on some preference information. In this tutorial, I will present repair-based semantics, then review different kinds of preferred repairs that have been considered in the literature.
Bio: Camille Bourgaux is a researcher at the CNRS (French National Center for Scientific Research), based at the DI ENS lab in Paris, France. She obtained her PhD in 2016 from University of Paris-Saclay, and was a post-doctoral researcher at TU Dresden and Télécom ParisTech before joining the CNRS in 2018. Her research domain is knowledge representation and reasoning, and her research interests include in particular ontology-mediated query answering, handling of incomplete, uncertain or inconsistent data, and description logics.
Human-Centered ASP Applications: Representation & Reasoning
Esra Erdem, Aysu Bogatarkan, & Muge Fidan, Sabanci University, Istanbul, Turkey
Abstract: As the definition of artificial intelligence (AI) changes towards building rational agents that are provably beneficial for humans, knowledge representation and reasoning (KRR) plays an important role in addressing the user-oriented challenges in applications, such as generality, flexibility, provability, hybridity, bi-directional interactions, robustness, and explainability. In this tutorial, we will introduce participants to modeling and solving problems in human-centered real-world applications using KRR methods and tools provided by answer set programming, while addressing such challenges for AI.
Bios: Esra Erdem is a professor in computer science and engineering at Sabanci University. She received her Ph.D. in computer sciences at the University of Texas at Austin, and carried out postdoctoral research at the University of Toronto and Vienna University of Technology. Her research is in the area of artificial intelligence, in particular, the mathematical foundations of knowledge representation and automated reasoning, and their applications to various domains, including robotics, bioinformatics, logistics, and economics. Dr. Erdem was a general co-chair of ICLP 2013, a program co-chair of ICLP 2019, KR 2020 and PADL 2025, the general chair of KR 2021, and the president of KR Inc. She is an associate editor for Artificial Intelligence (AIJ) and Theory and Practice of Logic Programming (TPLP).
Aysu Bogatarkan is a PhD candidate in Computer Science and Engineering at Sabanci University. She earned her BSc degrees in Computer Science and Engineering and Mechatronics Engineering from Sabancı University. Her current research focuses on developing flexible, explainable, and robust solutions to multi-agent path finding problems, utilizing knowledge representation and reasoning methods. She is interested in demonstrating the applicability and usefulness of her methods in real-world applications, such as robotics and logistics.
Müge Fidan is a PhD candidate in Computer Science and Engineering at Sabancı University. She received her BSc and MSc degrees in Mathematics from Bilkent University. Her research primarily focuses on the application of knowledge representation and reasoning methods to solve matching problems, particularly in real-world scenarios that consider user-specific preferences and constraints.
Explaining Reasoning Results for Description Logic Ontologies
Patrick Koopmann, Vrije Universiteit Amsterdam, Netherlands
Abstract: The Web Ontology Language (OWL), grounded in description logics, enables reasoning systems to infer implicit knowledge in a transparent manner. However, the expressivity of OWL and the complexity of many ontologies often result in reasoning outcomes that are not immediately intuitive. Explanations of these outcomes are essential for users to understand ontology content, communicate its structure and behavior effectively, and debug undesired or missing inferences. This tutorial provides an overview of advanced explanation techniques that have been developed in the recent years, such as proofs, counterexamples and abduction, and have a closer look at theoretical properties and practical methods for different types of explanations for description logic reasoning results.
Bio: Patrick Koopmann completed his PhD at the University of Manchester and subsequently held postdoctoral positions at the University of Oxford and the Technische Universität Dresden. He is currently an assistant professor at the Vrije Universiteit Amsterdam, where he teaches courses on logic and symbolic AI as part of the artificial intelligence program. His research focuses on knowledge representation, automated reasoning, and theoretical computer science, with a particular emphasis on description logics and related formalisms. A central theme of his current work is the development of techniques to explain reasoning results in description logics, and to suggest fixes in case of undesired or missing reasoning results. To address these questions, he investigates novel ways of computing proofs & counterinterpretations, abduction, and interpolation, examining both theoretical foundations and practical implementations. Additionally, he is involved in the development of graphical frontends to these algorithms, to make these techniques accessible and enhance the ontology development experience for users.
Modern Datalog: Concepts, Methods, Applications
Markus Krötzsch, TU Dresden, Germany
Abstract: Pure Datalog is arguably the most fundamental rule language, elegant and simple, but also often too limited to be useful in practice. This has motivated the introduction of many new expressive features, ranging from datatypes and related functions, over aggregates and semi-ring generalisations, to existential quantifiers and complex terms. In spite of their variety, all these approaches remain true to the nature of Datalog as a direct, pattern-based way of computing on structured data. We therefore find that a modern notion of Datalog is emerging, distinctly different from other approaches of logic programming and with its own set of related methods and applications. In this course, we will introduce Datalog and its most common extensions, and explain when and how these features can be used together (which is often, but not always, safe to do). We will further look at modern Datalog systems and some of their primary use cases. Hands-on work with Datalog and its extensions is done through with the free Datalog engine Nemo. The course is free to all audiences and does not assume specific prior knowledge.
Bio: Markus Krötzsch is a full professor at the Faculty of Computer Science of TU Dresden and director of the SECAI Konrad Zuse School of Excellence in AI. He obtained his Ph.D. from the Institute AIFB of Karlsruhe Institute of Technology (KIT) in 2010, and thereafter worked at the Department of Computer Science of the University of Oxford until October 2013, before founding the Knowledge-Based Systems group at TU Dresden. His primary research contributions are in rule languages, ontologies, and the development and analysis of knowledge modeling languages (including the W3C OWL standard), inference methods, and automated reasoners. He is also a co-founder of the influential knowledge graph Wikidata. He has given many summer school lectures and tutorials related to these topics at leading international events.
From One-Level to Multi-Level Ontology-Based Data Management
Antonella Poggi, Sapienza University of Rome, Italy
Abstract: Ontology-Based Data Management (OBDM) is a well-founded paradigm to manage large amount of data through an ontology, i.e., a unique shared conceptual view of a domain of interest, abstracting from the actual data organization and distribution. Besides an ontology and a set of data sources, an OBDM system comprises a so-called mapping, that is a declarative specification of the relationship between the ontology and the data. Multi-level OBDM is the rather unexplored OBDM setting where both ontologies and mappings are multi-level. An ontology is multi-level if instead of resorting to a clear distinction between elements on which it predicates, typically represented as individuals, and predicates themselves, represented by classes (relations) to which individuals (pairs thereof) belong, it deals with predicates as domain elements, whose features and properties are of interest. Thus, multi-level ontologies comprise meta-classes and meta-properties, where a meta-class is a class whose instances can be in turn classes, and a metaproperty is a property (or relation) whose instances are properties holding between classes, rather individuals. A mapping is multi-level if besides specifying which individuals are derived from data, it further specifies which predicates are derived from data, thus enabling the OBDM system to capture contexts where classes and relations may depend on data and vary as the data change. This course provides an overview of the major results related to the traditional OBDM setting, which we call “one-level”, and then concentrate on recent advancements concerning the multi-level OBDM setting.
Bio: Antonella Poggi is an Associate Professor in Computer Engineering at Sapienza University of Rome. Her research interests include: knowledge representation and reasoning, and more specifically, knowledge graphs and Description Logics ontologies; database theory, and more specifically, data integration and exchange, data provenance, and semi-structured data. She obtained her PhD degree at both Sapienza and University of Paris-Sud, defending a thesis about the integration of structured and semi-structured data. Her thesis led to the journal paper “Linking data to ontologies”, appeared in 2008 in the Journal on Data Semantics, which, according to Google Scholar, has more than 1100 citations (still 203 in 2024). She is (co)author of 79 publications indexed within DBLP, in journals, books chapters, international conferences, and workshops, including some of the most prestigious ones in the above mentioned areas, such as the Journal of ACM and Artificial Intelligence, and the conferences ACM PODS, IJCAI, AAAI, and KR. In 2017, she co-founded the start-up “OBDA Systems Srl”, which originates from the research of the group she belongs to at Sapienza.
ASP Essentials: Modelling & Efficient Solving
Francesco Ricca & Giuseppe Mazzotta, University of Calabria, Italy
Abstract: Answer Set Programming (ASP) is a logic-based Knowledge Representation and Reasoning (KRR) paradigm that facilitates rapid prototyping of solutions for complex problems. It is particularly effective for tackling Deep Reasoning tasks involving exponentially large search spaces, such as combinatorial search and optimization. While getting started with ASP is relatively easy, mastering its advanced constructs and scaling solutions to real-world problem sizes can be challenging. This course will introduce participants to the usage of ASP, from computing answer sets with basic ASP systems to addressing more advanced scenarios where standard modelling and solving techniques may fall short. Additionally, it will briefly explore the usage of LLMs with ASP highlighting how these technologies can complement each other.
Bios: Francesco Ricca is Full Professor of Computer Science at the Department of Mathematics of the University of Calabria, Italy. He received his Laurea Degree in Computer Science Engineering (2002) and a PhD in Computer Science and Mathematics (2006) from the University of Calabria, Italy. Francesco Ricca’s research interests belong to the AI area of knowledge representation and reasoning. His particular research focus lies on Answer Set Programming. Francesco Ricca is co-author of more than 150 refereed research articles published in international journals (40+), collections, and conference proceedings. He is member of the Executive Board of the Italian Association of Logic Programming, and member of the steering committee of the Rule and Reasoning Association (RRA) and member of the Steering Committee of the Italian Association for Computational Logic (GULP) He is Editorial Advisor for the Journal of Theory and Practice of Logic Programming, and Editorial board member of Intelligenza Artificiale. He was program (co-)chair of ICLP20, RuleML+RR18 and is general chair of ICLP25.
Giuseppe Mazzotta is a Post-Doctoral Researcher in Computer Science at the Department of Mathematics, University of Calabria, Italy. He obtained his MSc Degree in Computer Science in 2020 and completed his PhD in Computer Science and Mathematics in 2023 at the same institution. Giuseppe's research focuses on knowledge representation and reasoning, particularly in Answer Set Programming (ASP). During his PhD, he specialized in developing efficient techniques for evaluating ASP programs affected by the grounding bottleneck problem. His work has been published in international conferences, earning him recognition such as the "AAAI Outstanding Student Paper Honorable Mention" at the 36th AAAI Conference on Artificial Intelligence.
Neuro-Symbolic Artificial Intelligence
Luciano Serafini, Fondazione Bruno Kessler, Trento, Italy
Abstract: Neuro-Symbolic Artificial Intelligence (NeSy) is a field of AI that explores how neural network architectures can be integrated with symbolic systems to solve complex tasks that require both numerical computation and symbolic manipulation. The current state-of-the-art in NeSy encompasses a wide range of approaches, depending on the types of neural architectures and symbolic systems integrated, making it almost impossible to provide a complete overview of the field. In this tutorial, we focus specifically on NeSy systems that combine feedforward neural networks with logical systems based on propositional and first-order languages. In these systems, gradient-based learning is integrated with logical reasoning to address tasks such as data classification and link prediction with logical constraints, symbolic rule learning, symbol learning, data clustering with logical constraints, and query-answering on high-dimensional data. The tutorial aims to offer a unified perspective on these different systems.
Bio: Luciano Serafini is the Head of the Data and Knowledge Management Research Unit at the Fondazione Bruno Kessler (FBK), a position he has held since 2007 doing research in Artificial Intelligence, with a focus on areas such as logic-based knowledge representation, neuro-symbolic integration, and embodied AI. He is also an adjoint professor at the University of Padova, where he teaches knowledge representation and learning. Since 2021 he is EurAI Fellow. With a career spanning over 35 years, Luciano has made significant contributions to the field of AI, including the development of Multi-Context (MC) Systems for modular knowledge representation, innovations in quantum computing languages, and the integration of machine learning with logical reasoning. He is currently working on an innovative approach called Planning and Acting to Learn (PAL).
Reasoning About Time in DatalogMTL
Przemyslaw Walega, Queen Mary University of London, United Kingdom
Abstract: DatalogMTL is a powerful extension of Datalog, designed to handle complex temporal reasoning. In this framework, a temporal dataset consists of facts that hold over intervals along a rational timeline. A DatalogMTL program enables recursive reasoning over these facts, by incorporating operators known from metric temporal logic (MTL). This results in an expressive language that enhances Datalog's capabilities and opens the door to a range of potential applications. While the temporal aspect introduces significant computational challenges, several reasoning approaches for DatalogMTL have been proposed, with some already seeing implementation. DatalogMTL has also been further extended with features such as non-monotonic negation, existential rules, and temporal aggregation, introducing even more complex behaviours. During the tutorial, I will introduce the line of research on DatalogMTL. I will discuss properties of DatalogMTL, focusing on the reasoning algorithms and analysis of computational complexity.
Bio: I am an Associate Professor (Senior Lecturer) in Queen Mary University of London, Centre for Fundamental Computer Science. I was a Senior Researcher in University of Oxford, Department of Computer Science and received PhD in Logics in the Institute of Philosophy at the University of Warsaw. My research is devoted to designing formal logical languages, studying their computational properties, and developing efficient reasoning algorithms for them. I am especially interested in methods for complex reasoning about time. Time is ubiquitous in our everyday lives, in the way we perceive and reason about the surrounding world, as well as how our AI algorithms do it. The topic of time brings together computer scientists, mathematical logicians, and philosophers, which makes this research area really fascinating to me.
Endorsed by:
