27TH EUROPEAN CONFERENCE ON ARTIFICIAL INTELLIGENCE

This tutorial will explore the promises and shortcomings of the field of explainable AI. Through the lens of post-hoc explanation methods and interpretable-by-design models, attendees will learn which answers to expect when dealing with explainable AI techniques.

This tutorial introduces attendees to the principles and best practices of the Data-Centric AI paradigm in a hands-on approach. We’ll cover topics such as data complexity, missing data, imbalanced data, data fairness, data privacy, and data explainability, discussing suitable techniques and open-source tools, and illustrate applications in real-world domains such as healthcare and finance.

This half-day tutorial delves into the transformative role of Agent AI in finance, covering the development, deployment, and evaluation of AI agents. It’s tailored for a wide audience, from PhD students to seasoned researchers, focusing on multimodal understanding, decision-making, and human-centric financial applications.

This tutorial aims at presenting automated planning to a broad audience, with a focus on “rich” objectives. Going beyond “classical planning”, the attendants will learn about multi-agent, flexible, temporal, epistemic and contingent aspects of automated planning.

The course offers an introduction to some recent advances in formal verification of intelligent
agents and multi-agent systems. The focus is on accessible presentation and simple examples, without going too deep into the involved mathematical machinery.

The 1985 paper by Carlos Alchourrón (1931–1996), Peter Gärdenfors, and David Makinson (AGM), “On the Logic of Theory Change: Partial Meet Contraction and Revision Functions” was the starting-point of a large and rapidly growing literature that employs formal models in the investigation of changes in belief states and databases.
In this tutorial, the 40 years of the logic of theory change (belief change or belief revision) are summarized.

Automated Reinforcement Learning aims to automatically make optimal design decisions for reinforcement learning agents to learn quickly and reliably. This tutorial will introduce the field, its methods, and provide hands-on examples of AutoRL tools to state-of-the-art reinforcement learning algorithms.

This tutorial aims to support ontology developers to reduce the time consumed in ontology implementation by generating graphical ontology conceptualizations that can be converted into code automatically. Participants will gain knowledge about the Chowlk visual notation for ontologies, how to use the Chowlk converter to accelerate ontology implementation based on the generated conceptualizations, how to use the related resources as templates and, finally, how to create new templates.

This tutorial introduces the framework of Causal Fairness Analysis, which can be used to address issues of fairness and equity in AI-driven decision-making systems. The key idea of the framework is to link disparities observed in the data to the underlying causal mechanisms, and provide formal algorithmic tools to detect and mitigate bias in AI systems.

We provide an overview of Explainable Constraint Solving within the context of eXplainable AI (XAI), focusing on providing explanations for constraint (optimization) problems and their solutions. We categorize explanations in deductive and contrastive explanations, reviewing old and new techniques and showcasing them in a live demonstration and hands-on experimentation on the well-known nurse rostering problem.

We will explain how to develop guarantees of robustness for stochastic models of multi-agent systems. We will present motivating applications of multi-agent systems, including social sensing (such as searching for missing persons or testing for COVID), shared mobility (such as Uber), and virtual power plants, where stochastic calculi seem crucial in modeling the response of the agents.

Answer set programming is a form of logic programming used in many areas of science and technology. This tutorial will describe the use of mathematical logic and automated reasoning in the process of designing and optimizing programs within this paradigm.

Knowledge Graphs are essential for many AI tasks as a reliable and structured data source, but their generation is not usually a straightforward process. In this tutorial, the attendees will learn best practices on how to construct KGs for any source through a sustainable and maintainable workflow.

The tutorial introduces the novel Hybrid Neural Networks (HyNNs), which combine a CNN branch for synthetic images and an MLP branch for tidy data. The practical sessions will cover methods to convert tabular data into synthetic images, implementation of HyNNs in classical machine learning projects, and a hands-on challenge where participants will compete to develop the most robust HyNN model using provided datasets.

Handling multiple objectives is a salient and ubiquitous characteristic of many, if not most, real-world decision problems. This tutorial explores the challenges and solution concepts for multi-objective decision-making through the lens of multi-agent and population-based learning approaches.

This tutorial focuses on the integral role of computational argumentation in AI, illustrating how it aids in reasoning with incomplete information and evaluating the epistemic state of discussed topics. The half-day session aims to deliver a dynamic learning experience through lectures, games and demos, bridging the logical foundations of argumentation with human-like reasoning to enhance transparent decision-making in AI systems.

This tutorial will cover the use of quantum annealing to model and solve constraint satisfaction and constrained optimization problems. We will present various techniques to model problems in QUBO (Quadratic Unconstrained Binary Optimization), the input language for quantum annealing computers.

The aim of this tutorial is to discuss the genesis, state of the art and prospects of an emerging topic known as Quantum Artificial Intelligence. This research area combines quantum computing and artificial intelligence by considering two possible directions: quantum computing used to improve the implementation of artificial intelligence techniques, and artificial intelligence methods used to address quantum computing challenges.

The tutorial introduces the audience to the concept of referring expressions, formulæ that can be used to communicate identities of otherwise abstract objects. The formalism provides foundations for a successful and unambiguous exchange of information about individuals between agents sharing common knowledge about such individuals, a task that is indispensable in most modern applications of knowledge representation and semantic technologies.

Communicating correctly and timely about AI is too important to be neglected by AI researchers. This tutorial gives you a basic understanding of what it takes to talk to journalists about your work and AI in general.

Motivated by the observation that machine learning can be thought of as “a continuation of epistemology by other means” (liberally adapted from Carl von Clausewitz) in this tutorial we will try to critically examine some of the most fundamental (and often tacit) assumptions of machine learning through a philosopher’s lens. In particular, we will focus on key questions pertaining to induction, confirmation, causality, and explanation.

This tutorial will provide the foundational knowledge for understanding the symbiosis of neural networks and differential equations with emphasis on learning differential equations from data and solving differential equations. The tutorial unfolds by introducing the fundamental theory of differential equations, followed by an exploration of the motivation, intricacies, and hands-on applications of two pioneering architectures: Physics-Informed Neural Networks and Neural Ordinary Differential Equations.

Recommender Systems focus users on a small selection of items from a much larger catalog, alleviating information overload and boosting sales for Internet retailers. This tutorial aims to provide a comprehensive and in-depth exploration of the complexities involved in adapting language modeling paradigms to improve the performance and ethical considerations of recommender systems.