Symbolic AI:
Objectives. The objective of the first part of the course is to introduce the formal approaches to symbolic reasoning allowing the student to combine their algorithms and methods with deep learning techniques to devise a more general approach to artificial intelligence.
[Knowledge and understanding] Understanding the fundamental topics in logic, knowledge representation and symbolic reasoning which are the fundamentals of the symbolic approach to traditional artificial intelligence
[Applying knowledge and understanding] Being capable of combining several methods of symbolic reasoning with different learning paradigms to create models with improved efficiency and interpretability. Being able to use state of the art tools, including SAT and SMT Solvers.
[Making judgments] being capable of applying and combining symbolic and neural components in a critical way, identifying the most effective approaches to solve a given problem. Being able to critically compare different methods to evaluate their effectiveness.
[Communication skills] being able to explain the basic ideas of symbolic methods and communicate the results to a literate public.
[Learning skills] being capable of exploring literature on symbolic computing to find alternative approaches and combine them to solve complex problems.

Reliable AI:
[Knowledge and understanding] Understanding desiderata for sociotechnical systems beyond test set accuracy. Comprehending fairness definitions and the trade-offs between them. Recognizing the role of technical and organizational interventions in mitigating algorithmic bias. Being aware of the recent legislation on anti-discrimination and AI.
[Applying knowledge and understanding] Being able to identify risk factors for algorithmic discrimination in real-world systems. Selecting and applying bias mitigation strategies in practical settings.
[Making judgments] Evaluating and comparing fairness measures and mitigation strategies. Assessing the strengths and limitations of tools such as algorithmic audits and formal verification methods.
[Communication skills] Being able to present findings on AI ethics with clarity, nuance, and rigour. Bridging technical language with societal concerns in an accessible yet precise way.
[Learning skills] Engaging with emerging debates around transparency, accountability, and ethical design in AI systems. Being capable of exploring multidisciplinary topics at the intersection of computer science, law, philosophy, and the social sciences.

Symbolic AI:
Knowledge of the standard Machine Learning Approaches. Knowledge of Python and scientific Python. Basic knowledge of mathematical language and notation.

Reliable AI:
Basic knowledge of machine learning.

Symbolic AI:
The module on symbolic AI offers in its first part the basic principles of Symbolic Artificial Intelligence focusing in particular on the logic approaches for knowledge representation and reasoning. In particular, students will gain foundational knowledge about propositional logic, predicate logic, logic programming and their computational and complexity problems. Environments supporting formal reasoning will be introduced (e.g. SAT solvers, Z3 etc). It is also shown how to use logic to specify problems and properties of systems.

Reliable AI:
This module equips students with the tools to critically design, evaluate, and improve machine learning systems from a fairness and accountability perspective. It emphasizes both technical precision and multidisciplinary insight, fostering a deep understanding of how AI interacts with legal, ethical, and social domains.
Verification and reliability: formal verification of neural networks.
Algorithmic fairness: measures, sources of unfairness, and bias-mitigating interventions.
Fairness in practice: real-world challenges, anti-discrimination law and approaches for algorithmic auditing.
Ethics of AI: selected topics including transparency, agency, and autonomy.
Every topic is presented with real-world examples of tech applications based on machine learning and large language models in important social contexts.

Symbolic AI:
References. - Norvig, P. Russel, and S. Artificial Intelligence. "A modern approach." (Part III) - Mordechai Ben-Ari, Mathematical Logic for Computer Science, Springer. Nikolaj Bjørner, Leonardo de Moura, Lev Nachmanson, and Christoph Wintersteiger, Programming Z3 (Tutorial)

Reliable AI:
Slides and lecture notes provided throughout the course. The Fair ML book serves as a recommended, non-compulsory foundation.
[Fair ML book] Solon Barocas, Moritz Hardt, Arvind Narayanan. Fairness and machine learning: Limitations and Opportunities. MIT Press. 2023

Symbolic AI:
Review of Propositional Logic SAT algorithms and Solvers Review of Predicate Calculus (FOL) SMT Solvers (Z3) Logic Programming (e.g. Prolog and/or Answer Set Programs) Knowledge Graphs: graphical representation of knowledge Modal logic (Temporal logics)

Reliable AI:
Verification & reliability. Robustness and safety of deep learning systems. Formal verification methods for neural networks.
Introduction to algorithmic fairness. Motivation: examples and historical context. Sources of bias and unfairness (status quo, data, problem formulation).
Measuring fairness. Axes of unfairness. Formal non-discrimination criteria. Strengths and tradeoffs of fairness measures.
Datasets & bias. Types of dataset bias and their consequences. Tools for data documentation and analysis. Defining and measuring data diversity. Perspectives from critical data studies.
Fairness interventions. Technical interventions (pre-, in-, and post-processing). Organizational and design strategies. Role of interdisciplinary collaboration.
Real-world challenges to algorithmic fairness. Gaining stakeholder support. Picking appropriate measures and interventions. Fairness strategies of tech companies. Limitations to demographic data and how to handle them.
AI law & governance. Legal definitions of discrimination. Overview of EU Anti-Discrimination Law and AI-specific regulation. Case studies and compliance strategies.
Algorithmic auditing. What audits aim to achieve. Audit methodologies and constraints. Legal and organizational perspectives. Qualitative studies, commentaries, and perspectives.
Ethics of AI. Ethical desiderata for sociotechnical systems. AI risk taxonomies. Selected examples of risks and mitigation strategies.

Symbolic AI:
Frontal lectures and hands-on sessions, both individually and in groups. Hands-on activities typically involve experimenting with SMT solvers and Python libraries, developing or using/testing tools to implement the methodologies seen during lectures.

Reliable AI:
The course combines lectures with interactive, hands-on learning. Technical Labs include Python-based exercises exploring fairness metrics and bias mitigation strategies. Ethics simulations consist of group role-playing activities that place students in real-world scenarios to debate, diagnose, and propose solutions to algorithmic bias.

Symbolic AI:
The examination will consist of an oral interview where the student presents either a relevant paper about symbolic computing not seen in detail during the lectures or a project where the topics seen during the course are applied to a novel problem. The main assessment points are the clarity and precision of the answers, the technical understanding of the methods, and the understanding of their conditions of applicability.

Reliable AI:
The exam will consist of two parts:
1. a group project work, in groups of 2-3 students. Each group works on an original project on algorithmic fairness and presents it with supporting slides. The project topic is proposed by students and validated by the lecturer. The main points of evaluation are the rigor and originality of the performed analyses, the interdisciplinary integration, and the clarity of the presentation.
2. an oral interview to assess the individual's understanding of the course topics. The main points of evaluation are the understanding of the course topics across disciplines, conveyed with nuanced and precise answers.
The two parts can be done in the same session or in separate sessions; the group presentation requires all group members to be present simultaneously. The final mark is obtained as a weighted average of the project (70%) and oral examination (30%).

This course explores topics closely related to one or more goals of the United Nations 2030 Agenda for Sustainable Development (SDGs)