Explainability in AI Systems

About This Course

This course addresses the challenges of building ethical and transparent AI-backed systems across various sectors. It introduces explainability techniques in AI and machine learning (explainable AI, XAI), ranging from interpretable methods by design to utilizing large language models to interpret and explain predictions.

Program Outcomes

This course teaches students how to design and evaluate transparent, trustworthy AI systems by applying interpretable machine learning methods and modern explainability techniques. Through hands-on exercises and real-world case studies, they will learn to generate clear, meaningful explanations of model behavior and understand the strengths and limitations of different XAI approaches.

Learning Objectives

• Understand why interpretability matters and how different interpretable models and XAI techniques support responsible AI development
• Practice applying XAI methods using Python-based tools and libraries, implementing interpretable models, generating explanations, and evaluating their quality in real-world scenarios.
• Present clear, well-structured explanations of model behavior and communicate XAI insights effectively.

Course Staff

Prof. Dr. Ekaterina Artemova is a researcher specializing in natural language processing (NLP), with a focus on the post-training and evaluation of large language models (LLMs). She earned her degree from the Higher School of Economics (HSE) and completed a postdoctoral fellowship at LMU Munich. Her research interests include artificial text detection, neural architecture search, and multilingual language understanding benchmarks. Some of her notable contributions involve studies on non-English languages, the topology of attention maps in NLP models, and advancements in benchmarking methodologies.