Skip to content

Interpreting Machine Learning: A Gentle Introduction

Welcome to Interpreting Machine Learning, a collaborative book exploring the growing field of eXplainable Artificial Intelligence (XAI).

This book was developed by seminar students of the seminar course: Seminar on XAI: Concepts, Applications and Future Directions), during the Spring of 2026 at the German University in Cairo (GUC).

Instructor: Amr S. Mohamed

Email: amr[dot]saber[at]guc[dot]edu[dot]eg

About the Book

Artificial intelligence (AI) is increasingly shaping decisions. Yet, as AI models grow in complexity, understanding why these models make certain predictions becomes both a technical, regulatory, and ethical necessity. This book introduces readers to the principles, practices, and challenges in XAI.

Table of Contents

  • Chapter 1 – Introduction
    Provides an overview of XAI, including its motivations, definitions, and the importance of interpretability in modern machine learning. It sets the stage by explaining why transparency matters for trust, safety, and adoption of AI systems.

  • Chapter 2 – Regulations
    Examines the regulatory and ethical frameworks guiding the responsible use of AI. Topics include GDPR, EU AI Act, and the role of governance in regulating AI.

  • Chapter 3 – Feature-based Explanations
    Introduces techniques such as PFI, SHAP, LIME, and feature importance metrics. The chapter explains how these methods help identify which input features most influence a model’s predictions.

  • Chapter 4 – Contrastive and Counterfactual Explanations
    Explores explanations that answer “Why not?” and “What if?”. Readers learn how contrastive reasoning highlights alternative outcomes and how counterfactuals provide actionable insights for decision-making. Methods discussed include DiCE and FACE.

  • Chapter 5 – Example and Case-based Explanations
    Discusses exemplar-based reasoning, where models justify predictions by referencing similar past cases. Methods discussed include prototypes and criticisms, ProtoPNet, TCAV, and influence functions.

  • Chapter 6 – Deep Learning Interpretability
    Focuses on methods for interpreting complex neural networks, including saliency maps, layer-wise propagation, activation maximization, attention mechanisms, and visualization techniques.

  • Chapter 7 – Mechanistic Interpretability in Transformers
    Analyzes transformer architectures at the circuit level. The chapter explains how attention heads, layers, and pathways can be dissected to reveal the inner workings of large language models.

  • Chapter 8 – Explainability with Federated Learning
    Investigates the challenges of distributed learning, where data remains decentralized and the challenges that XAI faces in federated learning

  • Chapter 9 – Applications in Healthcare
    🚧 [Still in construction] Discusses the challenges that XAI faces in healthcare.

  • Chapter 10 – Gaps and Future Directions
    🚧 [Still in construction] Identifies open challenges, research gaps, and opportunities for future exploration in XAI.

Citation

Each chapter has its own BibTeX entry. To cite chapters or sections from this book, please use the BibTeX entry at the end of each chapter.

Corrections

For corrections or suggestions to the text, please email the course instructor at the email provided above.