Explainable AI with Python

When a machine-learning model produces an answer, understanding how it arrived there can matter as much as the answer itself. Explainable AI with Python, Second Edition, explores the methods and ideas that help make complex AI systems more interpretable—and the questions practitioners should ask when an explanation is offered.

Antonio Di Cecco and Leonida Gianfagna connect foundational concepts with practical approaches to explainable AI (XAI). This updated edition extends the discussion to large language models, generative and multimodal systems, and the interpretability of additive models.

From black-box models to meaningful explanations

The book introduces the landscape of XAI and considers why transparency matters in machine learning. It examines both intrinsic approaches, where interpretability is part of a model’s design, and model-agnostic methods, which can be applied without relying on a model’s internal workings. The distinction gives readers a useful way to think about which kinds of explanation fit different situations.

Methods across machine learning

Chapters address deep-learning explanations, additive models, and adversarial machine learning. The treatment also considers how explanations can differ between training and test data, a practical concern when assessing how reliably an interpretation reflects a model’s behavior.

XAI meets language models and multimodal AI

Expanded coverage explores explainability for language models and large language models, along with generative and multimodal systems that work across forms of data. The book also considers the relationship between XAI, privacy, and the possible development of artificial general intelligence—questions that broaden the discussion beyond explaining a single prediction.

For readers working with AI

With theoretical discussion alongside Python-focused practical material, this edition is suited to machine-learning practitioners, students, and readers seeking a structured introduction to the technical and broader implications of explainable AI. Its central concern is not simply how to produce an explanation, but how to make explanations useful in the pursuit of transparent and responsible systems.

Explore a field where understanding a model is part of building and evaluating it. 💡