Why do intelligent and well-informed people examine the same evidence and come to different conclusions? Why do political debates, public health controversies, and media conflicts become more polarized as more data becomes available?

This book argues that ideological rigidity is not merely a matter of bias or bad faith. Rather, it is a predictable outcome of how human cognition processes information. Drawing clear parallels between machine learning and belief formation, the book shows how the same mechanisms that cause AI systems to overfit data also cause human worldviews to become confident, narrow, and resistant to correction.

By advancing from moral diagnosis to computational explanation, the book provides a fresh perspective on polarization and echo chambers. It also explores how techniques used in machine learning can suggest practical strategies for reducing rigidity in institutions and public life. This book addresses students, scholars, researchers, and professionals from various disciplines who are working on topics related to belief, polarization, and AI.

Orestis Delardas is an independent scholar whose work bridges quantitative analysis, computational thinking, and public policy. He holds an MSc in Economics and Policy of Energy and the Environment from University College London and a BEng in Electrical and Electronics Engineering from Swansea University. His research spans energy markets, institutional design, and data-driven decision-making, and has been published across disciplines reflecting the same cross-domain curiosity that animates this book. Professionally, he works in data analysis and corporate sustainability in the private sector. This book grew out of his long-standing interest in why evidence rarely changes minds, and what computational models of learning reveal about that failure. He is based in London.