Ideological Overfitting: Understanding Belief Rigidity Through Machine Learning

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.