On Credit: Scoring Foundations in the Age of AI

This book traces the mathematical lineage connecting Alan Turing's wartime cryptanalysis to modern credit scoring systems. It reveals how weight of evidence, developed to break enemy codes at Bletchley Park near London in the early 1940s, connects through maximum likelihood and logistic regression to neural networks, ensemble methods, and the explainability frameworks used in today's financial institutions.Written for data scientists, risk professionals, and students in financial services, the book follows a single thread: the credit assignment problem. Whether allocating credit to loan applicants, attributing predictions to features, or back-propagating errors through neural networks, the fundamental challenge remains the same: fairly assigning credit across contributing factors.Each chapter includes historical context, mathematical intuition, and practical applications, making complex concepts accessible to both technical practitioners and senior risk leaders. The book provides an essential toolkit for practitioners building credit models and necessary context for leaders evaluating AI systems in regulated financial services. By revealing the deep connections between classical statistics and modern machine learning, it equips readers to make informed modeling decisions. Read more