Simply stated, this book bridges the gap between statistics and philosophy. It does this by delineating the conceptual cores of various statistical methodologies (Bayesian/frequentist statistics, model selection, machine learning, causal inference, etc.) and drawing out their philosophical implications. Portraying statistical inference as an epistemic endeavor to justify hypotheses about a probabilistic model of a given empirical problem, the book explains the role of ontological, semantic, and epistemological assumptions that make such inductive inference possible. From this perspective, various statistical methodologies are characterized by their epistemological nature: Bayesian statistics by internalist epistemology, classical statistics by externalist epistemology, model selection by pragmatist epistemology, and deep learning by virtue epistemology.
Another highlight of the book is its analysis of the ontological assumptions that underpin statistical reasoning, such as the uniformity of nature, natural kinds, real patterns, possible worlds, causal structures, etc. Moreover, recent developments in deep learning indicate that machines are carving out their own "ontology" (representations) from data, and better understanding this—a key objective of the book—is crucial for improving these machines’ performance and intelligibility.
Key Features
- Without assuming any prior knowledge of statistics, discusses philosophical aspects of traditional as well as cutting-edge statistical methodologies.
- Draws parallels between various methods of statistics and philosophical epistemology, revealing previously ignored connections between the two disciplines.
- Written for students, researchers, and professionals in a wide range of fields, including philosophy, biology, medicine, statistics and other social sciences, and business.
- Originally published in Japanese with widespread success, has been translated into English by the author.
Introduction
1. The Paradigm of Modern Statistics
2. Bayesian Statistics
3. Classical Statistics
4. Model Selection and Machine Learning
5. Causal Inference
6. The Ontology, Semantics, and Epistemology of Statistics
Biography
Jun Otsuka is Associate Professor of Philosophy at Kyoto University and a visiting researcher at the RIKEN Center for Advanced Intelligence Project in Saitama, Japan. He is the author of The Role of Mathematics in Evolutionary Theory (Cambridge UP, 2019).
"Statistics are being used ever more widely in AI, climate studies, medicine and other areas. Yet they are hard to understand both mathematically and conceptually. Jun Otsuka has the answer to this problem. He has a remarkable ability to explain statistical techniques clearly and accurately with a minimal use of mathematics. At the same time he gives lucid discussions of why they work. He deals not only with the long-standing controversy between Bayesianism and classical statistics, but also with such recent topics as causality and deep learning by computers. His book is the perfect guide to those perplexed by statistics." -- Donald Gillies, University College London
“Otsuka’s excellent book is mostly organized around the idea that different statistical approaches can be illuminated by linking them to different ideas in general epistemology. Otsuka connects Bayesianism to internalism and foundationalism, frequentism to reliabilism, and the Akaike Information Criterion in model selection theory to instrumentalism. This useful mapping doesn’t cover all the interesting ideas he presents. His discussions of causal inference and machine learning are philosophically insightful, as is his idea that statisticians embrace an assumption that is similar to Hume’s Principle of the Uniformity of Nature.” -- Elliott Sober, University of Wisconsin-Madison
