On The Path To AI: Law’s Prophecies And The Conceptual Foundations Of The Machine Learning Age

Description

This open access book explores machine learning and its impact on how we make sense of the world. It does so by bringing together two ‘revolutions’ in a surprising analogy: the revolution of machine learning, which has placed computing on the path to artificial intelligence, and the revolution in thinking about the law that was spurred by Oliver Wendell Holmes Jr in the last two decades of the 19th century. Holmes reconceived law as prophecy based on experience, prefiguring the buzzwords of the machine learning age—prediction based on datasets. On the path to AI introduces readers to the key concepts of machine learning, discusses the potential applications and limitations of predictions generated by machines using data, and informs current debates amongst scholars, lawyers and policy makers on how it should be used and regulated wisely. Technologists will also find useful lessons learned from the last 120 years of legal grappling with accountability, explainability, and biased data. Prologue—Starting with Logic......Page 5 Holmes and His Legacy......Page 7 A Note on Terminology: Machine Learning, Artificial Intelligence, and Neural Networks......Page 9 Notes......Page 11 Contents......Page 15 About the Authors......Page 18 Abbreviations......Page 19 1 Two Revolutions......Page 21 1.1 An Analogy and Why We’re Making It......Page 23 1.2 What the Analogy Between a Nineteenth Century Jurist and Machine Learning Can Tell Us......Page 24 1.3 Applications of Machine Learning in Law—And Everywhere Else......Page 27 1.4 Two Revolutions with a Common Ancestor......Page 29 2 Getting Past Logic......Page 38 2.1 Formalism in Law and Algorithms in Computing......Page 39 2.2 Getting Past Algorithms......Page 41 2.3 The Persistence of Algorithmic Logic......Page 43 3 Experience and Data as Input......Page 51 3.1 Experience Is Input for Law......Page 52 3.2 Data Is Input for Machine Learning......Page 53 3.3 The Breadth of Experience and the Limits of Data......Page 56 4 Finding Patterns as the Path from Input to Output......Page 59 4.1 Pattern Finding in Law......Page 60 4.2 So Many Problems Can Be Solved by Pure Curve Fitting......Page 62 4.3 Noisy Data, Contested Patterns......Page 64 5 Output as Prophecy......Page 67 5.1 Prophecies Are What Law Is......Page 68 5.2 Prediction Is What Machine Learning Output Is......Page 72 5.3 Limits of the Analogy......Page 75 5.4 Probabilistic Reasoning and Prediction......Page 77 6 Explanations of Machine Learning......Page 85 6.1 Holmes’s “Inarticulate Major Premise”......Page 86 6.2 Machine Learning’s Inarticulate Major Premise......Page 88 6.3 The Two Cultures: Scientific Explanation Versus Machine Learning Prediction......Page 89 6.4 Why We Still Want Explanations......Page 93 7.1 Problems with Juries, Problems with Machines......Page 99 7.2 What to Do About the Predictors?......Page 102 8 Poisonous Datasets, Poisonous Trees......Page 107 8.1 The Problem of Bad Evidence......Page 108 8.2 Data Pruning......Page 110 8.3 Inferential Restraint......Page 111 8.4 Executional Restraint......Page 112 8.5 Poisonous Pasts and Future Growth......Page 113 9 From Holmes to AlphaGo......Page 120 9.1 Accumulating Experience......Page 121 9.2 Legal Explanations, Decisions, and Predictions......Page 124 9.3 Gödel, Turing, and Holmes......Page 126 9.4 What Machine Learning Can Learn from Holmes and Turing......Page 127 10 Conclusion......Page 130 10.1 Holmes as Futurist......Page 131 10.2 Where Did Holmes Think Law Was Going, and Might Computer Science Follow?......Page 136 10.3 Lessons for Lawyers and Other Laypeople......Page 138 A Data Scientist’s View......Page 146 A Lawyer’s View......Page 147 Selected Bibliography......Page 150 Index......Page 159