Academic journal article The Yale Law Journal

Bias in, Bias Out

Academic journal article The Yale Law Journal

Bias in, Bias Out

Article excerpt

ARTICLE CONTENTS  INTRODUCTION                                                        2221   I.  THE IMPOSSIBILITY OF RACE NEUTRALITY                          2227       A. The Risk-Assessment-and-Race Debate                        2227       B. The Problem of Equality Trade-offs                         2233       C. Charting Predictive Equality                               2238          1. Disparate Treatment (Input-Equality) Metrics            2240          2. Disparate Impact (Output-Equality) Metrics              2241             a. Statistical Parity                                   2242             b. Predictive Parity                                    2243             c. Equal False-Positive and True-Negative Rates         2243                (Equal Specificity)             d. Equal False-Negative and True-Positive Rates         2244                (Equal Sensitivity)             e. Equal Rate of Correct Classification                 2245             f. Equal Cost Ratios (Ratio of False Positives to       2245                False Negatives)             g. Area-Under-the-Curve (AUC) Parity                    2246       D. Trade-offs, Reprise                                        2248          1. Equality/Accuracy Trade-offs                            2249          2. Equality/Equality Trade-offs                            2249  II.  PREDICTION AS A MIRROR                                        2251       A. The Premise of Prediction                                  2251       B. Racial Disparity in Past-Crime Data                        2251       C. Two Possible Sources of Disparity                          2254          1. Disparate Law Enforcement Practice?                     2255          2. Disparate Rates of Crime Commission?                    2257          3. The Broader Framework: Distortion Versus Disparity      2259             in the Event of Concern III.  NO EASY FIXES                                                 2262       A. Regulating Input Variables                                 2263       B. Equalizing (Some) Outputs                                  2267          1. Equalizing Outputs to Remedy Distortion                 2268          2. Equalizing Outputs in the Case of Differential          2270             Offending Rates             a. Practical Problems                                   2271             b. Conceptual Problems                                  2272       C. Rejecting Algorithmic Methods                              2277  IV.  RETHINKING RISK                                               2281       A. Risk as the Product of Structural Forces                   2282       B. Algorithmic Prediction as Diagnostic                       2284       C. A Supportive Response to Risk                              2286          1. Objections                                              2287          2. Theoretical Framework                                   2288          3. Examples                                                2290       D. The Case for Predictive Honesty                            2294 CONCLUSION                                                          2296 APPENDIX: THE PRACTICAL CASE AGAINST ALGORITHMIC AFFIRMATIVE        2298 ACTION--AN ILLUSTRATION 

INTRODUCTION

"There's software used across the country to predict future criminals. And it's biased against blacks." (1) So proclaimed an expose by the news outlet ProPublica in the summer of 2016. The story focused on a particular algorithmic tool, COMPAS, (2) but its ambition and effect was to stir alarm about the ascendance of algorithmic crime prediction overall.

The ProPublica story, Machine Bias, was emblematic of broader trends. The age of algorithms is upon us. Automated prediction programs now make decisions that affect every aspect of our lives. Soon such programs will drive our cars, but for now they shape advertising, credit lending, hiring, policing--just about any governmental or commercial activity that has some predictive component. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.