About This Report ii

Summary iv

Tables viii

Chapter 1. Introduction 1

Defining Artificial Intelligence (AI) 2

Designing an AI Taxonomy for the Criminal Justice Context 4

Organization of this Report 4

Chapter 2. Taxonomy Overview 6

Taxonomy Design Approach 6

1. Risk Assessment and Prediction 9

2. Surveillance, Monitoring and Identification 16

3. Analysis and Decision Support 24

4. Operations and Case Management 30

5. Training, Treatment, and Services 35

Summary 39

Chapter 3. Observations 41

Complex AI Applications Exist Where Stakes are Highest 41

AI is Cautiously Applied to Contestable, Formalized, and High-Visibility Functions 42

Structural Equity Risks are Higher in Upstream and Midstream Applications 43

Summary 45

Chapter 4. Conclusions and Recommendations 46

Opportunity Areas and Strategic Questions 46

Recommendations 53

Conclusion 54

Appendix A. AI Taxonomy for the Criminal Justice Context 56

Appendix B. Data Types 63

Appendix C. Methodological Approach to Assessing Equity 65

Appendix D. Primary Fairness & Distributional Risk Dimensions of Equity Risk Score 67

Abbreviations 75

References 76