CJ lecture one
Shift from Subjective Prosecution to Data-Driven Oversight
Prosecutorial Background:
Professional career began as a criminal prosecutor in the Manhattan District Attorney's Office.
Served subsequently at the United States Department of Justice.
Appointed as Attorney General of the state of New Jersey in .
Initial Systemic Inquiries:
Initial inquiries upon taking office focused on identifying core metrics: who was being arrested, who was being charged, who was being put into jails and prisons, and whether decisions were improving public safety.
Systemic tracking failure: Major criminal justice agencies, including the New Jersey Attorney General's office, lacked mechanisms and infrastructure to track critical operational metrics and outcomes.
Manual Audit Procedure: Detectives conducted a case-by-case manual review of five years of case files using yellow legal pads in a conference room to extract basic data.
Findings from Audit: Prosecutorial resources were heavily concentrated on prosecuting low-level street drug cases occurring immediately adjacent to the state office building in Trenton.
Realities of Police Operations and the Moneyball Paradigm
Operational Conditions in Camden, New Jersey:
Camden, New Jersey was designated as the most dangerous city in America, leading the New Jersey Attorney General to directly run and oversee the Camden Police Department.
Senior police officials attempted to reduce crime using physical yellow sticky notes placed on wall boards (e.g., noting a robbery occurring weeks prior with zero suspects, or a shooting in a neighborhood week prior with zero suspects).
Modern data-driven policing, analytics tools, and cross-agency data sharing were entirely absent across police departments, prosecutors' offices, courts, and correctional facilities.
The Moneyball Framework Applied to Criminal Justice:
Subjective decision-making across the criminal justice system closely mirrored traditional baseball scouts who relied exclusively on personal instinct and experience.
Adapted the analytical model used by the Oakland A's, leveraging statistical data to replace subjective evaluations with objective analytics to maximize performance outcomes.
Empirical outcomes of data-driven reform in Camden, New Jersey:
Removed Camden from the top position on the list of the most dangerous cities in America.
Reduced the murder rate in Camden by , representing lives saved.
Reduced total overall crime in Camden by .
Realigned statewide prosecution priorities away from local street-level drug offenses toward high-impact matters: violent crime reduction targeting the most violent offenders, street gang prosecutions, gun and drug trafficking, and political corruption.
Macro-Level Statistics and Systemic Pretrial Inefficiencies
Fundamental Role of Public Safety:
Public safety is the most essential function of government, serving as an absolute requirement before citizens can be educated, maintain health, or fulfill daily activities.
National Criminal Justice Statistics:
Annual volume of arrests in the United States: total arrests.
Proportion of misdemeanor and low-level offenses: to of all annual arrests.
Proportion of violent crimes: Less than () of total annual arrests.
Annual state and local corrections expenditures: ( billion).
Total population currently in United States jails and prisons: individuals.
Pretrial detainee proportion: Two-thirds () of all individuals in local jails are unconvicted detainees awaiting trial.
National recidivism rate: of individuals released from prison are rearrested, establishing a cycle where nearly out of released individuals return to custody.
The Pretrial Risk Determination Problem
Central Importance of the Pretrial Decision:
Evaluating whether a newly arrested individual poses a public safety risk requiring pretrial detention or can be safely released is the single most critical decision in public safety.
This core decision dictates downstream outcomes across the entire justice system, including sentence lengths, access to drug treatment programs, community crime rates, and overall violence levels.
Disconnect Between Judicial Intent and Empirical Reality:
Judges consistently express the belief that they detain dangerous people and release nonviolent individuals.
Empirical data reveals widespread system failures caused by subjective decision-making:
Low-risk offenders comprise of the total criminal justice population, yet significant numbers remain incarcerated pretrial due to an inability to pay bail.
Case Example (Leslie Chu): A Texas resident arrested for stealing blankets on a cold winter night. Kept in jail on a bail bond he could not afford, remaining incarcerated for months until trial at a total taxpayer cost exceeding .
High-risk offenders: Nationally, of the highest-risk individuals evaluated as having the highest likelihood of committing new crimes are released pretrial.
Architecture of the Universal Risk Assessment Tool
Deficiencies of Pre-Existing Risk Assessment Tools:
Only to of United States jurisdictions historically utilized any formal risk assessment tool.
Existing tools were exceptionally expensive to administer, time-consuming, restricted to local jurisdictions, and non-scalable.
Development by the Arnold Foundation:
Assembled a team of data scientists, statisticians, and researchers to construct a universal, objective risk assessment tool.
Analyzed a dataset of cases collected across municipal, county, state, and federal jurisdictions nationwide (the largest pretrial dataset in the United States).
Evaluated over potential risk factors to identify specific factors that were universally predictive of risk across the entire country.
Key Inputs and Mechanics of the Universal Tool:
Simple input factors focused on objective historical metrics:
Defendant's prior criminal convictions.
Prior sentences resulting in incarceration.
Past engagement in violent behavior.
History of failure to appear in court.
Objective predictions generated by the tool:
New Criminal Activity Score: Predicts the likelihood of committing a new crime if released.
Elevated Risk of Violence: Predicts whether an individual will commit an act of violence upon release.
Failure to Appear Score: Predicts the likelihood of returning to court.
User Interface and Strategic Integration:
The tool provides judges with a clear dashboard displaying the New Criminal Activity Score (maximum score of ), an Elevated Risk of Violence flag, and a Failure to Appear Score.
Integrates objective data science with judicial instinct and experience rather than replacing judicial discretion entirely.
Implementation rollout:
Launched statewide in Kentucky on July 1.
Target goal: Standardize data-driven risk tools for every judge in the United States within years.
Active development underway to create specialized risk assessment tools for prosecutors and law enforcement officers.