Parole Risk Assessment Tools: Scoring, Bias, and Appeals

Parole risk assessment tools are standardized instruments that predict how likely someone is to reoffend after release, using data about criminal history, prison conduct, and personal circumstances to produce a score that parole boards weigh when deciding whether to grant early release and what supervision conditions to impose. Every federal prisoner receives a classification under the First Step Act, and most state corrections systems use at least one validated instrument. The score matters: it shapes whether a person goes home with monthly check-ins or faces GPS monitoring, curfews, and mandatory programming for years.

What Goes Into the Score

These tools pull from two kinds of information, and the distinction between them is the most useful thing to understand about how scoring works.

Static factors are facts about the past that cannot change no matter how much rehabilitation work someone does. Age at first arrest, total prior convictions, offense type, past failures to appear in court, and previous parole violations all fall here. A person first arrested at 16 with five prior felonies looks statistically different from someone with a single conviction at 40, and the tools weigh that difference heavily.

Dynamic factors can shift over time. Corrections professionals often call these “criminogenic needs” because they’re the conditions most closely linked to reoffending: employment, education, substance use history, housing stability, peer associations, and attitudes toward authority. Someone who enters prison unemployed and addicted and leaves with a vocational certificate and two years of sobriety will score differently at reassessment. That potential for change is why dynamic factors matter so much to parole boards. They reveal whether someone has done the work, not just served the time.

The Tools in Common Use

Different jurisdictions use different instruments. Four come up most often.

COMPAS

The Correctional Offender Management Profiling for Alternative Sanctions is one of the most widely used and most criticized tools in American corrections. A proprietary algorithm combines questionnaire answers with data from criminal records to generate separate scores for general and violent recidivism risk. Much of its analysis pulls from existing digital records, including factors like residential stability and whether family members have criminal histories. Because the algorithm’s internal weighting is owned by a private company, defendants and their attorneys typically cannot see exactly how individual factors influenced a particular score.

LSI-R

The Level of Service Inventory-Revised is a 54-item instrument covering ten domains: criminal history, education and employment, finances, family and marital relationships, housing, leisure activities, peer associations, substance abuse, emotional health, and attitudes. Scoring usually involves a structured interview with a trained practitioner who scores responses in real time. The face-to-face component adds a qualitative layer that pure data-crunching misses.

Public Safety Assessment

The PSA, developed by what is now Arnold Ventures, was designed for pretrial decisions rather than parole, but its methodology illustrates how modern tools work. Nine factors drawn entirely from age and criminal history predict three outcomes: failure to appear, new criminal activity, and new violent criminal activity. The PSA deliberately excludes personal interviews, community ties, and marital status, aiming for a tool that scores quickly and consistently across jurisdictions.

PATTERN

The First Step Act of 2018 required the U.S. Attorney General to develop and publicly release a risk and needs assessment system for the federal Bureau of Prisons. PATTERN (Prisoner Assessment Tool Targeting Estimated Risk and Needs), currently on version 1.3, classifies every federal inmate as minimum, low, medium, or high risk. Unlike proprietary tools, the Bureau of Prisons has published PATTERN’s scoring forms, factor lists, and cut-point thresholds, which allows researchers and defense attorneys to examine how scores are generated.

How Evaluators Gather the Information

The data feeding these tools comes from several overlapping sources. The Pre-Sentence Investigation report, prepared by a probation officer at the original conviction, provides the foundation: circumstances of the crime, victim account, and a detailed personal background.

Institutional records add a second layer. Evaluators review disciplinary history during incarceration (fighting, contraband, rule violations) and programming participation (substance abuse treatment, vocational certificates, educational degrees, work assignments). A clean record paired with completed programming sends a different signal than repeated infractions and no programming.

Clinical interviews round out the picture. Evaluators ask targeted questions about attitudes toward the offense, post-release plans, and support network, and they verify self-reported answers against official records and police reports. Someone who claims no substance abuse history when their file shows three drug-related infractions will lose credibility fast. That cross-referencing is what prevents the assessment from becoming a test of how well someone tells a story.

What a Risk Level Actually Controls

After scoring, people are sorted into tiers. State systems commonly use three (low, medium, high); federal PATTERN uses four (minimum, low, medium, high). The classification directly shapes release conditions.

A low-risk designation typically means lighter supervision: monthly check-ins, fewer travel restrictions, minimal mandated programming. Medium-risk individuals usually face required substance abuse counseling, vocational training, or cognitive-behavioral programs. High-risk classifications bring the heaviest restrictions: GPS ankle monitoring, frequent unannounced home visits, mandatory curfews, and strict limits on associations. Daily costs for electronic monitoring vary by jurisdiction, and in many states the parolee pays part or all of them.

In the federal system the score also controls release timing directly. Under 18 U.S.C. ยง 3632, any federal prisoner who successfully participates in approved programming earns 10 days of time credits for every 30 days of participation. Prisoners classified as minimum or low risk who maintain that classification over two consecutive assessments earn an additional 5 days, for a total of 15 days per 30-day period. Those credits can go toward transfer to a halfway house, home confinement, or early supervised release.

Scores influence decisions; they don’t make them. Parole boards retain discretion to weigh the number alongside the nature of the crime, victim input, release plans, and their own judgment. Predictive accuracy is better than unaided human judgment in the aggregate but remains imprecise enough that scores should not substitute for the full deliberation.

Reassessment and Changing the Score

A risk score is not a permanent label. The federal system requires reassessment at regular intervals, and most state systems follow suit. Federal courts have found that reassessment every 6 to 12 months works well for tracking changes in stable criminogenic needs like employment and substance abuse. For acute factors that shift rapidly, some research suggests evaluations every one to two weeks improve the ability to predict imminent reoffending.

Reassessment creates a path forward. Someone classified as high risk at intake who completes programming, maintains clean conduct, and develops a solid release plan should see the score drop. In the federal system, moving from medium to low or minimum unlocks the higher earned-time-credit rate, which can meaningfully shorten time in custody. The reverse is also true: new infractions or dropping out of programming can push a score up and delay release eligibility.

Bias and Accuracy Concerns

Independent analyses have found that some widely used tools produce higher false-positive rates for Black defendants than for white defendants, meaning they incorrectly classify Black individuals as high risk at elevated rates. One prominent analysis found Black defendants were nearly twice as likely as white defendants to be incorrectly flagged as high risk despite not going on to reoffend.

The problem is not limited to one tool. A 2023 federal review of PATTERN found that while the instrument performed accurately overall, it continued to overpredict risk for Black, Hispanic, and Asian men and women relative to white individuals. Overprediction means people in those groups may face harsher supervision, longer incarceration, and fewer earned time credits than their actual risk warrants.

Defenders point out that actuarial tools still outperform unstructured human judgment, which carries its own biases. That is true, but “better than gut feelings” is a low bar when the stakes include years of someone’s freedom. Agencies using these tools are expected to evaluate performance across racial and ethnic subgroups and recalibrate when disparities appear. Whether that happens with enough rigor is a separate question.

There is also a legal dimension to the opacity of proprietary tools. In State v. Loomis (2016), the Wisconsin Supreme Court held that using a COMPAS score at sentencing does not violate due process, but only with safeguards: the sentencing court must be warned that the algorithm is proprietary and cannot be fully disclosed, that scores reflect group-level data, that studies have raised questions about disproportionate classification of minority defendants, and that risk tools require ongoing monitoring. The court also ruled that a risk score may never be the sole factor determining incarceration or sentence severity. The decision addresses sentencing rather than parole directly, but the warnings travel with the score.

Challenging a Classification

People facing an unfavorable risk score are not without options, though the process varies by jurisdiction. The most straightforward challenges involve factual errors: a prior conviction that actually belongs to a relative with the same name, an incorrect birthdate, or a miscalculated score on a hand-scored instrument. These mistakes happen more often than you might expect, and correcting them can change a classification.

Beyond clerical errors, defense counsel can present protective and promotive factors specific to the individual that the standardized tool may not capture: a strong employment offer, family support, programming completion beyond what was required, or decades of clean conduct since a distant offense. The work resembles sentencing mitigation. You’re putting the full picture of a person in front of the decisionmaker rather than letting a number stand alone.

In the federal system, inmates receive their PATTERN risk level during intake and review an individualized needs plan. Those who believe the classification is wrong can use existing administrative remedy procedures to challenge it. Whether those grievance processes are truly accessible for people without legal representation is a legitimate concern, but the formal mechanism exists.

The Financial Side

Risk classification affects more than supervision intensity. Roughly 33 states charge a monthly supervision fee to people on parole, and total costs can range from hundreds to thousands of dollars depending on jurisdiction and supervision level. Higher-risk classifications typically carry more expensive conditions: electronic monitoring fees, program co-pays, and drug testing all add up.

The burden falls hardest on people who are already economically disadvantaged, which creates a troubling feedback loop. Many of the dynamic factors that drive higher scores, like unemployment and unstable housing, are closely tied to poverty. Imposing significant financial obligations on people already struggling in those areas can make successful reintegration harder, not easier. Some jurisdictions offer fee waivers or sliding-scale payments, but the practice is far from universal.