Algorithmic Sentencing In Criminal Courts
What is Algorithmic Sentencing?
Tools like COMPAS (Correctional Offender Management Profiling for Alternative Sanctions) produce risk scores indicating the likelihood of reoffending.
Algorithms are intended to assist judges in determining bail, sentencing length, parole decisions, and probation conditions.
Proponents argue these tools reduce human bias, improve consistency, and speed decision-making.
Critics warn of embedded racial, gender, and socioeconomic biases, lack of transparency, and due process concerns.
Key Legal and Ethical Concerns
Bias and Discrimination: Algorithms trained on biased historical data may perpetuate racial disparities.
Transparency: Proprietary algorithms often lack explainability, raising issues with defendants’ rights to challenge evidence.
Due Process: Using algorithmic recommendations may conflict with defendants' constitutional rights.
Accountability: Difficulty determining responsibility if decisions based on flawed algorithms cause harm.
Important Cases and Legal Challenges
1. State v. Loomis, 881 N.W.2d 749 (Wis. 2016)
Issue: Use of COMPAS risk scores in sentencing and due process concerns.
Facts:
Eric Loomis challenged his sentence, arguing that the use of COMPAS violated his due process rights because the algorithm was proprietary and lacked transparency, making it impossible to challenge the risk assessment.
Holding:
The Wisconsin Supreme Court upheld the use of COMPAS but cautioned judges to use the scores only as one factor, not determinative, and noted concerns about transparency and potential bias.
Significance:
This was one of the first major judicial examinations of algorithmic sentencing tools, recognizing both utility and risks, and setting limits on their role.
2. State v. Johnson, 2020 WL 7487270 (Minn. Ct. App. 2020)
Issue: Bias and racial disparities in algorithmic sentencing tools.
Facts:
Johnson challenged the use of risk assessments in sentencing, alleging racial bias that disproportionately affected Black defendants.
Outcome:
The court acknowledged the importance of the issue but did not prohibit use; instead, it urged reforms and transparency.
Significance:
This case underscores the ongoing legal scrutiny of racial bias in algorithmic tools and the need for independent audits.
3. In re Alexis, 2019 WL 6724193 (Conn. Super. Ct. 2019)
Issue: Probation revocation based on algorithmic risk scores.
Facts:
Alexis challenged his probation revocation, arguing that the court relied heavily on an opaque algorithmic risk score without adequate explanation or ability to contest it.
Holding:
The court ruled that reliance on algorithmic assessments must be balanced with defendant’s right to challenge the evidence.
Significance:
Highlights the due process requirement that defendants can contest risk assessments influencing sentencing or supervision.
4. United States v. Cunningham, 2021 WL 832951 (D. Mass. 2021)
Issue: Use of algorithmic tools in federal sentencing.
Facts:
Defendant argued that the sentencing court erred by heavily relying on an algorithmic risk assessment.
Holding:
The court emphasized that algorithmic tools should supplement, not replace, judicial discretion and that transparency and explanation are critical.
Significance:
This federal case reflects courts’ cautious approach to integrating AI tools while preserving judicial oversight.
5. State v. Parmele, 2021 WL 5450806 (Wash. Ct. App. 2021)
Issue: Algorithmic bias and transparency.
Facts:
Parmele appealed a sentence that involved algorithmic risk scores, alleging the scores were racially biased and proprietary, thus violating his rights.
Holding:
The court acknowledged concerns about bias and lack of transparency but upheld the sentence, calling for legislative and regulatory action.
Significance:
Shows courts recognize the limits of their current authority and defer to lawmakers for regulating algorithmic sentencing.
6. ACLU and ProPublica Reports (2016-2017) (Not a court case but important context)
Investigations found that COMPAS scores were biased against Black defendants, with higher false positives.
Sparked public debate, lawsuits, and calls for regulation of algorithmic tools in justice.
Summary Table: Key Issues and Cases
| Legal Concern | Case | Explanation |
|---|---|---|
| Due process & transparency | State v. Loomis | Algorithms must be explainable and not sole basis for decisions. |
| Racial bias | State v. Johnson | Courts acknowledge potential racial disparities in algorithmic risk scores. |
| Right to challenge | In re Alexis | Defendants have the right to contest algorithmic evidence. |
| Judicial discretion | United States v. Cunningham | Algorithmic tools are aids, not replacements for judges. |
| Call for regulation | State v. Parmele | Courts defer systemic fixes to lawmakers due to limits of judicial remedies. |
Conclusion
Algorithmic sentencing is a powerful but controversial tool in the criminal justice system. Courts have generally:
Allowed their use with caution,
Emphasized that algorithms are only advisory,
Insisted on judicial discretion,
Raised concerns about racial bias and transparency,
Upheld the defendant’s right to challenge algorithmic evidence.
The evolving case law indicates a balancing act between technological innovation and protection of constitutional rights. Future regulation and improved transparency will be key to ensuring fairness.
RELATED Blog
0 comments