Ai For Social Good: Bridging Innovation And Ethical Ai Decision-Making
Year: 2025
Type: article
Abstract: Artificial Intelligence (AI) is increasingly influencing high-stakes decision-making across various domains, particularly within the criminal justice system, where predictive risk assessment models inform decisions regarding parole, sentencing, and bail. The Correctional Offender Management Profiling for Alternative Sanctions (COMPAS) tool has been widely utilized to assess recidivism risk; however, it has also faced scrutiny due to potential racial biases, especially its disproportionate false positive rates (FPR) for African-American individuals compared to Caucasian and Hispanic counterparts. This study conducts a fairness analysis of COMPAS-based predictions using machine learning models, evaluating the impact of different classifiers on racial groups and examining whether mitigation techniques can improve fairness outcomes. Our analysis employs Logistic Regression, Decision Trees, and Random Forest classifiers to evaluate the fairness of risk assessments. To address these biases, we implement fairness-aware adjustments, which progressively reduce disparities within each classifier. After mitigation, the FPR for African-Americans decreases by 11%, and the Disparate Impact Ratio improves significantly, reducing from 0.11 to 0.08 (Random Forest). These reductions indicate that fairness-aware methods can enhance equitable outcomes while maintaining model performance. We advocate for continued fairness interventions, policy regulations, and interdisciplinary efforts to ensure the responsible deployment of AI in real-world decision-making processes for high-impact applications such as criminal justice, where biased decisions can have severe consequences.
Authors: Nitin Kumar, Vipin Kataria
Language: English
Subfield: Safety Research
Field: Social Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W6893161847
FWCI:
Cites:
Cited by:
Related to: