STOR390 HOMEWORK 5 Solution

Abstract: This homework is meant to give you practice in creating and defending a position with both statistical and philosophical evidence. We have now extensively talked about the COMPAS data set, the flaws in applying it but also its potential upside if its shortcomings can be overlooked. We have also spent time in class verbally assessing positions both for an against applying this data set in real life. In no more than two pages (knit to a pdf to ensure page count) take the persona of a statistical consultant advising a judge as to whether they should include the results of the COMPAS algorithm in their decision making process for granting parole. First clearly articulate your position (whether the algorithm should be used or not) and then defend said position using both statistical and philosophical evidence. Your paper will be graded both on the merits of its persuasive appeal but also the applicability of the

statistical and philosophical evidence cited.
Before implementing COMPAS or any criminal justice algorithm, several prerequisites must be met. Firstly, the accuracy of FPR and TNR between different groups, such as race or gender, or proxy variables that combine several immutable characteristics, must be considered. Using a veil of ignorance, one would oppose an algorithm like COMPAS, which is more likely to predict a false positive and less likely to predict a true negative for recidivism amongminorityindividuals, especiallyifplaced intoadisadvantaged group in society. Ideally, an algorithm would perform similarly across these groups in terms of FPR and TNR. However, an algorithm that performs equally poorly across groups is not a solution. Flipping a coin, which would be correct only 50% of the time, is only slightly less accurate than the COMPAS algorithm but without bias. The reliance on an algorithm that cannot decisively outperform a simplistic method such as coin flipping, which boasts a 50% accuracy rate, poses a profound ethical dilemma. The use of such an algorithm, under the guise of efficiency or objectivity, could inadvertently institutionalize bias within judicial processes, perpetuating cycles of inequality. Further, it could mask the subjective nature of decision-making under a veneer of technological impartiality. In light of these considerations, it is crucial that any algorithm deployed in such high-stakes contexts undergoes rigorous, ongoing evaluation not only of its predictive accuracy but also of its fairness and impact on all segments of society. Furthermore, the discussion extends beyond mere statistical validation to encompass the broader implications of algorithmic governance in justice. The integration of such technologies into the judicial system demands transparency, accountability, and a commitment to rectifying identified biases. Stakeholders, particularly those from marginalized communities, must have a voice in how these tools are developed, deployed, and audited. The ultimate goal should not be to replace human judgment with algorithmic determinations but to enhance the former with the latter’s insights, ensuring that technology serves as a tool for justice enhancement rather than an instrument of inequity. In envisioning a future where algorithms like COMPAS are part of the criminal justice landscape, it is essential to establish robust frameworks that include independent oversight, public reporting of algorithmic performance, and mechanisms for addressing disparities when they are identified. By adopting a holistic approach that considers the statistical, ethical, and societal dimensions of algorithmic implementation, the justice system can leverage technology to improve decisionmaking processes while steadfastly upholding the principles of fairness and equity.
Overall, while COMPAS and similar algorithms used to predict recidivism have room for improvement, they can provide valuable insights into risk factors associated with recidivism when used judiciously. This should be complemented by a robust framework that includes qualitative assessments, personal circumstances, and rehabilitative efforts. The goal is to employ a holistic decision-making process that respects the complexity of human behavior while striving for a just and equitable penal system that does not dehumanize the individual under accusation. After all, we are a country founded on the principle of innocence until proven guilty. The development and advancement of technology should enhance, not undermine, this principle for everyone, regardless of their skin color or gender.

References in the Discussion