Creating Unbiased Machine Learning Models by Design

Unintended bias against protected groups has become a key obstacle to the widespread adoption of machine learning methods. This work presents a modeling procedure that carefully builds models around protected class information in order to make sure that the final machine learning model is independent of protected class status, even in a nonlinear sense. This procedure works for any machine learning method. The procedure was tested on subprime credit card data combined with demographic data by zip code from the US Census. The census data serves as an imperfect proxy for borrower demographics but serves to illustrate the procedure.

Destroying Disability: Expanding Application of the Genocide Convention

Summary Disability is not a protected class under the Genocide Convention, even though disabled people across the world frequently face egregious human rights violations. Many of those practices should be considered genocide because they meet the criteria listed in the definition. In order to amount to genocide, an action must be committed with the intent to destroy a group, in whole or in part, by killing, causing serious harm, inflicting conditions of life calculated to bring about destruction of the group, prevent births, or forcibly transfer children out of the group. Disabled people have been subjected to all these actions. By refusing to grant this group status as a protected class, the international community has allowed acts of genocide to continue into the twenty first century. To prevent future genocides against this group, and advance disability rights on a global scale, disabled people need the protections provided in the Genocide Convention.

Assessing Algorithmic Fairness with Unobserved Protected Class Using Data Combination

The increasing impact of algorithmic decisions on people’s lives compels us to scrutinize their fairness and, in particular, the disparate impacts that ostensibly color-blind algorithms can have on different groups. Examples include credit decisioning, hiring, advertising, criminal justice, personalized medicine, and targeted policy making, where in some cases legislative or regulatory frameworks for fairness exist and define specific protected classes. In this paper we study a fundamental challenge to assessing disparate impacts in practice: protected class membership is often not observed in the data. This is particularly a problem in lending and healthcare. We consider the use of an auxiliary data set, such as the U.S. census, to construct models that predict the protected class from proxy variables, such as surname and geolocation. We show that even with such data, a variety of common disparity measures are generally unidentifiable, providing a new perspective on the documented biases of popular proxy-based methods. We provide exact characterizations of the tightest possible set of all possible true disparities that are consistent with the data (and possibly additional assumptions). We further provide optimization-based algorithms for computing and visualizing these sets and statistical tools to assess sampling uncertainty. Together, these enable reliable and robust assessments of disparities—an important tool when disparity assessment can have far-reaching policy implications. We demonstrate this in two case studies with real data: mortgage lending and personalized medicine dosing. This paper was accepted by Hamid Nazerzadeh, Guest Editor for the Special Issue on Data-Driven Prescriptive Analytics.

Is Affirmative Action Still Necessary? Analysis of Minority-Working Adults’ Perceptions

Over the past three decades, the Unites States has struggle valiantly to overcome that disgusting legacy as it moves toward to eliminate race, and gender inequality, and the uprooting of prejudice and discrimination. Out of this struggle, came the birth of affirmative action. It has left politicians, social scientists, and economists debating its merits and possible alternatives. From the Supreme Court to the dinner table, the potential effects of this policy on our legal, political and social system have been argued. This study analyzes the perceptions protected class employees in terms of the affirmative action in employment. Utilizing a sample of 151 protected class working adults, data analysis provided mixed support to the stated hypotheses which suggested that affirmative action had eliminated most discriminatory practices in corporate America. The results of this study answer the question of this study asserting the necessity of the affirmative action.

Conclusion: Diversity Intelligence as a Core of Diversity Training and Leadership Development

Problem: Leadership development is a core part of training, education, and career management strategies in organizations. Yet, leaders are not translating what they learn about protected class employees during leadership development initiatives back to the workplace. Solution: Diversity intelligence should be added to organizational diversity and leadership development training and education initiatives. With DQ as a core of the training and education initiatives, leaders may acquire the needed ability to translate what they learn to actual practice. A conceptual model for DQ as a core of leadership development and typology of leaders with low and high DQ are provided. They will be able to better lead their protected class followers because they will know who they are and how to enhance their performance. Stakeholders: Workplace leaders, diversity trainers, educators, and career management professionals are provided ideas for enhancing their diversity improvement efforts. Implications for Human Resource Development professionals and researchers are also offered.