Aditya Mittal

arXiv | DOI

A. Mittal and N. Matloff (2025). "TowerDebias: A Novel Unfairness Removal Method Based on the Tower Property." arXiv:2411.08297
Status: Under review.

Abstract

Decision-making processes have increasingly come to rely on sophisticated machine learning tools, raising critical concerns about the fairness of their predictions with respect to sensitive groups. The widespread adoption of commercial "black-box" models necessitates careful consideration of their legal and ethical implications for consumers. When users interact with such black-box models, a key challenge arises: how can the influence of sensitive attributes, such as race or gender, be mitigated or removed from its predictions? We propose towerDebias (tDB), a novel post-processing method designed to reduce the influence of sensitive attributes in predictions made by black-box models. Our tDB approach leverages the Tower Property from probability theory to improve prediction fairness without requiring retraining of the original model. This method is highly versatile, as it requires no prior knowledge of the original algorithm's internal structure and is adaptable to a diverse range of applications. We present a formal fairness improvement theorem for tDB and showcase its effectiveness in both regression and classification tasks using multiple real-world datasets.

Code

Github Repository:

arXiv | DOI

A. Mittal, T. Abdullah, A. Ashok, B. Zarate Estrada, S. Martha, B. Ouattara, J. Tran, and N. Matloff (2025) "dsld: A Socially Relevant Tool for Teaching Statistics." arXiv:2411.04228
Status: Under Review.

Abstract

The growing influence of data science in statistics education requires tools that make key concepts accessible through real-world applications. We introduce "Data Science Looks At Discrimination" (dsld), an R package that provides a comprehensive set of analytical and graphical methods for examining issues of discrimination involving attributes such as race, gender, and age. By positioning fairness analysis as a teaching tool, the package enables instructors to demonstrate confounder effects, model bias, and related topics through applied examples. An accompanying 80-page Quarto book guides students and legal professionals in understanding these principles and applying them to real data. We describe the implementation of the package functions and illustrate their use with examples. Python interfaces are also available.

Code

Github Repository:
Quarto Book:

Conference: Directed Reading Program - UC Davis Department of Mathematics
Session: Poster Session
Location: Davis, California, USA
Date: December 2024

Description

Presented mathematical foundations of algorithmic fairness and methods to reduce bias in model predictions across demographic groups.

Materials

Poster:

Conference: Joint Statistical Meetings (JSM) - 2024
Session: Rethinking Statistics and Data Science Education: Incorporating Changing Technology and Encouraging Critical Thinking
Location: Portland, Oregon, USA
Date: August 2024

Description

Presented the dsld R package for discrimination analysis in education, emphasizing how it teaches statistical concepts through real-world fairness and bias examples.

Materials

Slides:

Conference: Undergraduate Research, Scholarship, and Creative Activities (URSCA) Conference
Session: Oral Session
Location: Davis, California, USA
Date: April 26-27, 2024

Description

Presented TowerDebias, a post-processing method that removes sensitive-variable effects from black-box models and improves fairness without retraining.

Materials

Slides:

Conference: UC Davis Scholarship of Teaching and Learning
Session: Poster Session
Location: Davis, California, USA
Date: December 1, 2023

Description

Presented the dsld package for discrimination analysis in education, showing how it supports teaching statistical concepts through real-world fairness examples.

Materials

Poster: