Manjish Pal

Research Statement

My research lies at the intersection of algorithmic fairness, machine learning, and graph data analysis. I focus on developing principled methods to ensure fairness in machine learning models, particularly in structured and unstructured data settings. My work addresses biases in graph-based learning tasks, including fair link prediction, fair subgraph representation learning, and fairness in submodular maximization. I investigate how algorithmic decisions impact marginalized groups, with a special emphasis on intersectional fairness, ensuring equitable treatment across multiple demographic attributes.

A core aspect of my research involves optimizing fairness constraints in learning algorithms while maintaining model performance and interpretability. I employ techniques from convex and combinatorial optimization, multilinear relaxation, and deep learning to design fairness-aware algorithms. 

In addition, I explore the calibration of graph neural networks, ensuring that predictive confidence levels are well-calibrated and unbiased. This is crucial in high-stakes applications where trustworthiness and reliability are paramount. Through interdisciplinary collaborations, I aim to bridge theoretical advancements with practical implementations, contributing to the broader goal of responsible AI. My research has applications in various domains, including social network analysis, recommendation systems, and healthcare informatics, where fairness and accountability are critical.