Engineering AI-Enabled Software with an Eye on Fairness
The fairness of AI-enabled software has emerged as a pressing concern recently. Despite the availability of several metrics and bias mitigation algorithms, the software engineering (SE) process still grapples with effectively identifying, debugging, and designing for fairness, especially regarding protected attributes like race, gender, and age. In this presentation, I'll dive into the SE challenges of ensuring fairness in machine learning models and systems. I'll outline two innovative solution approaches: firstly, the formal verification of neural networks to ensure individual fairness properties, and secondly, the use of causal reasoning within the ML pipeline to address group fairness issues at the component level. These strategies highlight the advantages of a modular design in helping developers pinpoint and rectify unfairness. I'll also share new endeavors and future directions in the rigorous analysis for fair software development.
I'm Sumon Biswas, a Postdoctoral Researcher at Carnegie Mellon University, with a Ph.D. and M.S. in Computer Science from Iowa State University. My research sits at the nexus of Software Engineering (SE) and Artificial Intelligence (AI), specifically focusing on SE practices for AI. I've pioneered methods to integrate algorithmic fairness into the AI-enabled software development pipeline through comprehensive analysis and verification strategies. My early research efforts advocated for fairness as a foundational element in software design, paving the way for subsequent studies. My contributions have been recognized in premier SE forums, including FSE (2020-24) and ICSE (2022-24). I am honored to serve the SE community as a Program Committee member for various editions of ICSE, ASE, FSE, and as a distinguished reviewer for TOSEM. My work supports the NSF TRIPODS Institute's mission on dependable data-driven software and contributes to DARPA projects on verifying legacy software systems. For more about my journey and projects, visit my website at https://sumonbis.github.io.