ANR JCJC AI-PULSE

Welcome to the AI-PULSE Project Website, an ANR JCJC project led by Héber H. Arcolezi at ÉTS Montréal, within the Trustworthy Information Systems Lab (TISL).

AI-PULSE investigates how Differential Privacy (DP) and fairness can be jointly integrated into Machine Learning (ML) systems, with the goal of developing efficient, scalable, and responsible AI methods.

The project started on March 1, 2025.

Motivation

AI-PULSE addresses the growing need for responsible AI in settings where the benefits of data-driven systems must be balanced against concerns related to privacy and fairness. Privacy-preserving techniques such as Differential Privacy can limit information leakage, while fairness-aware learning aims to reduce disparities across individuals and groups. However, these objectives can interact in complex ways. AI-PULSE studies these interactions and develops methods that jointly account for privacy, fairness, and predictive utility.

Goal

AI-PULSE aims to develop methodological and theoretical foundations for Machine Learning systems that jointly account for Differential Privacy, fairness, and utility. The project investigates when privacy and fairness objectives reinforce or conflict with one another, and how these trade-offs can be characterized and mitigated. Through new algorithms, theoretical analyses, empirical evaluations, and practical tools, AI-PULSE seeks to contribute methods for responsible privacy-preserving Machine Learning.

Principal Activities

The AI-PULSE project is organized around three main tasks:

  • Task 1. Develop and evaluate methodologies that combine Differential Privacy (DP) and fairness mechanisms in Machine Learning. This includes pre-processing, in-processing, and post-processing approaches, as well as extensions to decentralized settings such as Federated Learning.
  • Task 2. Develop theoretical foundations for understanding and balancing privacy, fairness, and utility under Differential Privacy and Local Differential Privacy (LDP). This includes studying the impact of privacy mechanisms on fairness and designing privacy-preserving mechanisms that account for fairness constraints.
  • Task 3. Translate the project’s theoretical and methodological advances into practical tools, including an open-source toolkit integrating privacy-preserving and fairness-aware Machine Learning methods.

ANR JCJC AI-PULSE project is funded by ANR, the French National Research Agency.