Algorithmic auditing through synthetic data

Algorithm auditing refers to the study and evaluation of algorithmic systems to ensure their transparency, fairness, legality and compliance with ethical standards. Our project focuses on the acceptability of practical audits where platforms provide synthetic data about algorithms, instead of the traditional approach with external audits without considering the collaboration of the platform. Technical implications for auditors, users, and privacy need exploration, considering reliability, accuracy, and manipulation concerns. Current audits lack collaboration with platforms, assuming a deceitful intent, contrary to the spirit of IA law. Addressing gaps, such as platform privacy, is crucial for platforms to provide certificates without compromising customer privacy.

Faculty Supervisor:

Sébastien Gambs

Student:

Partner:

INRIA

Discipline:

Computer science

Sector:

Artificial Intelligence

University:

Université du Québec à Montréal

Program: