Transparency in fairness is an integral, yet so far overlooked, ingredient for facilitating public conversations and productive public debates regarding fair machine learning systems.

In this talk, I will present several of our work in facilitating transparency in fairness.

These include learning to disentangle non-spurious from spurious correlations in the data domain, and providing an individual-level explanation of fair systems via imagined examples.

Supervised by Novi Quadrianto

For more information, please refer to our research work at Predictive Analytics (PAL) Lab of the University of Sussex.