Can You Fake It Until You Make It? - Impacts of Differentially Private Synthetic Data on Downstream Classification Fairness

Explore the impacts of differentially private synthetic data on downstream classification fairness in this 18-minute conference talk from FAccT 2021. Delve into the research presented by V. Cheng, V. Suriyakumar, N. Dullerud, S. Joshi, and M. Ghassemi as they investigate whether synthetic data can effectively replace real data in machine learning applications while maintaining fairness. Learn about their experimental pipeline, which includes data generation, model training, and fairness evaluation. Examine the results of their study and understand the implications for using synthetic data in various classification tasks. Gain insights into the potential benefits and limitations of "faking it until you make it" with synthetic data in the context of machine learning fairness.