Leveraging Multi-Task Learning to Transfer Demographic Fairness Across Tasks with Limited Demographic Information

Multi-task learning can be used to transfer demographic fairness from a task with available demographic information to a related task without such information, enabling fairer models for the target task.