There exist many different philosophies and methods to discover user requirements for a software system, and most of them have been extensively researched in the field of Requirements Engineering. However, as the world now demands Machine Learning based software, are these methods still usable to discover and understand user requirements? Scholars point toward that ML based software is inherently so different in nature from traditional software systems that new strategies and methods are needed. In this PhD, I propose doing empirical studies to explore emerging methods and philosophies for user requirements discovery amongst practitioners. In this Ph.D., I propose doing empirical studies to explore emerging methods and philosophies for discovering user requirements in the context of ML-based software. The goal is to bridge the gap between theoretical knowledge and real-world practices employed by practitioners in this evolving field.