Presentations will address whether non-random selection due to conditioning on an exposure level generates selection bias, how leveraging different graphical models can more effectively depict selection mechanisms and their implications for causal identification and estimation, and how selection relates to representativeness in ways that can guide inference and policy translation. The session will also highlight how insights into mechanisms of selection bias connect and can be applied to missing data and effect heterogeneity. By integrating methodological innovation with simulated and applied examples, the symposium will provide a forward-looking view on how selection processes shape validity in modern epidemiology.