Deuterium has a special place in cosmology, nuclear astrophysics, and galactic chemical evolution, because of its unique property that it is only created in the big bang nucleosynthesis while all other processes result in its net destruction. For this reason, among other things, deuterium abundance measurements in the interstellar medium (ISM) allow us to determine the fraction of interstellar gas that has been cycled through stars, and set constraints and learn about different Galactic chemical evolution (GCE) models. However, recent indications that deuterium might be preferentially depleted onto dust grains complicate our understanding about the meaning of measured ISM deuterium abundances. For this reason, recent estimates by Linsky et al. (2006) have yielded a lower bound to the “true”, undepleted, ISM deuterium abundance that is very close to the primordial abundance, indicating a small deuterium astration factor contrary to the demands of many GCE models. To avoid any prejudice about deuterium dust depletion along different lines of sight that are used to determine the “true” D abundance, we propose a model-independent, statistical Bayesian method to address this issue and determine in a model-independent manner the undepleted ISM D abundance. We find the best estimate for the gas-phase ISM deuterium abundance to be (D/H)ISM ≥ (2.0 ± 0.1) × 10-5. Presented are the results of Prodanović et al. (2009).