The Kepler Mission revolutionized exoplanet science by obtaining highly precise photometry of 170,000 stars over 4 years. A critical piece of information needed to accurately exploit Kepler data is the Kepler selection function, since nearly 200,000 targets had to be selected for observation from a sample of over a million stars positioned over the Kepler CCDs with minimal information of their evolutionary state, stellar multiplicity, or proper motions. In this talk I will explain how I use Gaia DR2 to reconstruct the Kepler selection function and explore its biases with respect to evolutionary state, stellar multiplicity, and proper motions. By comparing the stars positioned on the Kepler CCDs that were not chosen for observation to those that were, I will present the first evaluation of the biases of the Kepler target selection function. I find that Kepler was complete for stars brighter than Kp = 14 magnitude, unbiased with respect to proper motions, and that the selection function shows some bias against stellar multiplicity, in the sense that stars without companions were preferentially selected for observation.