Empirically Verifying Hypotheses Using Reinforcement Learning