Science Out of Its Ivory Tower: Improving Accessibility with Reinforcement Learning