Who Benefits from AI Explanations? Towards Accessible and Interpretable Systems