Gravar-mail: Explainable Agents for Less Bias in Human-Agent Decision Making