Opening the black box: The effect of SHAP and counterfactuals on AI novices' trust in and understanding of black box models