Leveraging Contextual Counterfactuals Toward Belief Calibration