[DS Interface] Large Pre-trained Language Models Contain Human-like Biases ofWhat is Right and Wrong