This position paper proposes using Natural Language Processing (NLP) for causal analysis and perception mining to infer mental health from social media. It argues for more explainable AI models in mental health analysis, focusing on two dimensions: 1) Causal analysis to identify cause-and-effect relationships in user-generated text; and 2) Perception mining to understand psychological perspectives influencing online users' intentions. The paper explores critical areas within NLP, particularly discourse analysis, to advance the development of conversational agents for mental health inference from social media.
