AXAI-CDSS : An Affective Explainable AI-Driven Clinical Decision Support System for Cannabis Use

Transparency matters for clinical adoption: The paper demonstrates that AI models used in healthcare decisions often lack transparency, which limits doctors and researchers from trusting and using them in real-world medical settings. By making AI decisions more explainable, the system aims to increase adoption of AI-driven clinical tools.

Multiple explanation methods provide better understanding: Rather than relying on a single explanation technique, the system combines several Explainable AI (XAI) methods alongside causal inference to clarify why the AI makes specific predictions about cannabis use, helping clinicians understand the reasoning behind recommendations.

AI should adapt to emotional and personal context: The system uses facial emotion recognition and sentiment analysis to detect a user's emotional state and adjusts its responses accordingly, recognizing that effective clinical decision support requires empathy and personalization, not just technical accuracy.

Plain language explanations bridge the AI gap: By using Large Language Models to translate complex AI outputs into accessible, personalized insights, the system helps non-technical users (clinicians and researchers) interact effectively with AI without requiring deep technical expertise—addressing a key barrier to responsible AI deployment in healthcare.

User-centered design enhances trustworthiness: The paper emphasizes that improving perceived trustworthiness and usability of AI systems requires designing them around actual user needs and emotional responses, not just optimizing model performance, ultimately increasing the real-world impact of clinical decision support tools.