• Can we develop a trustworthy AI system that reliably screens for substance use risk and delivers appropriate, safety-guarded interventions without direct physician oversight?
• Does the use of an AI interface (chatbot) increase the willingness of adolescents to disclose sensitive information compared to human providers due to a lack of perceived judgment?
• Can an AI-driven implementation of SBIRT reduce provider burden while improving systematic follow-up and patient outcomes compared to the current standard of care

The research proposes developing an AI-assisted tool to automate the SBIRT process for adolescents. This involves a tiered technical approach: utilizing deterministic, rule-based algorithms for the initial screening and risk stratification (low, moderate, high), and exploring Large Language Models (LLMs) to conduct the conversational "brief intervention" for higher-risk patients6666. The methodology includes fine-tuning existing models using transcripts or "GAN-like" adversarial training (AI patient vs. AI therapist) to establish safety guardrails.