Constitutional AI and Behavior Change: Ethical Frameworks for Trust and Adoption in Clinical LLM Deployment

The use of Large Language Models (LLMs) in healthcare settings presents a fundamental question: how do we ensure that AI systems promote positive behavior change and maintain ethical boundaries? This position paper argues that Constitutional AI (ConAI) [3] is a promising design approach where ethical constraints are encoded into AI behavior and become facilitators of behavior change [9]. Reflecting on early implementation experiences with the TREAT (Transforming Healthcare Through Semantic Interoperability and Self-Efficacy) remote patient monitoring (RPM) project [7], we view ConAI as facilitating four aspects of behavior change: clinical trust, patient engagement, ethical accountability, and behavior change opportunity. Early findings suggest that AI effectiveness is improved by ethical alignment. The workshop community is invited to explore how ethical AI design can help address complex challenges in healthcare technology adoption.