Ethics in conversation: Building an ethics assurance case for autonomous AI-enabled voice agents in healthcare

Ethics frameworks need practical application: The paper demonstrates that principles-based ethics assurance isn't just theoretical—it can be systematically applied to real AI systems like Dora (an AI telemedicine platform) to identify and address ethical concerns before deployment.

AI benefits and risks are unevenly distributed: The case study revealed that while the AI system provided positive ethical impacts, frontline clinicians faced unexpected risks, particularly regarding their autonomy and decision-making authority—highlighting that different stakeholders experience AI systems very differently.

Clinician autonomy deserves priority attention: The research identified risks to healthcare workers' professional autonomy as a critical area requiring evaluation and safeguards, suggesting that AI deployment in healthcare must actively protect human professionals' ability to make independent clinical judgments.

Comprehensive stakeholder evaluation is essential: The framework's value lies in its ability to structure reasoning about how benefits, risks, and constraints on autonomy are distributed across all affected groups—not just patients or organizations—leading to more inclusive AI assessments.

Ethical assurance requires ongoing adjustment: Rather than a one-time approval process, the framework supports constructing solutions through iterative adjustments to how AI systems operate, suggesting that responsible AI deployment is an evolving process of refinement based on real-world evidence.