Care-Aware Decision Layer: Embedding Ethics of Care into AI-Supported Clinical Discharge Decisions

Artificial intelligence (AI) systems are increasingly used to support high-stakes clinical decisions, including hospital discharge. Yet, most predictive models rely on statistical risk scores and structured variables while overlooking relational, emotional, and normative dimensions that shape patient readiness and safety. This paper introduces the Care-Aware Decision Layer (CADL), a modular, post hoc care-ethical evaluative layer grounded in the ethics of care . Rather than producing predictions, CADL interprets AI-generated discharge recommendations in light of patient-specific factors such as trust, anxiety, caregiver availability, and preference. We present a weighted ethical reasoning model that operationalizes care ethics through interpretable, rule-based reasoning. CADL’s architecture supports modular integration and ethical auditability, offering justifications for affirming, modifying, or overriding AI recommendations. Through six illustrative clinical scenarios, we demonstrate CADL’s potential to address moral blind spots in traditional AI systems, enhancing deliberation, clinician trust, and patient-centered care. Finally, we propose an evaluation framework for care-aware AI that extends conventional metrics to include ethical defensibility, contextual fit, and human–AI alignment. CADL thus provides a reproducible pathway to improve normative reasoning and ethical judgements by embedding care ethics into clinical AI, bridging predictive accuracy with moral attentiveness in healthcare environments.