Ethical Risks and Governance Directions of Conversational Artificial Intelligence in Nursing Triage and Patient Education: Shared Foundations and Contextual Differences
Yucheng Cao, Yang Tang, Lili Deng, Xusheng Liu, Zhixian Feng, Yu GaoABSTRACT
Aim
To provide an early‐stage integrative synthesis of shared and scenario‐specific ethical risks of Conversational Artificial Intelligence in nursing triage and patient education, and to synthesize governance directions and limitations discussed in the current literature.
Design
A systematic integrative review following the Whittemore–Knafl framework and PRISMA guidelines.
Methods
Eight databases were searched. Two researchers independently conducted screening, data extraction, and thematic coding, followed by inductive synthesis. Quality appraisal used design‐appropriate tools according to article type. Ethical risks were analysed within and across scenarios. Registered in PROSPERO (CRD420251079144).
Results
Nine articles were included (four on nursing triage; five on patient education), comprising two empirical studies, four reviews, two randomized controlled trial protocols, and one debate paper. Both scenarios shared six common ethical challenges: data privacy and security, over‐reliance and deskilling, training‐data bias and stigma reproduction, lack of empathy and emotional interaction capability, algorithmic black box and insufficient interpretability, and blurred accountability and regulatory gaps. Nursing triage presented additional risks including assessment inaccuracy, contextual misunderstanding, lack of personalization, superficially plausible misguidance and insufficient clinician trust. Patient education revealed four distinct issues: misleading information, digital accessibility gaps, cross‐cultural and multilingual adaptation, and fairness and health inequality. Six shared governance directions were synthesized—human oversight and manual review, improvement of legal policies and industry standards, enhanced transparency and interpretability, continuous algorithm optimization and scientific validation, the human–machine balance principle, and capacity building for healthcare professionals. The literature also suggested scenario‐specific reinforcements for triage and education.
Conclusion
Ethical risks of Conversational Artificial Intelligence in nursing show both common and scenario‐dependent patterns. Given the limited and heterogeneous evidence base, the identified governance directions should be viewed as preliminary pathways requiring further validation.
Impact
This review offers evidence‐informed ethical insights and scenario‐based governance references to support safe, equitable and human‐centred application of Conversational Artificial Intelligence in nursing practice.
Patient or Public Contribution
Not applicable.