Culturally Sensitive Artificial Intelligence in Nursing: A Transparency Imperative for Equitable and Safe Care

Abstract 

Artificial intelligence (AI) is rapidly being integrated into clinical decision support systems across European healthcare settings, yet most existing systems fail to accommodate the cultural diversity of contemporary patient populations. This perspective article argues that cultural sensitivity and transparency are not optional enhancements to AI in nursing — they are structural prerequisites for safe and equitable care. Drawing on Leininger’s Culture Care Theory, the EU AI Act (Regulation 2024/1689), and recent empirical evidence, this article examines why symbolic AI architectures — those grounded in traceable IF-THEN rule systems — offer a more ethically sound and clinically trustworthy framework for transcultural nursing than generative or connectionist approaches. The argument advances three interconnected claims: first, that cultural blindness in AI constitutes a form of invisible discrimination with measurable public health consequences; second, that the apparent efficiency of generative AI in clinical settings is a false illusion that displaces rather than eliminates safety risks; and third, that nursing must emerge as the ethical guardian of AI in health systems. The concept of the AI Nurse Specialist, currently being proposed in the literature, is highlighted as a promising institutional response to this governance responsibility. This article contributes a nursing-centred perspective to an emerging global debate on the responsible integration of AI in culturally diverse health systems. 

Keywords: transcultural nursing; artificial intelligence transparency; cultural competence; clinical decision support; algorithmic bias; health equity 

Culturally Sensitive Artificial Intelligence in Nursing: A Transparency Imperative for Equitable and Safe Care 

Healthcare systems across Europe are navigating two simultaneous structural transformations. The first is demographic: international migration, labor mobility, and the rise of global health tourism have produced patient populations of unprecedented cultural diversity. In Portugal alone, the foreign resident population increased by 33.6% in 2023, surpassing one million individuals — a reality felt daily in hospital wards and private healthcare groups alike.^[1] The second transformation is technological: the integration of AI-based clinical decision support systems (CDSS) into clinical workflows is accelerating, driven by institutional pressures for efficiency and standardization. 

These two transformations are on a collision course. Current AI systems are overwhelmingly designed with a culturally homogeneous user and patient base in mind.^[2,3] When deployed in multicultural clinical environments, they risk not only failing culturally diverse patients but also actively amplifying the health inequalities they were meant to reduce. This is not a hypothetical concern — it is an evidenced risk with documented clinical consequences.^[4] 

This article takes the position that transparency and cultural sensitivity are non-negotiable conditions for responsible AI deployment in nursing. It presents a nursing-centered perspective on why symbolic AI — grounded in explicit, auditable rule structures — is the ethically and clinically superior framework for transcultural decision support, and why nursing professionals must lead the governance of these systems rather than merely receive them. 

The Cultural Gap in Clinical AI: Invisible Discrimination at Scale 

Leininger’s Culture Care Theory (CCT) establishes a foundational principle confirmed across four decades of transcultural nursing research: care that disregards cultural context is not neutral — it is harmful.^[5,6] The Sunrise Enabler, the CCT’s operational tool, maps the domains of cultural influence on care — religious practices, family structures, dietary norms, traditional medicine, communication styles — into a structured, clinically actionable framework. 

Current AI systems applied to nursing largely ignore these domains. A scoping review confirmed that existing AI tools present persistent inadequacies in both transparency and cultural adaptation.^[2] An integrative review of nursing care for migrant populations corroborated this finding, identifying the absence of structured cultural support tools at the point of care as a persistent and underaddressed obstacle.^[7] 

The consequences are measurable and clinically significant. Culturally unaware AI systems produce delayed access to care, lower therapeutic adherence, higher rates of adverse events, and systematic underassessment of pain in populations with culturally restrained symptom expression.^[8] The risks span several domains. In the area of religious dietary practice, prescription systems that do not confirm individual patient attributes may administer porcine-derived heparin to Muslim or Jewish patients without prior verification — a preventable safety event.^[7,9] Where ritual fasting is not assessed, such as during Ramadan, standard insulin regimens may precipitate severe hypoglycemia in patients who have not disclosed a change in eating patterns.^[9] Traditional medicine use presents a pharmacological dimension: documented interactions between Ginkgo biloba and antiplatelet agents carry hemorrhagic risk that standard CDSS fail to flag when cultural history is absent from the record.^[10] Culturally restrained pain expression — common across several populations — leads to systematic underscoring on numeric rating scales and consequent under-treatment.^[8] Finally, unassessed gender preferences in care can violate patient dignity and fracture the therapeutic relationship before clinical trust has been established.^[7] Taken together, these are not isolated incidents. They represent systematic, predictable failures that occur whenever cultural attributes are not individually confirmed at the point of care. At scale, this constitutes a structural form of discrimination — one that is invisible precisely because it is embedded in the architecture of the tools clinicians are trained to trust. 

The False Illusion of Efficiency: Why Generative AI Is Not the Answer 

The dominant narrative in healthcare AI discourse frames large language models (LLMs) and generative AI as efficiency solutions: systems that automatically infer context, generate recommendations rapidly, and reduce the manual burden on clinicians. This narrative is seductive — and in high-risk clinical settings involving cultural diversity, it is also dangerous. 

A critical counterargument lies in the nature of professional accountability: nurses retain an ethical obligation to validate, cross-reference, and critically appraise every output of an AI system, regardless of its source.^[11] In connectionist and generative architectures, this validation obligation is not eliminated — it is exponentially amplified. The opacity of LLM reasoning means that a clinician receiving a culturally contextualized recommendation cannot inspect the chain of logic that produced it. The risk of hallucination — plausible-sounding but factually or clinically incorrect outputs — is structurally inherent to these systems.^[12] Auditing an opaque output for hidden errors consumes more cognitive time and introduces more clinical risk than the time ostensibly saved by automatic inference. 

Furthermore, generative systems that infer cultural attributes from demographic proxies — such as nationality, surname, or country of origin — do not provide personalized care. They produce statistical stereotyping on a clinical scale. A system that automatically recommends Halal dietary accommodations to any patient with an Arabic surname, without individual confirmation, violates the foundational principle of Leininger’s theory and constitutes a structural mechanism for cultural discrimination, regardless of intent.^[13] 

The alternative — symbolic AI grounded in explicit IF-THEN rule structures — requires the nurse to manually and individually confirm each cultural attribute. This is not inefficiency; it is clinical rigor. The investment of time at admission guarantees that all subsequent inferences are traceable, hallucination-free, and genuinely individualized. Rule-based CDSS have been shown to be structurally more transparent, interpretable, and clinically trustworthy than deep learning alternatives.^[14] A meta-analysis of explainable AI in clinical settings further identifies interpretability as the single factor most consistently associated with clinician adoption of AI in practice.^[15] 

Transparency as a Public Health Determinant 

Transparency in AI is frequently discussed as a regulatory or technical requirement. This article argues that it must be reframed as a public health determinant in its own right. Opaque AI systems — whose decisional logic is inaccessible to clinicians and unexplained to patients — erode two foundational pillars of public health: community trust and clinical accountability. 

The EU AI Act (Regulation 2024/1689) classifies AI systems used in healthcare as high-risk, mandating architectural transparency (Article 13), effective human oversight (Article 14), and decision traceability (Article 12).^[16] These are not bureaucratic obligations — they are codified expressions of a fundamental ethical principle: that decisions affecting individuals' health must be explicable, challengeable, and humanly governed. Symbolic AI architectures satisfy these requirements natively. Connectionist systems require costly post-hoc instrumentation that approximates — but never fully achieves — the intrinsic interpretability of rule-based logic.^[12,17] 

From a community perspective, the consequences of opaque AI are far from abstract. Culturally diverse populations consistently report experiences of institutional invisibility and cultural disrespect in healthcare settings.^[8] When these populations encounter AI-mediated care without any visible mechanism for cultural recognition, the effect on trust — and consequently on care-seeking behavior, therapeutic adherence, and health outcomes — is clinically consequential. Transparency is not merely a compliance requirement; it is the mechanism by which AI systems earn their social legitimacy in diverse communities. 

Nursing as the Ethical Guardian of AI in Health Systems 

The integration of AI into clinical care cannot be a passive process for nursing. The concept of the AI Nurse Specialist has been proposed in the literature as a professional who would combine advanced clinical competencies with AI literacy to supervise, govern, and shape the responsible implementation of AI systems.^[18] This role is not yet formally recognized as an established specialization, but it is increasingly advocated in response to an urgent and growing professional need. 

Nursing is uniquely positioned to assume this governance role. As the profession that operates at the sustained interface between technical knowledge, clinical decision-making, and human relationships, nurses are the primary actors who experience both the cultural complexity of patient populations and the operational limitations of existing support tools. They are also the professionals most directly accountable for the cultural safety and individual dignity of every patient in their care. 

The proposed governance model for culturally sensitive AI in nursing should be distributed and multi-layered. In such a model, the AI Nurse Specialist would assume responsibility for curating and formalizing the knowledge base, translating validated evidence into auditable rule structures. Generalist nurses would provide operational feedback on knowledge gaps identified at the point of care. Community cultural experts would hold advisory and veto power over rules that risk stereotyping. An external audit body would verify ongoing compliance with the AI Act. 

This distributed governance model is not merely an institutional design choice — it is a structural safeguard against the risks of algorithmic bias, knowledge obsolescence, and uncritical AI adoption. Evidence demonstrates that bias mitigation is most effective when it occurs at the data input stage; the requirement for individual, manual confirmation of cultural attributes is the primary architectural mechanism for preventing stereotyping at its source.^[13] 

Implications for Practice and Policy 

The argument presented in this article carries concrete implications across clinical practice, professional education, and health system governance. 

At the clinical level, nurses and healthcare institutions should demand interpretability as a non-negotiable procurement criterion for any AI system used in direct patient care. A system whose decisional logic cannot be inspected and challenged by the nurse using it has no place in a culturally diverse clinical environment. The ability to understand, question, and override an AI recommendation is not a feature — it is an ethical baseline. 

At the educational level, AI literacy must become a core competency in nursing curricula. Structured educational interventions have demonstrated efficacy in improving cultural competence among nursing professionals.^[19] The same evidence base must now be extended to include the critical evaluation of AI tools, the identification of algorithmic bias, and the governance of knowledge-based systems. Future nurses should graduate prepared not merely to use AI, but to govern it. 

At the policy level, European health systems must resist institutional pressure to adopt the most commercially marketed AI solutions and instead develop procurement and governance frameworks aligned with the AI Act’s ethical imperatives. Consensus recommendations for minimizing bias in healthcare AI datasets — developed through international expert collaboration — offer a practical starting point for national and institutional policy development.^[20] 

Conclusion 

Artificial intelligence that does not see culture is clinically blind. Artificial intelligence that assumes cultural context without individual confirmation produces algorithmic stereotypes. The space between these two failures is where responsible AI in nursing must operate — and it is not a narrow space. It requires architectural intentionality, institutional governance, professional leadership, and a sustained commitment to transparency as an ethical practice rather than a regulatory checkbox. 

Nursing has both the professional mandate and the relational proximity to patients to lead this work. The prospect of the AI Nurse Specialist is not a distant archetype — it represents a potential next evolutionary step for a profession that has always placed the whole person, in their full cultural complexity, at the centre of care. The AI systems that nursing adopts must be built to honour that commitment. 

Ethical Statement 

This is a perspective article grounded in critical literature review and theoretical analysis. No empirical data were collected, no human participants were involved, and no clinical records were accessed. Ethics committee approval and informed consent were not required. The author declares no conflicts of interest of a financial, institutional, or personal nature. 

References 

1. Observatório das Migrações / AIMA. Relatório anual de indicadores de integração de imigrantes 2024. Lisbon: AIMA; 2024. Available from: https://www.om.acm.gov.pt 

2. Fernandes F, Shinners L, Mota M, Santos P, Sá L. Contributions of artificial intelligence to decision making in nursing: a scoping review. Nurs Health Sci. 2026;28(1):e70308. doi:10.1111/nhs.70308 

3. Marko JGO, Neagu CD, Anand PB. Examining inclusivity: AI and diverse populations in health. BMC Med Inform Decis Mak. 2025;25:57. doi:10.1186/s12911-025-02884-1 

4. Obermeyer Z, Powers B, Vogeli C, Mullainathan S. Dissecting racial bias in an algorithm used to manage the health of populations. Science. 2019;366(6464):447–453. doi:10.1126/science.aax2342 

5. Leininger MM. Leininger’s theory of nursing: cultural care diversity and universality. Nurs Sci Q. 1988;1(4):152–160. doi:10.1177/089431848800100408 

6. McFarland MR, Wehbe-Alamah HB. Leininger’s theory of culture care diversity and universality. J Transcult Nurs. 2019;30(6):540–557. doi:10.1177/1043659619867134 

7. Arias JMV, Peres AM, Bustamante CAC, Serrato JT, Rico FME. Cultural competence in nursing care for migrants: an integrative review. Int Nurs Rev. 2026;73(1):e70157. doi:10.1111/inr.70157 

8. Teixeira G, Picoito R, Gaspar F, Lucas P. Cultural competence and nursing work environment: impact on culturally congruent care in Portuguese multicultural healthcare units. Healthcare. 2024;12(23):2430. doi:10.3390/healthcare12232430 

9. Ibrahim M, Hassanein M, Akhtar S, et al. Recommendations for the management of diabetes during Ramadan: update 2025. Diabetes Metab Res Rev. 2025;e70057. doi:10.1002/dmrr.70057 

10. Herb-drug interactions in immigrant populations: systematic review. J Ethnopharmacol. 2023;310:116387. 

11. Nashwan A. The evolving role of nursing informatics in the era of AI. Int Nurs Rev. 2025. doi:10.1111/inr.13084 

12. Rudin C. Stop explaining black box machine learning models for high stakes decisions. Nat Mach Intell. 2019;1:206–215. doi:10.1038/s42256-019-0048-x 

13. Sasseville M, et al. Bias mitigation in primary health care AI models. JMIR Med Inform. 2025;27:e60269. doi:10.2196/60269 

14. Alnattah A, Jajroudi M, Manzari MN. Artificial intelligence in clinical decision-making: a scoping review of rule-based systems and their applications in medicine. Cureus. 2025;17(8):e91333. doi:10.7759/cureus.91333 

15. Giannopoulos P, Papadimitroulas P, Papageorgiou E, Kagadis GC. Explainable AI in clinical decision support systems: a meta-analysis. Healthcare. 2025;13(17):2154. doi:10.3390/healthcare13172154 

16. European Parliament & Council of the European Union. Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). Off J Eur Union. 2024. Available from: https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32024R1689 

17. Longo L. Explainable artificial intelligence (XAI) 2.0: concepts, taxonomies, opportunities and challenges. Inf Fusion. 2024;102:102095. doi:10.1016/j.inffus.2023.102095 

18. O’Connor S, Zhang M, Wicks D, Li X, Caton-Pete H, White P, et al. AI Nurse Specialists: transforming the nursing workforce for the age of AI. Int J Nurs Stud. 2026;105523. doi:10.1016/j.ijnurstu.2026.105523 

19. Osmancevic S, Zetterstrom Calling S, Ekblom-Bak E. The effectiveness of cultural competence interventions in nursing. Int J Nurs Stud. 2025;167:105079. doi:10.1016/j.ijnurstu.2025.105079 

20. STANDING Together Consortium. Consensus recommendations for minimising bias in healthcare datasets used for AI. Lancet Digit Health. 2024;6(12):e912–e933. doi:10.1016/S2589-7500(24)00224-3 

21. Anthropic. Claude (Version claude-sonnet-4-5) [large language model]. 2025. Available from: https://www.anthropic.com 

Author Note 

Cristina Maria Lima Pereira is a Registered Nurse with over 20 years of clinical practice (since 2004), holding a Bachelor’s Degree in Nursing from the School of Health Dr. Lopes Dias (ESALD), Polytechnic University of Castelo Branco (IPCB). She is currently a Master’s student in the thesis completion phase of the Master of Managing Digital Transformation in the Health Sector (ManagiDiTH, 2024–2026), an EU-funded international programme led by ISCTE — University Institute of Lisbon in consortium with Laurea University of Applied Sciences (Finland) and Aristotle University of Thessaloniki (Greece). She is also enrolled in the Postgraduate Programme in Health Innovation: Digital Technologies and Artificial Intelligence at the Escola Superior de Enfermagem de Coimbra (ESEnfC), University of Coimbra, and in the Specialisation Course in Artificial Intelligence in Healthcare — Development Profile at the Faculty of Medicine of the University of Coimbra (FMUC). This perspective article is an independent scholarly contribution grounded in a critical review of the literature. It does not present empirical data or results from any ongoing research. The author declares no conflicts of interest of a financial, institutional, or personal nature. 

AI Use Disclosure: The AI tool Claude^[21] was used exclusively to support translation from Portuguese to English and grammatical review. All intellectual content, arguments, literature selection, clinical reasoning, and conclusions are the sole responsibility of the author. The use of AI assistance was strictly limited to linguistic support and did not involve the generation of ideas, claims, or interpretations. 

Correspondence regarding this article should be addressed to Cristina Maria Lima Pereira, Viana do Castelo, Portugal.